Why you can no longer spot a poker bot at the table

Why human poker players can no longer visually detect AI poker bots at the table in 2026 — detection has moved to the room's security server, poker-ai.org

TL;DR. Throw out the guides. Seriously, delete them. The ones about “identical timing” or “silence in chat” are dead weight. It’s 2026. You sit at a table and you cannot tell a top-tier bot from a live player. Not by eye. Not anymore. Modern engines randomize their pauses — and not dumb randomization, the pauses scale with how complex the spot is. They mimic reg bet sizings. They answer in chat, and the answer comes either from an LLM or from a live operator. Does that make bots invulnerable? No. The catching happens on the room’s server now, away from your table. And not every room does this equally well. Some do it properly. Some — we’ll get there.

Every couple of weeks I get the same DM, different wording: “look, I spotted a bot — all his pauses are identical.” Or: “caught one, didn’t respond to my chat jab.”

I’ve spent the last few years working on the architecture of detection systems, not on playing strategy, so every time I end up explaining the same thing. The heuristics that worked around 2018? Dead. I wrote about this back in “Ghosts at the Table” — and since then it only got worse.

WHAT’S WRONG WITH THE BEHAVIORAL-TELL GUIDES

Everybody links the same guide. The poker bot detection guide from Upswing Poker, August 2021 — probably the most cited one in the industry. Four working signals it gives you: consistent timing, no reaction to a moderator alert, sessions of 12+ hours, 25+ tables at once. (Bet sizing and silence in chat — to their credit, they admit those are unreliable.) Fine. Except five years passed since 2021. And in 2026 not one of those four delivers what it promised. Not one.

Timing first. People imagine old bots clicked at exactly 2.4 seconds on every decision. Nonsense. Even the OpenHoldem generation, the early Shanky Poker Bot stuff — randomization was already there. Crude, yes, identical for every situation, but there. Aggregate statistics across volume were what actually busted those bots. No hero with a stopwatch on a single hand. Today the range is context-aware. Garbage preflop hand — instant fold. Tough river all-in — nine seconds of “thinking.” They even fake tilt: loses a big pot, snap-calls right after. Like a human steaming. Cute.

The moderator test doesn’t work anymore either. Direct message with a notification — a modern bot answers just as convincingly as a human. The difference is the cost of the solution. Cheap option: an LLM answering autonomously, no human in the loop, and it slips on stuff unrelated to the game — recent news, local language nuances, the model starts to drift. Expensive option, the one serious farms use: a sniffer catches the system notification and forwards it instantly to an operator in Telegram. Guy replies from his phone. Good luck jab-testing that.

The 14-hour marathon was too obvious a tell on its own. So farms dropped it. Modern table-selection code hunts weak lineups by itself, and sessions cut off on a time cap or a win-rate threshold.

Same story with 25+ tables. Today that’s standard practice for professional mass-multitablers. Humans. The sheer table count gives a bot away about as much as wearing a hoodie does.

WHAT ACTUALLY HIDES THE BOT FROM THE ROOM

First, one question needs clearing up, otherwise it gets in the way: why does the anti-fraud system even let the bot in?

Here’s the thing. The technical perimeter isn’t aimed at you. It’s built to defeat the room’s anti-fraud system, and it stays active the entire time the game is running — at login and through the whole session. The standard setup is an Android emulator running a modified APK client of the room: a sniffer intercepts the game packets — cards, stacks, positions — before they’re even rendered on screen, forwards them to a remote decision server, and then the emulator taps the screen with the ready-made action. That’s the working mechanism. Nothing magical. And what hides it? Residential proxies instead of a data-center IP, plus FakeGPS with a realistic margin of error. To the room this player looks like an ordinary student on a smartphone in Lisbon. Not a VM in a data center. A student.

“LIQUIDITY BOTS”: WHY CLUB OWNERS RUN A BOT AT ZERO EV

Now an argument against manual detection that almost nobody brings up. Bots deliberately not programmed to win.

In WePoker, ClubGG, PPPoker there’s a practice called “liquidity bots.” They’re tuned to play at zero or a slight loss — 0 or −2 bb/100. Beating the regs isn’t the job. Keeping the illusion of a live game running 24/7 and generating rake for the club owner is. These bots make “human” mistakes. They toss emoji into chat. Tell me how you distinguish that script from a tipsy fish at the table. You can’t. There’s no way. The bot isn’t trying to play optimally. Optimal play is exactly what would give it away.

And here’s the part that stings. If you consistently run good against weak opponents at limits where the room earns on rake, don’t rush to thank luck. The bot is losing on purpose — but every pot you drag pays rake to the house. The club’s economy is fine. You’re the customer. You aren’t the beneficiary.

HOW BOTS GET CAUGHT

So if a person can’t spot a bot with their eyes — bots are invulnerable? No. They just get caught in places an ordinary player has no access to.

What happens on the room’s server

At the table you see a hundred hands against one suspicious opponent. The room sees every hand that player ever played, on every table. Plus IP, device telemetry, links to other accounts. Different game entirely.

Rooms themselves rarely disclose the details of their systems, so we go by their contractors. GeoComply, for instance — builds anti-fraud systems for rooms. It checks three things at once. First, what’s happening on the device itself: foreign code hooked into the game client, a solver renamed to look like a system utility, whether the whole thing is an emulator under the hood. Second, whether the physical location is real — it checks more than IP, including what GeoComply calls “impossible movement patterns.” An account was on GPS in Lisbon a minute ago and is now writing from a mobile network in Manila — that’s not a human, that’s a plane that doesn’t exist. Third, links between accounts: identical device fingerprints, a shared Wi-Fi point across supposedly different players. By their own description, that’s how they once caught 2,000 “unique” accounts physically sitting in a single location. Two thousand.

Of the rooms themselves, only partypoker has publicly disclosed the mechanics — on their Game Integrity team’s blog. Two mechanisms. The first is whole-pool comparison: the system matches every player against every other and calculates how similar their styles are; reports are generated automatically. The second is fingerprints. Once a bot has been identified, its profile is captured from the stats, and from then on the system fires a signal in real time as soon as anyone’s stats match it. Then a live analyst takes the account from there.

PokerStars keeps a dedicated Game Integrity team — data scientists, analysts, former pros — and states that 95% of caught bots are found by the team itself, with only 5% coming from player reports. Think about that next time you fire off an angry email to support. Their procedure is a cross-check: one analyst makes a call, a second one independently makes theirs. They disagree — a third joins in and they decide together. Why the paranoia? Because the actions are irreversible. Fund seizure. Insurance against a single analyst’s error, that’s what this is.

GGPoker, via its Poker Integrity Council, banned 42 accounts in a single 2026 investigation and seized $1.2M from those playing with AI. Partypoker’s numbers are more modest but published steadily, year after year: 291 accounts and $71,771 returned to players in 2024, and since 2018 — more than 2,540 accounts and over $2M.

And none of this is a fresh idea, by the way. Back in 2015, a master’s thesis at the University of Ottawa (“Hiding Behind Cards: Identifying Bots and Humans in Online Poker”) summarized earlier results on detection through server logs, timing, and network traffic — earlier work reported around 90% accuracy from traffic analysis and around 95% from decision logs. Catching a bot with statistics at scale isn’t new. Engineers have been systematically grinding this problem for more than a decade.

The same logic works against farms — only there, the target is coordination between accounts rather than a single bot. A 2022 paper catches that coordination with the Isolation Forest algorithm. Without labels. Without a pre-defined image of collusion. Just statistical outliers. Tested on 170,000+ players and 100,000+ matches.

An insider’s leak

Sometimes it isn’t a system that catches the bot. It’s chance. That’s what happened with Martin Zamani. On January 17, 2026, the poker pro published a 36-second video on X: rows of computers playing tables on Ignition and Bovada, not a single human at any keyboard. Zamani didn’t shoot the video — he got it from someone on the outside and posted it. Ignition claimed the footage was old, from 2022, and that all the accounts had long been closed. Zamani disagreed and showed a screenshot from the same source, this time from a 2024 tournament. Same farm.

Who’s right? From open data, you can’t tell. And that’s the point. Leaks aren’t a detection method. They’re luck plus somebody’s insider access, and even then they don’t give a definitive answer. Only grounds for an argument.

NOT ALL BOTS AND NOT ALL ROOMS ARE EQUAL

Behind all this engineering it’s easy to miss something simple: the majority of bots on the market aren’t the polished setup I described. Not even close. Building an undetectable system is expensive — development, infrastructure, keeping the thing alive through every room-client update. A significant share of bots would get caught on careful review. That’s my assessment based on such reviews — not hard numbers, I won’t pretend otherwise, but the kind of pattern you learn to see after enough of them. Either the mechanism through which the bot talks to the poker client lets them down. Or the operator does: broken schedule, an off-topic chat reply, a dropped proxy. Something always slips.

And the heavy tooling — style comparison, device telemetry — that’s mostly a thing at large public rooms. In closed club apps like WePoker, ClubGG, and PPPoker it’s different: the platform provides some tools, IP and geolocation restrictions and the like, but tracking games is up to the club itself. No unified system at the app level. Every club sorts it out on its own. And it’s not just a tools problem — the decision stays with the club owner, and he can turn a blind eye to a suspicious player as long as they bring in rake reliably. Server-side anti-fraud neural nets and cluster analysis? Only at the biggest rooms. The rest get by on player complaints and manual hand history reviews. Detection doesn’t disappear because of this. It’s just uneven. Strong in some places, nearly absent in others.

