Joe Ingram: Let's start with bots, because obviously bots are always one of the biggest and most controversial topics in terms of automated software. We saw that big article come out earlier that basically talked about how there's product offerings to poker players, where they can purchase these player profiles and purchase this automated software for their communities and their clubs. Obviously, this has been a popular topic over poker years when we talk about security.

How do you think about stopping that, and how do you think about policing your games against that type of activity?

John Andress: Yeah, maybe it's good to define the different problems first. I think compared to when I started playing, the problem has changed a lot. Maybe you can think of the fully automated bot — back in the day, PokerStars used to be the gold standard for security, and there was a lot of what we call contextual detection: invasive tracking on your computer or even some behavior tracking with mouse clicks.

As far as fully automated bots go, the technology has come a long way. Referencing that Bloomberg article, there is basically a huge organization out there that makes a lot of money doing this, and they invest a lot of resources into staying ahead of the curve as far as the cat-and-mouse game goes in security. Not only are they working to make sure they can get around the operator security protocols, they have also made a lot of advances in their gameplay too.

I think the way people traditionally view these GTO bots isn't actually what really exists out there now.

Editor's note – The September 2024 Bloomberg article John references here revealed a large, organized industry selling advanced poker bots and player profiles to online communities and private clubs. These bots use manual coding, precomputed solutions, and evolving strategies to bypass security measures, aiming to profit by exploiting specific player pools rather than strictly following GTO principles.

Joe Ingram: You're saying they have different personalities, right? Like, one guy could look like a typical passive loose player who's limping a lot and calling a lot, but really it all comes from the same situation? They've gotten a lot more advanced.

Thanh Tran: In that way, yes, they have evolved quite a bit. In short, the Bloomberg article talks about advanced AI — that's the research side of the problem. The practical solutions are quite different. I don't think they are very much AI-driven. They do have solvers and have precomputed solutions for certain games, but using solvers is not really scalable if you think about all the different game formats there are.

There's a lot of manual coding, working with poker players, domain experts, and really hardcoded logic. That part is actually quite important.

When John speaks to the different profiles of bots that exist, really it's about creating different rules that basically make the bots play high VPIP, high aggression factor preflop, but they are not as smart as you think — not this sort of blackbox AI that does all the magic. There's a lot of manual coding and manual expertise involved in crafting the bots. But they have evolved over almost 10 years now, right? I wouldn't say that the technology behind it is very advanced, but because there's quite a lot of money in that business, a lot of work has been put into the manual configuration and coding of a lot of different bot profiles.

John Andress: Yeah, I think maybe the way the poker community traditionally perceives it is that these GTO bots come in. But if you're building a solver, your goal is to solve for Nash equilibrium. Their goal is to make as much profit on whatever specific player pool they're targeting. Their approach is more about MDA and exploiting the player pool.

Thanh Tran: It's really funny when you look at these bot profiles — they are not even close to GTO.

Joe Ingram: Yeah, so GTO bot isn't an accurate way to describe any of this. It's more just players that have figured out different types of styles that can win or that can lose, and then they put those into an ecosystem. Some are winning, some are losing, and the losing ones are creating liquidity for the entire games.

Joe Ingram: Can you talk a little bit — I don't want you to give away the secrets — but can you share anything with the audience to give them a different way to feel safe? I think most players just want to feel safe.

Thanh Tran: Sure, we can talk about solutions. Maybe we can talk about the problem first and then move on to solutions. Why are we here? Why is it a big problem? Maybe we can even walk you through examples or issues we have seen in the past, indicating that this is actually a huge problem.

John mentioned bot collusion, multi-accounting, ghosting — a lot of these security concerns. Over the years, working with clients, we know it's a big issue.

John Andress: Yeah. Going back to the bot problem — the fully automated bot, like the Bloomberg article or maybe more niche groups — from my perspective, that's really the problem that sucks the most money out of the ecosystem. When those organizations can come onto a platform and scale, their main goal is profit. They're better at exploiting and winning money at scale than the regulars are, and we've kind of seen that from the win rates.