What the club owner is actually left with

If you play in the club apps, this section matters to you too — how the club owner handles this problem is what determines the quality of the pool you’re sitting in.

Here’s the honest picture: club owners aren’t looking for an absolutely bot-free room. That isn’t the goal. Often not even desirable. Bots — including a share of their own liquidity bots — mean around-the-clock action and traffic, which brings in real players and generates rake. The real threat is somebody else’s farm parking on the club without permission and quietly bleeding the regs dry. When that happens, the regs eventually notice, feel like they’re being farmed (because they are), and leave. Loyal regs gone — the recreational pool follows, and the rake dies. So what the club actually needs is control over external bot farms, with its own liquidity setup staying intact. Sterility isn’t the point.

The problem: a modern bot from an external farm mimics a live reg well enough that the club dashboard doesn’t distinguish it. Sizings normal. Timing looks emotional. Chat alive. Stats sit within the pool. By the built-in metrics — everything looks clean.

Which leaves the owner with two real options. Two.

The first is manual verification of suspicious players: ID request, gameplay video, a live voice check. It works. Technically. But using it on a real pool is problematic — every such check is friction with a player who may not turn out to be a bot at all, and a string of checks scares off loyal regs. And it’s the regs who bring in the rake. You see the trap.

The second is external analytics built specifically for the club’s job. Effectively the club needs its own miniature Game Integrity team. Building one from scratch is unrealistic, so the practical route is to rent that function as a service. That’s the model behind the poker bot detection service at pokerbot.com (I’m part of the team that builds it — judge accordingly): statistical analysis of the whole pool over long distance, hunting for teamplay patterns between accounts (joint play, coordinated lines of collusion), plus scoring based on the combined weight of statistically significant signals. What the club owner gets goes beyond a single “this player is suspicious” flag. It’s the whole pool annotated with review priorities — precisely the gap that internal platform tools can’t fill.

One honest caveat about this class of solution, because I keep my credibility by saying it out loud: statistical analysis catches external farms, coordinated groups, and profit-oriented bots because they leave a fingerprint at scale. It does not catch liquidity bots the club runs on itself. That’s a feature, not a bug — liquidity bots are, by design, statistically indistinguishable from a tipsy fish, and detection isn’t the layer where that decision belongs. That’s a business call for the owner. It isn’t a signal for a model.

What does all this mean for the ordinary reg? You may still notice something suspicious at the table, but reliably catching a top-tier bot with your eyes is out in 2026 — attention at the table is better spent on EV calculations, ranges, and the real weaknesses of the opponent. Suspicions are serious? Collect hand histories and file a report with the room’s or club’s support: security engineers have access to logs, telemetry, and cluster analysis you don’t. The hunt for the invisible in 2026 happens on the server, away from the table — and there, at best, you’re a signal source. Not the hunter.

Cognitive Biases at the Poker Table – Part 3: Rituals and Lucky Wins

Continuing this series of articles on cognitive biases in poker, I’d like to start—as is my custom—by explaining why it’s important to understand cognitive biases. After the first few parts, I hope you’ve come to realize that:

it’s impossible to turn off the inner monkey—the intuitive channel;
in certain situations, we really, really need the monkey.

System 1—or the intuitive decision-making system—is something we really, really need, and it’s precisely what distinguishes a professional from an amateur.

Imagine a professional boxer in the ring who, when his opponent’s gloved fist is flying toward him at tremendous speed, starts thinking: “I’ll dodge to the right now, then to the left, then I’ll duck, then maybe a double jab…”—and so on. If he thinks about that, he’ll be knocked out immediately—any athlete who’s ever boxed will tell you that. During training, reflexes are developed so that during a fight, the boxer thinks only strategically and positionally. At the moment of a punch or an attack, as a rule, he isn’t thinking about where to dodge, where to tilt his head, at what angle, how low to crouch, or where to extend his elbow… All of this happens automatically. And that’s exactly what our “inner monkey” does. It works quickly, without thinking—exactly what’s needed during a fight.

That much is more or less clear, but the question arises: do we need System 1 in other areas unrelated to motor coordination? Maybe we should turn it off in all other cases? Well, no!

Take a look, for example, at a game between top-level chess players. What distinguishes a grandmaster from a master, a master from a club player, and a club player from a beginner? You probably know that in chess there’s a clock with a flag. If the flag falls, time is up—which means you’ve lost. A certain amount of time is allotted for each game. The entire game depends on the speed with which the player finds strong moves. Does that mean they think better (faster) than an ordinary person—someone who isn’t a chess player? No.

Strange as it may seem, for a grandmaster, a titled master, or any trained chess player, all these moves come almost automatically; only afterward, with the help of System 2—our rational mind—does the player reject some, accept others, and decide which move to make. But there are a huge number of different moves that can be made on the board. How do you choose the optimal one from this multitude? The fact is that a chess player begins by considering the most important moves, and that is precisely what saves them time. That’s what the monkey does. And this happens not only in chess. As you might have guessed, it happens in poker, too.

System 1 (the inner monkey) is capable of making the right decisions only when a person has a certain amount of experience. And the more experience a person has, the more the energy-intensive and time-consuming System 2 delegates tasks to System 1. So I’ll repeat what I’ve said in previous articles. If you think you’ll find a “magic pill” in this text that will suddenly let you beat a poker bot—or, for that matter, defeat anyone at the poker table—you’re mistaken. My main point is this: to win, you need to make fewer mistakes, and to make fewer mistakes, we’re studying the key cognitive biases that apply to poker. In fact, I’m slightly shortening your path toward reducing systematic thinking errors when playing poker.

Let’s get to the point. Earlier, we looked at just two—but quite important—cognitive biases applicable to poker:

1. Gambler’s Fallacy

When you think that past events somehow influence the probability of future events.

2. Confirmation Bias

When, instead of accepting evidence that contradicts our established views, we seek only confirmation of our views while ignoring contradictory evidence.

Let’s examine a few more systematic cognitive biases related to our topic. Before we begin exploring anything new, please note that you should also study these biases in other sources of information. I also want to emphasize that I’m only discussing the cognitive biases I’ve personally encountered and was able to identify while analyzing my own thinking at the poker table. Who knows—maybe this knowledge won’t apply to you. That’s exactly why I recommend reading other articles and publications on this topic.

3. Illusion of Control

Do you tend to believe in superstitions? You probably knock on wood to ward off bad luck or keep good fortune from leaving. And if you spill salt, it’s said to lead to a quarrel. To counteract this, you toss a pinch over your left shoulder.

When it comes to poker, this cognitive bias occurs when you think you can influence a random event through your actions and rituals, even though the outcome is actually completely random.

Rituals can be absolutely any action you perform while playing poker that you believe will bring you luck—even if the failure rate is many times higher.

Some people tap the table a few times before making an important decision or switch the hand they use to bet. The most dangerous thing in this case is when your brain receives random positive reinforcement for your illusion of control.

Well, there’s really nothing more to say here: just next time, if you catch yourself performing some kind of ritual, you need to stop yourself from thinking that you can somehow influence the hand. Because that’s not the case.

4. Winner’s Curse / Survivorship Bias

Let’s take a closer look at this.

This cognitive bias is also known as “survivorship bias.” I’ll provide the most appropriate definition below.

Survivorship bias is a type of sampling bias that occurs when a person makes a decision based solely on examples of “survivors” (those who succeeded) but fails to consider statistics on “casualties” (those who did not achieve the same result), because data on the latter is scarce or nonexistent. People are subject to this cognitive bias in various situations: when making decisions in everyday life, in financial planning, in scientific research, in business, and, of course, in poker.

This cognitive bias was first identified during World War II by mathematician Abraham Wald. The U.S. Army was losing a huge number of aircraft. To improve the survivability of combat aircraft, the command tasked Wald with strengthening the aircraft’s structure while adhering to weight requirements. In other words, it was not possible to fully reinforce the aircraft’s armor.

Military engineers examined the damage to aircraft that had returned from the battlefield and assumed that the most heavily damaged parts needed to be reinforced. However, Abraham Wald pointed out that if an aircraft was able to return to base, then the hits to those parts were not critical. It was more important to focus on the damage that prevented aircraft from returning. In other words, armor should be added to the areas where the “surviving” aircraft had no holes. This helped reduce American aviation losses and save the lives of many pilots.

Cognitive Biases in Poker

Now, let’s move on to poker. In poker, this bias manifests very often—even among regs. The standard way to combat this bias is to analyze your decisions after receiving this information and “slap yourself on the wrist” the next time you try to give in to your inner monkey. After such practice, the monkey will give you more accurate cues.

Only successful hands are remembered

Let me give an example from my own playing career. Once, at a tournament, I was playing absolutely recklessly. My VPIP was around 80, and I was playing very aggressively. I frequently went all-in on bluffs; a couple of times, that bet was called, and by some miracle, I made it to the final table and won the tournament. And what was the result? System 1 “remembered” this tactic as a winning one; many tournaments later, in my mind, it was still considered a winning strategy. It was probably after the tenth tournament—when I was eliminated almost immediately—that I realized this style of play wasn’t winning at all, and my victory had been a fluke.

A player remembers that single instance of success but completely ignores the dozens of times he did the same thing and lost all his chips. Only the “surviving” successful episode remains in his memory.

Imitating Someone Else’s Playing Style

Probably every reasonably good poker player works on correcting their mistakes, which includes watching stronger players and analyzing their style. In doing so, they might notice that a poker champion with junk cards in hand makes a super-aggressive bet and takes down a huge pot. System 1 also “bookmarks” this for itself, without even paying attention to the fact that 100 other hands played in this style would have been losing hands.

I’ll also include taking the advice of “seasoned” players at face value under this subpoint. Every strong player has their own well-honed System 1. It’s capable of producing the right moves in a short period of time. This is achieved through extensive experience and many hours at the table. But few professionals even realize that they have an “inner monkey” capable of providing them with the right decisions. And these players may believe that some particular style or ritual is what made them champions. And if they share their insights with beginners, it can backfire. After all, we now know that the whole secret lies not in rituals or specific tactics, but in the well-trained System 1 of experienced players.

Now you know two more systematic errors. I strongly recommend looking into them in other sources; you can ask ChatGPT or Claude for a few more examples. Once you have this important information, all that’s left is to observe yourself and your thought processes at the poker table (and, by the way, in everyday life as well). And you’ll soon spot these two new systemic errors during one of your hands. Forewarned is forearmed.

Cognitive Biases at the Poker Table

When I wrote the previous article, my goal was to keep it within a certain length so as not to overwhelm the reader with too much information. I wanted to write a comprehensive article that fully explored the topic, but as I wrote, I realized that the subject was quite interesting, and as I delved deeper and wrote the first article, I realized that systematic errors of perception (or cognitive biases) is a very broad topic. So I decided to write a series of short articles on making incorrect decisions related to this topic, using poker as an example.

In fact, if we compare modern humans to computers, and mental patterns and ways of thinking to an operating system, the following analogy best explains what is happening. The core of a modern person’s operating system is identical to that of an ancient person’s. In modern people, only the system utilities, libraries, and user applications in their minds have changed, been updated, or been rolled out. In other words, we have a program embedded within us that was formed long ago. And, unfortunately, we’ve also inherited the bugs in this software, which is what we’re talking about. What’s most interesting is that, even though we’re all different, these cognitive traps are passed down to us, and we don’t even suspect it. But, as they say, forewarned is forearmed.

As I mentioned, we have System 1 (which I will sometimes call the “inner monkey”), whose job is to conserve mental resources and make quick decisions when necessary, drawing on life experience and the “ancient program” within us. And the cumbersome System 2, which is very labor-intensive, kicks in when we’re really thinking something through, in situations where the intuitive and fast System 1 can’t handle the task. It’s as if we’re switching from autopilot to manual control. So what exactly is the main point we need to clearly grasp for ourselves? The “ancient program” is designed to conserve energy, so your brain will try to make decisions using the inner monkey rather than System 2. And the inner monkey really, really loves to make mistakes. In this situation, I can’t help but give another example of how System 1 makes the wrong decisions. Here’s a simple picture. Which line is longer?

Cognitive Biases

System 1—or as I call it, the “inner monkey”—gives the answer: the top line is shorter than the bottom one. Now scroll to the very end of this article and see for yourself that System 1 gave the wrong answer. In fact, the lines are the same length.

It’s worth noting that no matter how hard you try, it is impossible to turn off System 1. Try looking for 5–10 seconds, for example, at a store sign in a language you know. Did it work? =)

There is also an effect known as the Stroop Effect, which was allegedly used during the Cold War to identify Soviet spies. A person suspected of espionage was shown a text in Russian; the text consisted of a word denoting a color, for example, “Yellow,” but the text itself was printed in a different color. The person had to name the color in which the word was printed, not the color the word denoted. Try practicing.

Stroop Effect

Not an easy task at all, is it? Well, for the average American, Russian text is nothing more than a bunch of squiggles. Thus, any American could easily name the color of the text. A Soviet spy, however, who knew Russian, already faced difficulties. Using this simple example, you can observe for yourself the difficulties the brain experiences when switching from System 1 to System 2. System 1 sees the text and is ready to give an answer, but you realize—the task has a catch—so you have to switch to System 2, and that’s when dissonance occurs.

Let’s consider and note the following. Absolutely everyone is susceptible to cognitive biases. You, me, and even the founder of behavioral economics, Daniel Kahneman, who wrote a monumental work on cognitive biases. People are prone to cognitive biases because their brains are wired exactly the same way they were 200, 300, 400, and 5,000 years ago. Understanding how the brain works will help you build what’s called “immunity” to bad decisions at the poker table.

Now, I think we’re ready to look at some of the thinking traps or blind spots that arise when making decisions at the poker table. IMPORTANT to understand! One of the main differences between a bot and us is that it isn’t prone to cognitive biases; it doesn’t make the mistakes that meatbags might make.

Gambler’s Fallacy

The essence of this bias is that we believe past events somehow influence future ones. It’s hardwired into our ancient programming to look for patterns. That’s why we believe this.

For example, during an MTT tournament, you’re dealt pocket aces for the third time. The previous two times, with two aces in your hand, you played aggressively and lost almost your entire stack over those two hands. At that moment, your brain thinks, “I have the strongest hand preflop; the last two times I lost with the same hand, so the third time I’m bound to get lucky.”

In reality, however, a specific hand has absolutely nothing to do with the previous ones!!! Let me try to explain it another way. The Gambler’s Fallacy is when the brain believes that if a random event has occurred more often than usual, it will happen less often in the future (or vice versa).

Let me give an example with roulette at a casino. If red has come up 5 times in a row on the roulette wheel, that doesn’t mean you should bet all your chips on black. The chance of red is ALWAYS 47.4% on American roulette (the two green zeros take the rest).

“I haven’t gotten any aces in 20 hands in a row—so they’re definitely coming soon” is far from a fact.

Therefore, it’s very helpful to remind yourself that hands have no memory at all. And every event in poker is in no way connected to the previous one.

Confirmation Bias

A cognitive bias that favors one’s own worldview. As I mentioned earlier, our brain is very lazy. And if you hold certain beliefs, it’s easier for the brain to seek out and collect evidence supporting those beliefs than to accept evidence to the contrary and reconsider its views. I think this is one of the most costly mistakes in poker. I often catch myself falling into this bias when I play poker.

A player makes a decision (often the wrong one), and then selectively remembers only the instances when it worked, while forgetting or downplaying the instances when it failed. Here are a few examples.

After studying this cognitive bias, I noticed two distinct patterns in my own perception at the poker table. I considered card sequences like 3/4, A/2 (especially), or 8/9 to be very strong combinations, since you can build an excellent straight. And can you believe it, I only remembered the hands where I actually made a straight, while my brain simply ignored the other 95% of hands where I didn’t. I had the same thing with suited cards; I always considered them strong because it was so easy to make a flush. But now we know that’s not the case.

Confirmation bias is especially dangerous because the brain doesn’t like to admit its failures, while protecting your ego.

Here’s another example. I was playing a cash game against a typical nit — never bluffs, rarely enters a pot. I was so bored that I remember this instance very clearly. Suddenly, I noticed that my opponent had stolen the blinds twice and made a couple of raises that allowed him to take the pot without a showdown. From that moment on, my brain labeled this player as a major bluffer. And a couple of hands later, I bluffed on the river, he went all-in, we showed our cards, and he took my entire stack with a hand that was much stronger than mine.

In this article, I’ve cited only two examples of cognitive biases. I did this intentionally so that your brain can process the information better, bit by bit. I recommend studying these two cognitive biases further by looking them up online. And, of course, practice and applying this new knowledge are essential. Now, when you play poker, try to identify your own instances of the Gambler’s Fallacy and Confirmation Bias. And I’m sure you’ll find them.

And as promised, at the end of the article, I’ve included proof that the segments are identical:

Cognitive Biases Solved

One last thing, since I get asked about it often. Most of what I write here is from the player’s side — biases, decisions, the human stuff. But the actual work I’ve done for years is on the operator side: opponent modeling, behavioral calibration, and detection systems built to defend private-club ecosystems from external farms. It’s a completely different industry from the loud consumer “buy a bot for $100” market — quieter, older, and almost nobody writes about it openly. The place that finally does is pokerbot.com, a project I’ve been involved with. Worth a look if that side of the industry is closer to your world.

How to Play Against a Poker Bot

Probably every professional poker player—whom we commonly refer to as “regulars”—understands that poker isn’t just about luck. Poker is about probability, statistics, cold calculation, and emotional detachment. Emotions are an essential part of being human. On one hand, they’re a vital component of life that helps us survive and make quick decisions when necessary. You wouldn’t stop to think about what to do if a furious man with a knife suddenly charged at you on the street, would you? If you haven’t trained yourself for such a situation, you won’t even have time to think—a mix of fear and panic will make you turn and run. According to Nobel laureate Daniel Kahneman, the decision was made for you by what’s known as System 1. System 1 helps you pull your hand away if you touch a hot object; when you brush your teeth, zip up your favorite jacket, or play your favorite chords on the guitar—you don’t think about it or focus on the process.