Separately, RTA is a more bespoke problem where you might have some high-stakes pros that develop something to gain a small edge. It's still an issue for game integrity, but as far as the actual ecosystem, it's pulling out a lot less money than these organizations running the bots.

I feel like RTA is more about protecting the poker dream — so when people get to high stakes, they have a shot to win. Preventing these advancements from AI and destroying online poker is more about this automated, scalable bot problem that many companies are facing.

Thanh Tran: The sad reality is there are a lot of these operators that have big issues with these bot rings. When you really look through all the forums, all the things reported are true — there are bots running on these sites.

News
Bots in GGPoker Cash Games: TylerRM Finds 15 New Accounts
Read

The poker community’s leading sleuth appears to have bots on GGPoker. His April report outlined their habits, games, and +7bb/100 win rate.

Read

Joe Ingram: We didn't want to believe it. When this article came out, I was pretty mind-blown because we've heard theories — conspiracy theories — that there's a group making all the bots and putting them out there. A lot of people's minds were blown.

Then the GTO Wizard thing this summer — shout out to GTO Wizard — where Dominik Nitsche was using a solver on the rail, now people are becoming more aware of AI. A lot of people have known about this stuff; I've talked about it in my content for the past. But now it's coming into the mainstream.

From left to right: Dominik Nitsche in the white t-shirt, Joe McKeehen, and Jonathan Tamayo (the 2024 Main Event winner)

You touched on solutions — how do you even think about trying to battle against something like that?

Thanh Tran: The fight started for us six years ago, so this is not something new to us. I think recently it has become more evident.

I like to think the problem is huge because some operators just don't care. There are a lot of these small operators that come out, and their number one objective is to make money — and make money as quickly as possible. That's maybe the number one concern. Of course, there are big operators that have a long-term view — they want to protect ecology, they want to protect the poker community. That's why we're here.

Companies are not very high-tech. They think about building a poker solution as having the best wizards and this very nice client application — this app you can run — but they don't think about building the data infrastructure, the machine learning vehicle you need to fight bots, or even just dealing with simple problems like matchmaking to make the games more fun and fair for players.

Joe Ingram: To me, it seems like — I read through the A5 Labs white paper, and it's pretty interesting. You're breaking down your strategy for how you defend and detect, because you get all this data when you're the operator. You mentioned earlier about non-invasive detection — very interesting topic now as AI gets more advanced.

How much invasion do you do in terms of someone's privacy? You mentioned tracking systems you can put on a computer or a real-life device.

What's the thought process on the bots, then?

Thanh Tran: Really, when you ask the question, "How do we protect poker players?" — first of all, you need to make the poker operators care. We're a tech company, but we're lucky that we have partners and friends who share the same passion for the game. They do care. They invest a lot of money in tech to solve precisely this problem. That's number one.

When it comes to the tech solution, data matters. Broadly, you can think about fighting bots from two points of view: the context-based solution and the gameplay-based solution. There's the context-based solution, which, if you think about the major types of solutions in gaming, you have what is called kernel-based detection. What these types of solutions do is basically watch over the players. They look at the operating system, the software, the processes that are running, the app distribution — all the way down to the kernel level. That's why it's called kernel-based detection. It's very invasive.

Then the second type of solution is what we refer to as gameplay-based detection. This is where you need a large-scale data infrastructure. This is where you have to track behavior, gameplay, everything you can see, and build that sort of signature that distinguishes human from bot behavior.

We actually think that we have the most advanced holistic solution that takes contextual signals into account, but also has the data/AI-driven gameplay signatures. The reason we believe this holistic solution is important is that context-based solutions are very invasive. Recently, you have heard about how some systems have been hacked and data has been exposed. Clearly, you don't want to have this big brother always watching over you just because you play poker. It doesn't mean you want the poker operator to know everything about you.