In the other corner of the ring is System 2. Its operation requires a great deal of resources. And the human brain is wired such that, for the sake of conservation—in the context of evolutionary survival—not everyone likes to engage System 2. For example, when you’re learning to code, figuring out the rules of a new board game that the weird guy with greasy hair (but who, for some reason, is friends with your best friend) brought to the party, or watching a documentary about how a nuclear reactor works—System 2 is at work. System 2 is slow, conscious, analytical, and resource-intensive. It kicks in when you need to solve a complex problem, check logic, or suppress an impulse from System 1. Here’s a simple example illustrating when System 2 is active and when System 1 is active, using basic mathematical operations:

2 x 2
48 x 23

In the first case, you simply need to glance at this simple mathematical equation to come up with the answer 5 (just kidding, 4). In the second case, you need to break the equation down into its simplest components in your head. First, you’ll mentally multiply 48 x 20 separately, and then add the result to the result of the expression 48 x 3. It’s complicated, isn’t it? And if you were chewing gum or listening to music in the background at the same time, you’d stop chewing and wouldn’t be able to make out the sounds of the music.

Unfortunately, salespeople and marketers have adopted Kahneman’s work for manipulative purposes. The thing is, System 1 very often fails. Unscrupulous salespeople take advantage of this. Here are a few examples. The first price you see becomes the anchor against which you’ll evaluate the purchase. Displaying an old or inflated price next to the new one makes the discount seem very tempting. An expensive item in the window makes all the other prices seem more affordable. Or, for example, let me share a story from my own life. A long time ago, about 10 years back, when my belly was smaller, the grass was greener, and ice cream tasted better, I got my driver’s license and saved up for a car for about two years. So there I was at the dealership, on a friend’s recommendation, to buy a brand-new car, but one registered from the previous year. It was technically the end of the year, and in a couple of weeks, that car would be two years old on paper. That’s exactly why the dealership was offering a great discount. While I was finalizing the paperwork and drafting the sales contract, my personal sales representative’s colleague walked up to him and asked, loudly and pointedly: “Are you selling that last black Honda Civic right now? Because I already have customers here to buy it.” My first reaction was panic, and five minutes later I had already put a deposit down at the dealership’s cashier. A week later, when the car transporter delivered my car, the price had naturally gone up, and this purchase went from being a super-profitable deal to just a regular one. But what if I’d had time to read the deposit agreement and had time to make a decision? Most likely, I would have looked at offers from other dealerships and, quite possibly, bought it somewhere else.

Let’s gradually move on to poker. What happened to me at that car dealership 10 years ago? I let my emotions get the better of me; the salespeople—marketers—had me wrapped around their little finger. A small trick, all within the law, but it felt as though I’d been cheated (in fact, I still consider it a scam). Now, let’s try to apply or draw an analogy between this real-life situation and situations that often arise at the poker table. What exactly are these situations? They are situations where, under stress, our brain must make decisions under time pressure. In other words, only two factors are needed for a systematic error in decision-making. It turns out the formula for a wrong decision is quite simple:

Limited time + emotional component = wrong decision

Just a moment ago, I lost an online poker tournament organized by the company I work for, using the simple formula described above. The player sitting across from me—to whom I had lost a significant portion of my stack just two hands earlier—simply called my all-in bluff. And as a result, I lost and was eliminated from the tournament. I’m not claiming my opponent is a bot (though who knows). I simply made the wrong decision based on my emotions in a split second. I’ll attach the hand history below.

MTT NLH Sucks

And there are a huge number of such mistakes—often referred to as cognitive biases—in situations relevant to poker. So, let’s get down to business, since the article is titled “How to Play Against a Poker Bot.” And I’m sure you’re hoping for a “magic pill” or a simple guide on how to play against a poker bot in order to beat it. The answer is quite simple and straightforward: don’t make mistakes. It’s enough to simply reduce the number of mistakes to increase your win rate. The advantage of poker bots is that they don’t care at all who they’re playing against. It doesn’t matter to them that they lost half their stack two hands ago. A poker bot is simply a mathematical model capable of calculating probabilities and backed by statistical data. Over the short term, an average poker amateur, with a bit of luck, can beat a poker bot, but over the long term, the poker bot has a massive advantage.

Well, actually, do chess grandmasters really engage on System 2 to analyze moves? The answer is no. The secret to success lies in constant practice, analyzing games, moves, and situations. Thanks to this, System 1 provides professional chess players with a more accurate decision than it does for beginners.

In other words, practice to train System 1, along with an understanding of cognitive biases and the ability to apply that knowledge at the poker table, will help you make the right decisions. That’s when chance comes into play—what many call luck.

3UpGaming Bot Review: $1,800 for a “Pro” AI That Plays Like a Fish

TL;DR

A buddy of mine decided to try out a new player on the poker bot market — a “real AI bot running on poker AI + GTO that plays like top regs” — and paid $1,800 for it. Everything seemed serious and trustworthy, however, after a series of test runs (hand histories were scraped from all sessions + video recording was kept), non-stop patching of junky bugs from 3Up Gaming, we received a bot that plays worse than any fish and simply flushes your stack down the toilet — overall EV winrate: -87bb/100 hands and worse over a volume of ~10k+ hands.

I can say with confidence that this is just another well-hyped scam product on the poker bot market. 3UpGaming’s goal is to sell you the bot immediately (2-3+ units for the max price) which looks like a working product, but in reality, is just an imitation. You will only realize this after a couple of days of testing, while the 3UP manager keeps gaslighting you along with promises of updates. Stay away.

I know 3UpGaming managers might try to claim this review is fake. If anyone—potential buyers or developers—wants to see the unedited video recordings or the full database of hands, email me at [email protected]. I have nothing to hide.

The Red Flags

I rarely do these breakdowns of other people’s products, but I feel obligated to warn anyone wanting to buy a poker bot in 2026 about this scam. My friend fell for it (and he is far from a fool), so the chance of throwing a couple of thousand dollars into the void is high. Let’s get to the point.

The 3UpGaming project appeared in 2021. In 2024, judging by the Web Archive, they offered iGaming services — you buy a ready-made poker app that is customized for your brand (so-called poker skins). The contact info listed was American — phone number, email, Skype. Okay, you can find a presentation of their services on the internet, I have no questions there.

In mid-2025, global changes occur — the site is significantly reworked, the CMS changes, the text changes, new sections and services are added. It’s visible that a good SEO specialist took over the site, and copywriters are writing mountains of AI-generated articles for them weekly (processed via humanizer services, but after reading them it’s clear it’s all fluff). The site is actively promoted, and backlinks are being bought. Services for poker AI robots (bots) are added, as well as a service for club owners to fill tables with their AI bots.

That’s when my friend saw them on the internet, studied the site, and messaged Alex from 3UP Gaming on Telegram. So, here is what they promise:

“fully automated bots that are a mix of GTO and AI”
“They play exactly like a professional player”
“2–3 bots sit at a table, see each other’s cards and play team play”
“for each robot we give you a VM machine and a unique IP so that the percentage of robots being banned reaches zero.”
“Robots are trained and updated by AI every day.”

Setup of the bot for any poker room and poker format (Hold’em, Omaha).

And all this — for $1,200 setup + $600/month for each bot. Alex sent a few demo videos where their bots are playing on ClubGG.

Sounds not bad, right? But upon requests to show any HH (Hand History) graphs, game results, or successful cases, we get the answer:

“Most companies create proof for customers and you don’t understand whether this proof is for the robot or the player. The best proof for you is to test the robot and use it.”

When we asked for a trial:

“Unfortunately we only have 1 hour test. Our robots cost a lot to configure. We can’t prepare anything without money. We are not responsible for your bad experience. Please think carefully and then message me to rent robots.”

The maximum they can offer is a 1-hour test where we can watch their bots play against each other on ClubGG. And of course, they are “not responsible for your bad experience.” Doesn’t sound great, but given the specifics of the poker bot market, and considering they are newcomers — this is somewhat acceptable.

Anyway, the money was paid, and 5 days later pdf-instructions were issued for connecting to the VM via Web-VNC and launching the bot.

How we tested the AI “robot” from 3UP Gaming: Methodology

Room: PPPoker
Limits: NL4 cash games, 6-9-max tables

At the tables — a classic zoo of micro-stakes players:

Newbie fish
Amateurs
Old rule-based bots like Warbot, Shanky

There are no regs, poker sharks, or advanced poker AI bots here that profile players and find exploits. But there are cheap bots running on various profiles that can show +EV at micro-stakes.

The VM already had the PPPoker client installed, the whole environment was set up, we just needed to sit at the table and run the poker bot’s .exe file — from there it does everything itself. In the bot’s terminal window, logs were visible — OCR was recognizing cards at the tables, balances, buttons, and interface elements, sending all data to their “AI”, and then making an Action.

For game analysis and reviewing stats, we recorded the entire hand history using Hand2Note 4 and video recorded the entire gameplay — from start to finish.

Analysis of 3UpGaming Bot Sessions

At first, we received a raw product — their bot froze, didn’t take actions, lost connection with their servers, and did weird things. After a detailed description of the bugs to Alex, an updated stable version was released 2-3 days later.

Session 1: The “Stable” Bot

~2,100 hands, NL4 PPPoker.
Actual Result: about −$144.6.
EV: about −$121.8.

Stats:

Winrate: ~−309 bb/100
EV Winrate: ~−251 bb/100
VPIP / PFR / 3bet: 44 / 19 / 17
AF (Aggression Factor): 0.6
WWSF: 17%
Won w/o SD: ~−245 bb/100
WTSD / W$SD: 18% / 58%

Putting emotions aside, we get a bot that simply donates chips to opponents. And quite quickly. After a logical question to the 3UP Gaming manager: “WTF?!”, we received assurances that their AI had updated, try tomorrow (it updates and adapts daily), and everything will be fine. The bot didn’t have a good strategy, entered every second pot, called often, yet had a 3-bet of ~17%; the bot immediately surrendered after any aggression, even weak aggression, even with good cards. By then it became clear that things wouldn’t be “fine,” but there were some glimmers of hope (I apologize for any possible minor inaccuracies in the stats – I took the screenshots after writing the draft).

Session 2: Updated “Poker AI”

~2,300 hands, NL4 PPPoker.
Actual Result: about −$69.13.
EV: about −$45.99.

Stats:

Winrate: ~−87 bb/100
EV Winrate: ~−58 bb/100
VPIP / PFR / 3bet: 22 / 8.7 / 8
AF: 0.5
WWSF: 15%
Won w/o SD: ~−68 bb/100
WTSD / W$SD: 17% / 57%

The second version turned the bot into a less maniacal fish. Aggression is noticeably toned down, but the gap between VPIP and PFR remains large. Too many calls, few raises and isolates.

The bot plays loose but has zero postflop initiative. WWSF is 15%, whereas for regulars it’s 40-50%. The AI “robot” doesn’t fight for pots, it gives up and folds. In both sessions, the EV graph goes steadily and predictably down. It’s not a question of whether the 3upgaming poker bot is running bad or not, a systematic approach is clearly visible where the bot drives itself into negative situations and surrenders.

Other flaws of the 3UPGaming bot:

The bot aggressively raises by 1/2, 2/3, 3/4 of the stack — regardless of the pot size or stack depth (whether 50bb or 500bb). Meaning, the bot does not account for or rely on the limit, on others’ raises and calls — it is hard-coded to raise a certain percentage of its own stack.
The bot cannot auto-reload the balance, for example at =0 or when it drops below n-bb.
The bot is very easily interrupted by various pop-up messages from the rooms — these exceptions are not handled at all.
The bot often loses connection (Socket error, socket exception, invalid socket response) and mindlessly folds and sits out.
The bot FOLDS in spots where it can check. Imagine this — flop, no one raises, you can calmly check and see the next card. The bot clicks “Fold”, gets a question from PPPoker: “Are you sure you want to Fold???”, and the bot confidently clicks “YES”. Insane! They fixed this elegantly — now the bot simply clicks Cancel on the question from PPPoker. This clearly speaks to the “professionalism” (amateurism) of the 3Up gaming developers.
On the virtual machine provided by 3up gaming, free space constantly runs out, the game is interrupted by Windows update messages, offers to backup the system, and many others.
Control was often lost on the VM — you couldn’t move the mouse, although the server was active. The only solution — “press Reset and reboot the remote computer.”
They claim that “each robot we give you a VM machine and a unique IP so that the percentage of robots being banned reaches zero”. Yeah, except it’s a Hetzner data center IP in Finland, which immediately tells the security team of any poker room that this is a bot playing, not a human.

How do you like that? Does it look like a “professional AI-GTO bot” for $1800? Moreover, the managers strongly insisted that it is necessary to seat 2-3+ bots at a table ($600/month for each) to have a confident profit. Because they understand that after one payment, there will be no continuation.

Why 3UP Gaming Robot is not “Poker AI”, but an expensive fish-bot (My theories)

~4400 hands were played on a stable micro-limit field. For poker math, this isn’t a couple of hundred thousand hands, but it’s enough to see the “outlines” of the strategy and evaluate how the bot works — what it does in various situations, and to look at the trend of the EV line. If someone looks like a fish over 4k hands — that is already a signal, not noise.

I could assume that this is a quick student project slapped together: OCR + sending game state data to some publicly available LLM -> receiving an answer and executing the action. But even LLMs have a better strategy than the 3UP gaming poker bot. There is no GTO, no “Neural Network.” It’s a simulation of a product designed to milk the marks.

Possibly, the 3UpGaming project changed hands that invested in promoting the site and creating new “products.” Specifically described as a high-end AI solution with a high price tag to maximize earnings. But these are just my guesses.

Save your $1,800 and invest it in a coach or a solver instead. Stay away from 3UpGaming.

P.S. I wrote the drafts of this article a long time ago but didn’t publish them. At the moment, the bot has gone through several stages of “updates” solving problems, the “poker AI” improved every day, and in fact, 10k+ hands have already been played, which is statistically more significant. The result… EV winrate -127bb/100 hands.

How Far Are LLMs from Professional Poker Minds?

A focused poker player making a decision at the table, illustrating the challenge of LLMs in poker AI.

Poker is much more than just a game. It’s a proving ground for advanced machine reasoning, especially for evaluating the capabilities of LLMs in poker AI.

These features are problems for AI algorithms based on classical machine learning principles because all these features use the ability to think in time without time, bluff, and adapt. This is the reason we use poker as a yardstick to see how well an algorithm can perform in a game in which there is uncertainty and ambiguity.

Why Solvers Are Great Problem Solvers, But Not Instinct

The classical solvers are great at devising a balanced strategy where the aim is to discover a strategy that will prevent your opponents from being able to improve on the expected value (EV) in the long run at the tables. but they are not so good at making quick adjustments out of the emotions or instinct.

Human players on the other hand are great at making adjustments according to what is happening in the course of the game, and they tend to think instinctively or out of emotion in adjusting their strategies.

Why ToolPoker Was Invented for LLMs in Poker AI

As LLMs in poker AI continue to face the knowing-doing gap, new approaches are needed. The researchers invented ToolPoker to try to combine the best of both solvers (balance and non-exploitable) and the players (instinct) and adaptability.

The researchers did an experiment to compare the performance of six models at two games. The first game is a simplified version of poker called Leduc Hold’em. The second game is a version of Limit Texas Hold’em which presents more realistic poker games. Six different models (GPT-4, Qwen 2.5, LLaMA 3, o4-mini, and the baseline solvers: CFR+, NFSP) were given the same parameters and stakes and were judged on their performance on three criteria:

HR (Hand Rationality): How much does the model reason like a professional player?
FA (Final Action): Did the model make the GTO (game theoretic optimal) move?
AC (Average Chips): Did the model win?Why the Researchers Wanted to Test the LLMs Initially Before Making the Decisions

In order to understand how well the models would do the researchers had each model explain why it wanted to make a particular decision prior to making that decision. The results were astounding; many of the models gave perfect reasons for their decisions but then proceeded to make the opposite decision.

This phenomenon has been called the “knowing-doing gap” that is the difference between knowing something on an intellectual level and doing something on a practical level.

Three Major Ways LLMs Still Fail in Poker

While the researchers have improved LLMs with reinforcement learning (RL) training, they still possess great flaws as poker players. Heuristics-Based Logic

LLMs tend to rely on familiar linguistic heuristics. For example, a model may decide to bet aggressively in a situation because it has been exposed to examples of situations in which aggressive betting is successful even though the current situation may be very different. Poor Understanding of Mathematical Nuances

LLMs seem to understand mathematical concepts like pot odds and implied odds. But when it comes to employing those calculations in practice, they tend to be inconsistent.

Mot that an LLM will utilize a logical process to evaluate a decision to make, but that does not mean that it will execute that decision in the way that it ought to . This is due to the fact that the LLM makes a distinction between “the logical thought process it took to come to that decision” and “the actions it takes which flow logically from that decision.”

What ToolPoker Offers for LLMs in Poker AI

To improve the reliability of LLMs in poker AI, researchers introduced ToolPoker. It permits an LLM to take a break in the middle of a poker hand and consult a solver. Then when the LLM gets the benefits of the solver’s advice, it integrates that advice into its cognitive process for rendering the decision.

Think of ToolPoker like giving an overly-conversational student a crib-sheet, but only after they are able to give coherent and logical reasons for the answers generated by the crib sheet.

Here is how ToolPoker is used:

<think> – gives logic for the train of thought taken by the model
<tool>– consults a solver for GTO Advice
<output>– evaluates equity, pot odds and best possible move(s)
<answer>– acts, synching thought and deed

ToolPoker’s structured approach permits the separation of ‘hallucinations,’ so that the LLM’s developed reasoning tracks with the reasoning it puts to use subsequently in action.

Training LLMs in Poker AI with ToolPoker

ToolPoker is trained employing the two phased training methodology outlined:

Behavioral Cloning (BC): The LLM learns from annotated pro level reasonings plus solver use.
Reinforcement Learning (RL): Models are rewarded for accuracy of decisions, correctness of reasoning structure and correctness of tool usage.

Together, these phases guarantee that the model’s reasoning is both fluid and action-reachable. Performance Assessment

ToolPoker outperformed DQN and NFSP by greater than 60 chips in both games. It was within 3-5 chips of CFR+, in other words tied, when human-level variability is accounted for.

In terms of the quality of the explanation, ToolPoker’s performance was rated:

Hand Rationality (HR) was rated 2.0/2.0 which indicates expert reasoning.
Final Action Consistency (FA) was rated 1.9/2.0 indicating very strong correlation with optimum play.
Quality of the Reasoning Trace: Tool Poker’s traces were judged by human reviewers to be indistinguishable from those of a professional player using a solver.Comparing Humans vs LLMs in Poker AI

The results of ToolPoker’s decisions in comparison with professional poker players yielded the following ratings:

ToolPoker was the earth’s equivalent of matching 92 – 97% of expert EV (expected value).
Professional Poker players characterized ToolPoker’s play as being “on a par with” or “more clearly explained” than human commentators play in 70% of cases.
In 30% of the cases professional players rated ToolPoker’s play as “too good” missing the flexible aspect characterized by a human.