Joe Ingram: Some people want to be protected, but they also want their privacy. They don't want to be invaded, but they do consent to that type of behavior or tracking, especially if they feel like they're getting more security. It's a common thing we see in America and everywhere — this concept of, if you want to be safe, you have to accept some invasion into your life. In poker, it's like — how much, as the operator, do you want to do that? Some operators, that's the strategy they're using, I guess, is what you're saying.

Thanh Tran: So I think that's a choice for consumers, but we believe as a tech provider, you need to really give the user that choice. When you are a professional poker player, it's probably okay for you to have this big brother watching you. As a recreational player, we don't think you should have to install a kernel-based detection system. This is why we believe in a combination of context plus gameplay. Maybe John can talk more about the gameplay side.

John Andress: I think, just to wrap up the context-based side — when he says gaming, a lot of the big video game companies have these kernel-based anti-cheats.

The issue when we evaluate future-proofing the problem is that it's still a cat-and-mouse game. You have to do a lot of research. You have to figure out what applications are running. The assumption longer-term is that, especially when a bad operator is involved, they're going to change their tactics, figure out new ways, or remove hosting anything on their machine. That's kind of a dead end.

From the gameplay side, that's where our general hypothesis is that there is a definitive way to identify AI use from human use. We're predictable creatures, and that's evident in the way that we behave with a device — all the way down to how we make decisions in poker strategy.

Whether it's detecting someone using a GTO tool or some type of automated exploitative solution, by having rich data on the player population, we can understand where those human baselines are — for pros and for recreational players.

Additionally, we can reference data from the solutions that we build, whether it's exploitative or GTO, and have really comprehensive scoring that our models can consume to give us actionable insights.

The security team can really say, "This guy's a bot for X, Y, and Z."

Joe Ingram: That makes sense. So you're taking all the hand data from the player, running it through some type of algorithm. The algorithm gives you some type of score, potentially, if you should look at it. Then if you want to do more investigating from there, what happens?

Let's say you detect something — what's the next step of due diligence from your side to deal with these issues and make sure this is a real thing or not a real thing?

John Andress: Just like in real-life investigations with crime or legal issues, there's always a scale of evidence. There are some cases that are just so obvious — either their gameplay signature is so unique, or something stupid like they're all playing from the same type of hosting, like an Amazon server.

Obviously, those cases are really easy to validate.

Where it becomes more challenging longer-term — a lot of operators don't even do that very well.

Joe Ingram: It sounds pretty hard.

Thanh Tran: A lot of things go into this. When you think about poker operators, I think you should put them in three buckets.

You have poker operators that simply don't care — they don't really have anything in place to protect the players. Then there are the big operators that are fairly familiar with context-based solutions. If you have a software installed like RTA, when they check the system processes, they are able to detect and block some of the software or VPNs, stuff like that.

When it comes to gameplay-based solutions, this is why we need to think about this from a tech standpoint. For a poker operator, it's very hard to build such a solution because you need to have a data infrastructure in place so that you can track everything under the sun.

Whether it's John referring to building the signatures for the bots and different human profiles, now you need other sorts of technologies. Actually, we have our own solvers. With our own AI and solvers, we create bot signatures, we create GTO signatures, and we also run different sorts of exploits so that we can create exploit signatures. There's a lot of tech that goes into this.

In short, I think it's a very solvable problem. You need to make the investment in order to deal with that problem.

Joe Ingram: So you're saying you just have to be so good at both sides of things — you have to know what you're dealing with, and the only way you know what you're dealing with is if you are getting in the streets yourself and just figuring it out, building different strategies, building different ideas. That's crazy.

Thanh Tran: We look at GTO signatures. We look at different exploit signatures from real bots or from this sort of AI that we have trained to exploit certain human tendencies. It's based on these signatures that we can infer and draw the line between human profiles and bot profile.