In many cases, human pros now acknowledge the strength of LLMs in poker AI when combined with solver support. The characterization of ToolPoker as “too good” is important. It means that ToolPoker has no ego to serve, makes no bluff plays nor any tilt. It simply plays the math of the hand perfectly – each and every hand.

What ToolPoker Can’t (Yet) Do

While ToolPoker indeed is a great improvement in AI performance in respect to poker it has certain limitations:

No Psychological Model for Handling Opponents

ToolPoker does not adjust for loose opponents nor does it utilize tilt. As such, it is completely GTO reliant as such even when deviating from GTO would give it a greater EV.

Dependence on Solver

ToolPoker needs real time access to a solver in order to function. As such, it really is not scalable and the time for it to make decisions is longer than it could be because of this dependence.

Risk of Use

If left unchecked ToolPoker could facilitate RTA (real time assistance) cheating in live poker. Essentially an RTA bot which would use a solver to find the best decision in real time and send it to the player also in real time.

To eliminate this risk, the researchers have suggested audit trails, watermarking of the output of the solver and the implementation of ethics layers. Most importantly, they explicitly reject the idea of commercialization of the product and suggest strict regulation and control of its use.

LLMs in Poker AI: Why It Matters Beyond Poker

The impact of LLMs in poker AI goes far beyond the poker table. ToolPoker is much more than a poker playing AI. It represents a blueprint for developing AI systems that:

Think in a structured sequential manner.
Effectively use tools.
Have a transparent reasoning basis for their decisions.
Make high risk/high stakes decisions with transparent accountability.

Some areas where it could conceivably be applied include:

Finance: Automatic bids, fair price of a given market.
Policy: Adversarial negotiation models.
Cybersecurity: real time response models with full explainability.
Law and Medicine: Auditable models with human oversight.

In short poker is being used as a testing ground for AI accountability.

Synthetic Rationality and The End of Intuition

ToolPoker does not bluff, tilt or guess. It calculates. While it may lack the intuition of a human player, its calculation capabilities vastly outstrip that of the human player.

The authors try to define this by calling it synthetic rationality – the ability to reason without the benefit of experience and be able to make that reasoning known clearly.

They do not call the new systhesis the end of intuition, but the beginning of a new role for human intuition the editor, curator or overseer of AI systems.

p.s. updated:

PokerBattle.ai: A Live Head‑to‑Head of LLM Poker “Reasoning”

A head-to-head live test of the “reasoning” of top LLM’s poker strategy was conducted through PokerBattle.ai as a no-limit Texas Hold’em cash game lasting one week in which all models were given the same input prompts and played against each other without access to code or tools. All models were provided with their own “live” summary of their thought process throughout the session and were also given a “fold when I fail” feature to avoid taking unnecessary risks. At the end of the seven days of heads up competition, the three models had logged 3,799 hands each at 10/20 blinds, with both models beginning with equal bankrolls. The results of the event were that OpenAI’s o3 model won the most money ($36,691), followed by Claude Sonnet 4.5 who earned $33,641 and Grok who won $28,796. Meta Llama 4 lost the most money (-$100,000) and thus last place. This wide disparity in performance is evidence of a significant difference in the level of consistency and risk management between the four models. (About; Event Results) The event generated considerable interest — including front-page discussion at Hacker News regarding the strategic consistency of LLM’s and the randomness of LLM’s decisions, and even a tweet from Elon Musk boosting Grok’s brief lead — generating a large amount of debate and speculation about whether the current crop of LLM’s are capable of sustaining a GTO-like strategy that incorporates elements of both exploitable and non-exploitable mathematics/format-based errors in order to outperform humans at the table. (Reddit; PokerOrg recap of the event including the final standings; PokerNews coverage of the event.) In its design, PokerBattle.ai represents an innovative type of “open” comparison designed to provide an apples-to-apples evaluation of the “publicly available” decision-making processes of LLM’s that has been evaluated using solvers and/or dominant algorithms, but this time, PokerBattle.ai focuses on providing a publically accessible and inspectible view of the reasoning traces used by LLM’s to make decisions during the competition. Therefore, PokerBattle.ai serves as a timely reality-check: today’s general-purpose LLM’s have the ability to provide logical explanations of how they are thinking and adapting based upon prior experiences (via “notes”) however they continue to be unable to demonstrate the robustness and equilibirum-aware play necessary to beat professional players at the table on a regular basis. As such, it appears that PokerBattle.ai will likely be useful for instructional purposes and for analytical purposes related to understanding how LLM’s think about poker, but much less useful for developing models that can compete with professionals at the table on a regular basis.

Bluffing in Leduc Hold’em by AI: DQN vs CFR (2025 Study)

Player at poker table illustrating bluffing in Leduc Hold'em by AI

While analyzing bluffing in Leduc Hold’em by AI, the algorithms used in this study, the terminology applied to them, and the volume of simulations performed in the 2025 research provided evidence that a poetic truth exists — artificial intelligence lies; however, it does not lie because it has been instructed to do so, but rather, when there is insufficient information available, deception is the most rational choice.

As a result of 100,000 controlled simulations of Leduc Hold’em, two algorithms — DQN and CFR — produced bluffing strategies with approximately the same level of success (34-39%); however, the manner in which the two algorithms produced their bluffing strategies was distinct. CFR created bluffs randomly throughout the range of its strategy, much like a professional would disguise his hand strength. DQN produced fewer bluffs than CFR; however, the bluffs produced by DQN were significantly more precise and were produced at the perfect moment. It was akin to a machine version of a gut feeling.

Pure code can be made to simulate confidence when faced with the uncertainty of data. This is the essence of bluffing in Leduc Hold’em by AI, where deception arises naturally from uncertainty and not from intent.

Poker as a Psychological Arena

Poker is not merely a game; it is a psychological arena dressed as a card table. Each raise and/or fold is a motion in a silent drama of partial truths. That is the reason why researchers studying artificial intelligence are interested in poker, as it provides an environment that includes logic, un-predictability and strategy.

Using the Leduc Hold’em format of poker, the 2025 study, titled “Analysis of Bluffing by DQN and CFR in Leduc Hold’em Poker,” stripped poker to its bare essentials in order to determine if systems based upon math and feedback can develop deceptions without first learning to deceive.

Spoiler alert: they did.

Bluffing in Leduc Hold’em by AI: DQN vs CFR

CFR is the control freak. It analyzes each game decision in reverse, in order to find out what it should have done and adjust accordingly until the regret is eliminated. Eventually, CFR develops a level of balance in its play such that no player can take advantage of it. That is the theoretical concept behind CFR.

DQN is the experiential learner. It makes a guess, attempts the guess, fails, attempts again. It is not attempting to achieve perfection – it is attempting to achieve reward. It finds what works and focuses on that.

Neither CFR nor DQN had pre-programmed bluffs or pre-defined tactics. Both simply began with cards, rewards and logic.

Although neither CFR nor DQN were programmed to bluff, both developed bluffing strategies. These results are not merely impressive – they also reveal characteristics.

Experimental Design of Bluffing in Leduc Hold’em by AI

Two artificial intelligent agents. Equal stacks. Equal blinds. Each agent has one private card and one public card. Two betting rounds. No noise. No multi-player interaction. Simply, clear, distilled decision-making.

Each decision was recorded and analyzed for hand strength, bet size, and the response of the opponent. The objective of the experiment was to record the frequency of bluffing (a weak hand, a large bet), the effectiveness of bluffing, and the development of the bluffing behavior of each agent.

Again, no bluffing rules were defined within either CFR or DQN. Bluffing emerged as the only logical method to survive. In this distilled setup, bluffing in Leduc Hold’em by AI was not programmed—it evolved as the most viable response to incomplete data.

What CFR and DQN Teach Us About Bluffing in Leduc Hold’em by AI

CFR: Equilibrium and Unpredictability

CFR’s bluffing strategy was logically even-handed. It placed bluffs across all of its moderate-strength hands equally, not due to deceit, but because game theory dictates that a balanced bluffing strategy will make opponents less predictable. Bluffing was a necessary element of CFR’s equilibrium. Systematic. Calculating. Cold.

DQN: Opportunistic and Erratic

DQN’s bluffing strategy was more erratic. It bluffed in groups. After losing a hand. After an opponent folded. After “feeling” the correct moment. DQN was not consistently successful in its bluffing; however, it was adaptable. If it had been possible to listen to DQN, it may have sounded as though it was thinking: now is the time.

Identical Results; Distinct Methods

Both algorithms produced virtually identical levels of success in their bluffing strategies. That is the major point: bluffing is not an aberration of code. Bluffing is a requirement. In situations where the information is missing, deception is required to develop a viable strategy.

Strategic Lessons from Bluffing in Leduc Hold’em by AI

Why should we care that CFR and DQN lie?

Because it demonstrates that bluffing does not require intuition. Bluffing requires uncertainty. That is all. This principle underpins bluffing in Leduc Hold’em by AI, where lack of perfect information compels both biological and synthetic agents to deceive. With the absence of sufficient information and the requirement to win, any type of system (biological or synthetic) will create deception.

CFR produces bluffs based upon the requirements of mathematical certainty. DQN produces bluffs based upon experience that demonstrates that it is sometimes beneficial to pretend.

Bluffing is no longer unique to humans. Bluffing is simply…optimal.