Joe Ingram: Wow! One thing we talk about a lot is transparency with this stuff. Building trust with customers takes repetition, it takes consistency. In my mind, if I'm playing, I want to feel like other humans are winning in the game. I don't know how feasible that always is in 2024 these days, but how do you guys plan to build trust, or how do you think about that side of things?

John Andress: I think maybe — obviously the more sophisticated poker users have an idea about these problems and they want answers. Then there's the whole subset of people who are clueless and just want to go online and gamble. Maybe that's changing more recently.

I think the most important thing is that they feel that — and the lead metric for that, I guess, would be: can they actually win in the games?

You look at some of these other operators, and their games are like you have five highly skilled players and one big loser. If it's as simple as that — are the games fair? If the games are fair, you should feel like you have a chance to win.

I know there are other inputs, like how many pros versus recreationals are on there, but obviously any big poker operator is going to spend a lot of money to bring on recreational players and get new players into the game. If you step back and take a look — if you don't have a chance to win, there's probably a sign that there's something not right going on.

Thanh Tran: You really asked a very good and difficult question — how, as a player, do I develop that trust?

I think what John said makes the most sense. Number one is, when I play the game, I should be able to win. If I don't get the sense that I have a chance to win here, what's the point?

We also talk about how to make the system more transparent. How you can work with operators to not only rely on them using the tech — working with companies like us to say, "Okay, this is what we do to protect you from botting and all these problems" — but also make the data transparent.

Joe Ingram: Which data?

John Andress: We've been brainstorming a little bit about this. Maybe a simple example — for some of the club game platforms, we actually make a lot more of the data transparent and let the club owners or players see it. They can make decisions and get explanations — not giving personal data out, but giving explanations for why someone is labeled for collusion.

Mining in Poker for Player Data & Hand Histories
Read
Read

Also, maybe expose the hand histories after they happen, of course, in the collusion case. Or give them some metrics around bots that are easy to understand. For the more centralized platforms, it's a little trickier. But we've talked about working with panels of pro players to do an audit of some sort, where we talk with them about the tech.

The truth is a lot of people don't care until it affects them — whether they're losing on the site or they feel like they've gotten cheated. Like he said, we want to have a way to expose that this stuff works more transparently to the community.

Thanh Tran: We are seeking the community's opinion here. I think there are two solutions.

On the one hand, it's about the experience — creating a game that is really fair, where players feel they have a sense they can win. That's most important.

Maybe the other solution is being more transparent. From a player's point of view, what does it take for me to trust the system? I want to know who all these crushers are. I want to know if they are security vetted. When they are security vetted, what does that mean? Then the game operator and the tech company should come out and say, "Hey, this is when the player looks naked. This is a legit player."

Joe Ingram: Yeah, 100%, because I'm always wondering, who are these people? Who's really winning? I see these screen names on all these different sites. I don't know anything about these people.

You're right. So what's your plan for that? Is it to share more winners? Or is it to share more information on the players who are playing in the games? What's the thought process there?

Thanh Tran: I think at this point, it's more about us sharing ideas. Again, we are providing tech solutions. We don't go as far as implementing those solutions. This is the conversation we have with our club WPTGold partner.

Not only do we want to provide the tech where everything is under the hood, and provide a fair gaming experience, but if it takes building trust and confidence, we make club WPTGold more transparent.

John Andress: It's always a discussion about balancing maintaining or building liquidity versus how restrictive you want to be. Especially for higher-risk accounts, the best way to address this sometimes is with incentives. Do you want to undergo more advanced KYC? Do you want to get your account verified?

There should usually be some incentive in place for the person to do that — so they can play high stakes or get access to games they wouldn't normally have access to unless they do X, Y, and Z to get in there.

Obviously, with our partners, it's always a negotiation — getting people into the games versus keeping the games safe.

The best promotions now