Practical Implications of Bluffing in Leduc Hold’em by AI

Regardless of whether you are developing a poker-playing artificial intelligence or you are sitting at the poker table yourself, the implications of this research are significant:

• CFR-type bluffing is rhythmic. It creates bluffs using medium-strength hands and uses a pattern to predict its opponents’ responses. Identifying CFR-type bluffing involves recognizing patterns.

• DQN-type bluffing is emotional. It increases the rate of bluffs during winning streaks and responds to heat and cold. DQN-type bluffing mimics the tilt-recovery-tilt cycle that many humans experience.

Recognizing the rhythm will remove the mystery.

Limitations of the Study

This study is not without limitations.

It Is Not Real Poker

Leduc Hold’em is a small, clean format of poker. It does not replicate the complexity of No-Limit Texas Hold’em, including the depth of stacks, the flexibility of bet sizes and the complex interactions between players in No-Limit Texas Hold’em. Therefore, although the results are interesting, they may not accurately represent the chaos of actual poker.

Limited Training Data

100,000 games of poker may appear to be a significant amount of training data. However, 100,000 games represents only a short period of time in the training process for DQN. Additional training data may influence DQN’s behavior and potentially cause DQN to produce bluffs in a manner that is closer to CFR’s balanced bluffing strategy. Or, perhaps, DQN will not be influenced.

No Comparison to Human Players

The study did not include comparisons to the bluffing of human players. As a result, it is unknown whether DQN’s bluffing strategy appears to be bluffing in a human-like manner, only that the bluffing strategy of DQN has statistical similarities to the bluffing strategy of CFR.

Artificial Intelligence That Lies

When an artificial intelligence bluffs, it is not deceiving others. It is simply making decisions based upon probability. However, when two artificial intelligences deceive each other and succeed, something unsettling occurs. They cross a line.

Bluffing is not emotion. Bluffing is adaptation. And bluffing in Leduc Hold’em by AI proves that deception is not a bug—it’s a feature of intelligent behavior. When a series of code simulates confidence and succeeds in deceiving others, it is realized that deception is not a flaw of human nature. Rather, deception is a characteristic of intelligence.

When Poker AI Took Over: From Whiskey Nights to Code Wars

Poker cards on the table in front of the poker AI interface: real-time combination analysis

It didn’t begin with a roar, but the subdued whir of a server rack nestled in a research lab whose walls weren’t quite the color of casinos. This might very well be the first one you consider now – machines in a glow, grumbling engineers; not cigarette smoke and clinking glasses. And yet here we are in the world of artificial intelligence and we are discussing poker, a card game in which the theatrics of the bluff are matched by the elegance of an algorithm.

I recall the days a poker cheat was something you held in your hand – a creased corner, a thumbprint of grime, a pal who’d curse you with a “lucky” deck he’d purchased in Reno. Poker hacks now are phrases of software patch notes themselves, and the man across from you at the table no longer smells of booze; he smells of code, polished and optimized, a world made possible by poker A.I. research that’s taken late-night confrontations and turned them into a study in statistical perfection. This development owes largely to the emergence of poker AI, taking human intuition and turning it into algorithmic perfection.

Poker AI: The Algorithm Wears No Sunglasses

There’s something eerily uncomfortable about sitting and gazing at a screen on which a small avatar grins like a Cheshire cat, not once peeking at one’s own cards and not hesitating, while you fret whether you’re going to get beaten by a rather more politely-behaved bot than any human you’ve ever known. In poker, the bots were previously patsies – beats-by-blundy opponents of obvious flaws you could calibrate your play around. Now, as poker’s machine learning gets more advanced than acaffeinated 4 a.m. trader surging down Wall Street, we have poker AI beating pros, teaching solvers how to teach themselves and — here’s the kicker — making the term ‘best poker AI’ sound like a brag in a dating profile.

I’ve played against them. Or at least, I think I have. It’s difficult to know, as the entire idea behind a good poker AI bot is subtlety, like a thief who leaves your home cleaner than it was when he found it. One minute you have aces and you feel untouchable, and the next you’re staring at an empty stack full of embarrassment just because something in a line of code just made you feel like an amateur. Such is the power of modern poker AI, capable of exploiting every micro-leak in your strategy.

Here I am pouting in a Slack channel somewhere, someone no doubt calls it progress.

How Humans Built Monsters for Fun

So, the architects – the dreamers, the tinkerers, the MIT pokerbots team who asked themselves, why not make poker an AI dominance testing ground? From DeepStack AI to Pluribus poker, the very names sound like a Marvel comic and not around a felt table. These innovations are the golden age of poker AI research, defining the very essence of the game. But this is the reality: poker AI design was not about beating others with cards. It was about solving a centuries-old mental dance, about telling the world, “Yes, this – the holy game of patience, aggression, tequila shots occasionally – all this can be solved.”

And they did solve it, with poker AI algorithms so refined they make your old poker cheat sheet look like a cave drawing. These aren’t mere bots; these are relentless optimizers, born from reinforcement learning loops that chew through billions of hands, spit out poker AI strategies sharper than a scalpel, and then do it again, because why not be perfect twice?

I read the papers, by the way – the fat wedges of math in the semblance of naivete. They write about regret minimization, points of equilibrium, Texas Hold’em simulation software requiring more servers than my entire childhood town had light bulbs. And, yes, I played around with poker AI software myself. Huge mistake. It turns out learning the theory won’t keep you from tilting like a cheap barstool as the AI folds the river that you absolutely positively know it’s going to call.

Poker Hacks and Hidden AI Tools Nobody Admits Using

Poker hacks once referred to stories rumored in the back rooms, a whispered voice regarding a 90%-effective tell (spoiler alert: it didn’t work). Poker AI software nowadays are the very hacks, grounded in GTO solvers and machine learning. These are the times of “poker hacks” as installing the poker AI software you absolutely can’t trust, or the hunt for the best poker bot as you would the holy grail rare record. Poker bot cheats are asked of me like cottage gossip – the weather.

And then of course there is the whole gray market where you can download bot poker software gratis and hold’em apps for iPhone or Android and study GTO theory in the kitchen. Some even sentimentalize it as you coming in to study some kind of ancient martial art except nunchucks are replaced by neural networks.

Can’t fault them entirely. The appeal is real. I’ve tried trainer poker software, played around with poker bet size tables, even toyed around with something called PokerSnowie – whose amiable name you suspect until your bankroll gets slaughtered and you waste time deliberating life choices.

Are We Still Playing Poker in the Age of AI?

This is the question for which I return again and again to myself: are we playing poker still, or did we build a simulation so perfect it eliminated the messiness of humans that made the poker great? I mean, I get it – the world continues spinning whether or not we’d like to go back in the bad old days in which we reveled in the smell of cheap bourbon and bad decisions. But this is the point: poker wasn’t the cards. Poker was in a known sigh, a twitch, a story someone didn’t mean to tell.

The stories are coded now. Efficient, yes. Elegant, indeed. But does a poker AI bot gaze upon the sleek interface and discover any trace of the slow-burn tension, the inadvertent lyricism of the bumbling bluff?

I’m not nostalgic – okay, maybe I am – but I keep wondering if, in perfecting the game, we lost the game. And maybe that’s fine. Maybe the new frontier is learning how to play against machines, how to beat poker AI with a better AI, how to survive in a meta where the only real bluff is thinking you’re still the smartest one at the table.

And whether you’d like to know the answer to whether or not I gave it a try? Oh, yes. And no, I didn’t take the prize. But this is one thing I did discover: when poker began to think, poker didn’t grow any less human – it forced us to confront the question of what precisely the human at the table amounted to.

Poker AI Strategy: How Algorithms Are Changing the Game

Friends playing poker at a table, smiling while collecting chips, representing the psychology and unpredictability of poker AI strategy.

Have you ever sat down at a table and felt like everyone else is in on some secret that you’re not? It’s not just a card count, but a script, an algorithm, a silent hum of certainty trafficing in the dark behind their sunglasses. That’s where I ran into trouble — or, at least, that’s when the kinds of questions that prevent you from getting to sleep while the chips are still disappearing began to occur to me.

Back then, poker was theater. Men resting on elbows, women pretending not to show any interest, the barely audible clink of ice in a tumbler that spoke a thousand times more than any bet might. One raised eyebrow could send you folding kings without blinking. That was the game — reading faces, not code. But here we are, scrolling forums for poker hacks, downloading poker cheat sheets that promise mastery in six bullet points and whispering about bots in poker like they’re campfire ghosts. Today, poker AI strategy shapes the way we approach every decision at the table, blending old instincts with algorithm-driven precision.

I once tested one of these a.i. poker tools. For curiosity’s sake, not boredom, or at least that’s what I told myself. It wasn’t pretty, just a clonky piece of poker AI software you could download in three clicks, then install and — bam — now you have a digital consigliere. A voice that whispers, “Bet 37 percent here,” as if analyzing probability like some dispassionate surgeon. It didn’t blink, didn’t doubt. If only I could say the same for me.

Poker AI Strategy vs Human Instinct

You see, the psychology of poker was always a heady blend of confidence, pattern recognition and risk. Your neurons ignite in time with your heartbeat, the taffy of your cognitive load stretching and pulling whenever someone in your game check-raises on the river. There’s a name for it — executive function under uncertainty. Academics love phrases like that. Me? I refer to it as sweating bullets and pretending to drink bourbon.

Then machine learning in poker arrived. Deranged neural nets gobbling up hand histories, spitting out poker AI strategies so sharp it made Doyle Brunson sound like your uncle giving fishing advice. Pluribus poker, DeepStack AI, names more fit for a sci-fi movie than the felt. They aren’t tools, they’re predators—top poker AI ever coded to take advantage of every leak in your game before you even realize you’ve sprung one. These breakthroughs prove that poker AI strategy is no longer an abstract concept—it’s a dominant force changing competitive play.

I remember one time at the Bellagio I asked a guy if he was scared of bots. There was a harsh laugh, almost a scornful noise, followed by some muttered comment about “adapt or die.” Less than two weeks later, he was railing on a forum about online poker bots ruining his life. I didn’t say I told you so. Hell, I didn’t tell anyone.

When Poker Bots and RTA Change the Strategy

And here’s the tricky part. It’s not only bots that are playing online poker, it’s humans as well — using so-called poker AI algorithms in real time as they play. Real-time assistance, they call it. RTA poker. It’s a word that sounds sterile and clinical and authorities, like five-o’ and grandma saying baby, think it’s harmless — all until you realize the guy four seats over doesn’t just “know ranges” but is still in fact looking at a supercomputer every time he looks at his pocket. WarBot, Slumbot, PokerSnowie — the names read like figures in a comic book, and perhaps they are.

And I found myself wondering: Does that mean I’m the dinosaur? I used to be a tell guy, those micro-expressions that twitch like bum neon. And now I’m up against something that doesn’t twitch, doesn’t blush, doesn’t sense the glacial heft of a bad beat. You see, when you play poker against bots, you’re not just losing money; you’re losing the very thing that drives us humans to carry on.

And still, there is that pull, isn’t there? One night, drunk on bourbon and curiosity, I typed “best poker bot” into Google and stared at the pages promising unimaginable power, downloads for bot software that promised to change my life, even whispers about pokergpt like it was the dark oracle of GTO perfection. I did not click buy — but my cursor hovered for longer than I’m willing to admit.

The Beautiful Mess We Call Instinct

What’s ironic, or cruel, is that poker was always a psychology experiment in guise of a card game. The cognitive loops, the risk-reward models — they have been mapped out, measured and dissected into tidy tables and poker bet sizing charts that appear like commandments for the mathematically pious. But the mess? The stammer in your voice when you bluffed, the bead of sweat you prayed no one else saw, the time you read a smirk wrong because your ego fired off so loud — that’s what made it art.

Now the art is in code. The caffeinated cheetah that is poker AI development comes up with strategies faster than any mortal brain could attempt to replicate. The poker AI research community, dubbed “the poker brain trust” by one insider, drops white papers like mixtapes, each one sledging the new algorithms the group think that make yesterday’s edge a quaint relic.

I scroll through these studies occasionally, an audience of one with a cup of bad coffee, half-impressed, half-terrified. They’re talking about equilibrium strategies, Nash solutions, counterfactual regret minimization — which all sound more at home in a Pentagon briefing than a Saturday night game where, swear to God, someone’s uncle is still pulling for limping to be viable.

So Where Does That Leave Us?

Not at the end, not really. Just in a weird middle, where flesh-and-blood instincts grapple with silicon accuracy, where a poker bot can out-bluff a man who has read Caro’s Book of Tells cover to cover. The irony? We designed these things to know from us, and now we’re running to learn from them. Understanding poker AI strategy is now essential for anyone who wants to survive in this evolving game.

So I keep playing. Maybe stubbornly, maybe foolishly. I pore over the leaks, I fiddle with tools, I grumble about AI and poker the way an old coot gripes about smartphones. And every once in a while — just every once in a while — I see a flicker of the old game in the new. A human pause, a vagueness that no algorithm can fake. And, when it comes to that, I smile, slide one stack forward, and consider: maybe there’s still room for a little chaos in a world that can’t seem to get enough of perfect play.

Poker AI and Bots: Ghosts at the Table?

You wonder whether you felt the shift initially. At a quick glance, it appears to be no different from any other game. Some familiar avatars, the same old chat-box piffle, a few limp calls. But then something begins to scratch at the back of your mind. That player – screen name “RiverSaint88” – never hesitates. Not once. Bet sizing? Ideal, as though they’ve memorized all the poker cheat sheets ever made on Earth. They bet where they have no business betting, check-raise because they feel like raising, muck when most common folks would not be able to contain their curiosity chips and play the board. And there’s the thought, faint at first and then loud: this is not only skill. This is something else.

I’ve been streaming online for years, through every fad, and so-called “new edge.” First, it was HUDs. Then solvers. And then there were “training apps” that promised to make you a GTO wizard overnight. Hell, I even spent months getting obsessed with poker bot research again like, some caffeinated grad student, pouring over academic papers. I reassured myself that it was all in the name of the love of the game, even though, let’s face it, I simply wanted an edge that didn’t entail selling my soul.

And now here we are. Poker AI is no more the ghost in the forum or the bogeyman of paranoid regs. It’s here at the virtual table across from you, smiling facelessly.

When Bots Stopped Being Dumb

Remember when bots were jokes? From scripts that went all in on a pair of threes, to scripts that folded two aces due to a mistyped condition. Those days are gone. In other words, today we’ve got monsters like Pluribus, cooked up by Facebook and Carnegie Mellon, casually taking apart pros who thought variance was their best friend. And you DeepStack AI – don’t even get me started. Someone taught this machine how to do more than play poker: They taught it how to love breaking hearts.

The first time I ran up against what I later learned was a bot, I didn’t realize it immediately. I just remember thinking, wow, this guy’s good. Not great, in the soul-reading, Phil Ivey, fusillade-of-chips sense, but inevitable. Like gravity. Each bet was a size that was just right for maximum EV. No tilt. No hesitation. No chat-box complaining about bad beats. (And believe me, I tried. Dropped a classic “nice hand” after a brutal river – nothing. Not even a dot.)

Later, sifting through hand histories, that made sense. The lines were not just strong; they were surgical. Too regular for a human heartbeat. And that’s when I went down the rabbit hole – poker AI programming, poker AI algorithms and a seemingly endless march of software names that could double as secret government projects.

Machine Learning at Midnight

Wanna hear the scary part? As they don’t just play but also learn, these systems don’t mention the law or any concepts; they can’t. Machine learning in poker isn’t just some sort of academic buzzword, but something that’s silently plotting your doom each time you binge-watch that Dexter re-run on Netflix between sessions. They go through hands, identify patterns and develop counter-strategies while you war over limp-calling in the small blind on Reddit about whether this play is “balanced.”

I’ve read the studies – hell, I’ve been there. From basic range charts to neural nets which adjust on the fly mid-hand, it’s all out there. And if you fancy that you can outsmart it with a couple of “poker hacks,” well, God be with you. Those old tricks – timing tells, overbets designed to spook somebody, whatever – they rebound off of these things like pebbles against steel.

The Dirty Temptation of Hacks and Bots

Look, cards on the table (pun intended): I was tempted. After one too many sessions where I felt a chew toy for some unseen predator, I began Googling. “Poker now hack.” “WSOP cheats.” “Download bot poker.” You would not believe the promises that enter into a manual review of your Google Play store listing: the impersonations, the hundreds of fake email accounts, the scammy Chrome extensions and the full-on poker bot customers hoping to swindle you have been immense.

Did I cave? Maybe. Once. Pure research, I swear. (Okay, also ego. Fine.) I experimented with one of those, the so-called best poker bots. It was… awful. Played like my drunk uncle at chef’s house Thanksgiving, but not endearing at all. So the actual sharks aren’t posting their secret sauce on some seedy forum, it turns out. The legit stuff – the top bot for poker, the top poker bot software in the world – that’s deep in the black market, protected like nuclear codes.

And yet people buy it. They scour the internet looking for poker bot online, bot software download, poker AI tools like treasure hunters with a death wish. Some are seeking an edge, some revenge, and some, I think, simply want to watch the world burn in seven-bet pots.

Humans vs. Ghosts

Here’s the thing that kills me: poker was human. Messy. Beautiful. A fellow sweating bullets, as because the rent depended by it. A woman drinking whiskey slow, pretending she didn’t care even as her pulse hammered in her throat like a kick drum. Now? You’re eyeing an avatar called “BluffMaster2000,” and you’re trying to figure out if that’s a college kid in Kansas or a server farm in Shenzhen.

Sure, poker coaching software still was the promised land. GTO sims, poker bet sizing charts, apps that come with happy-go-lucky UIs that tell me “Beat the bots!” And maybe they help – maybe. But the gap’s widening. Bots don’t tilt. Bots don’t misclick. Bots don’t rage-shove because their dog barked at an inopportune time.

Me? I’m not done yet. I’ll keep on fighting, Even though it’s going to make me pass out from studying so much. I will fire up every poker tool, every trainer poker app, every scintilla of wisdom from the bowels of the internet. Will it matter? Hard to say. For in the end, this is no longer just poker – it’s poker versus ghosts. And ghosts don’t blink.

So Where Do We Go From Here?

I could bring this to a tidy conclusion, something optimistic, perhaps even inspiring. But that would feel dishonest. Truth be known, I do not know where this ends. Maybe humans adapt. Maybe the bots win. Perhaps one day we’re all just watching as AI poker bots vie for dominance, and the prize is … what? Bragging rights? A new firmware update?

For the time being, I’ll keep logging in, keep shuffling virtual chips, keep pretending the chap three-betting me isn’t running a poker AI bot coded by person(s) who probably don’t even play the game. For hey – what else am I gonna do? Quit? Yeah, right.