How to solve the real time solver issue

A solver is only as good as the player who uses it

Poker has always faced some sort of existential crisis. Online poker itself was once considered a threat to the game with the potential for collusion, superusers and the like. Then tracking software and game selection tools were thought to be killing the game. Now it is the turn of GTO solvers to come under the microscope.

A solver will take a hand and run a lot of simulations to discover what the best course of action would be, given the information it has been fed. This naturally has worried a lot of poker players but if you have spent anytime with solvers you know they are not the threat they sound on the surface. They are only as good as the assumptions the player feeds into them and it takes a particularly skilled mind to even understand why the solver did what it did. They are not much of a threat at the lower stakes because solvers tend to give you the best course of action for playing against strong players. They also, so far, take so much computational power that it is near impossible to use them in real time.

It does appear, however, there is a workaround to the last issue. High stakes player Fedor Kruse has been accused of using a ‘dream machine’ real time assistance (RTA) tool which is a database of thousands of presolved hands, enough that he has a reliably similar hand to compare to most of the spots that come up in his sessions. Rather than using a real time solver, he instead just consults a database.

A grey area

The alleged two PC set-up

Some people would argue that this is not a form of cheating and is no different to using a SnapShove chart or having a book by your PC while you play. The difference is that there is an assumption when a player does that they are using it as a memory aid for something they had already learned. If I tape a 10 big blind shoving range to my monitor then I have at least learned and agreed that this is a profitable range to play.

If the accusations against Kruse are true, the difference is that he is consulting something he never knew in the first place. Rather than reminding himself about what he learned he is being told the answer for the first time. He might not even know why he is taking a certain course of action.

It’s a very tricky area because there is a line between consulting something you have learned before and blindly following what a resource will tell you, I’m not sure anyone knows exactly that line is.

The more important question is how do we prevent players from consulting a database like the one Kruse is accused of doing? And how do we stop solvers from being used when they eventually become fast enough to be used in real time?

There will certainly be some measure of security implementations at poker rooms. It is already possible, for example, to track unusual mouse click activity when a person is playing. Kruse is said to have gotten around this by using two PCs and two mice. It seems inevitable that online poker rooms will have more access to information from your computer while you play, to prevent cheating from happening.

Another suggestion that does the rounds is making the game more like live poker. For example, insisting on webcams while you play. At one time that would have seemed ludicrous and a breach of privacy, but it is an idea that is gaining traction. Facial recognition security is already in the works, the rise of online home games means some people want them in the game for banter purposes and so many players Twitch stream it doesn’t even seem that big of a stretch that we would all do it.

Keep the game hard to solve

Keep the formats evolving to slow down the solvers

My personal preference is much more towards regularly changing up the formats just a little to make the solvers less useful. Kruse is said to have only ever played 100BB poker which he played near perfectly but then would make basic mistakes in 30BB effective pots. The suggestion is that his database was for 100BB spots of which he had a massive sample. If the games changed regularly – let’s say Zoom was made 125BB one week and 75BB the next, one week it had antes, the next week it had straddles – it would be hard for somebody using a solver to keep up with regular changes but a skilled player could adapt easily.

PokerStars caught a lot of cynicism for their trial formats like Fusion, Unfold and Split Hold’em, but I thought they were on to something regularly mixing up the formats. A good player can easily adapt to games like this because they have the fundamentals down, a weak reg using a solver in a paint-by-numbers fashion would be completely lost. It doesn’t take much to render a ‘dream machine’ database useless – bomb pots, antes, straddles, changing the number of blinds, etc.

This also makes the game a bit more fun for casual players, who won’t get crushed as much and have more moving parts to play with. This is why PKOs have become popular in tournaments, they are fun and harder to solve.

I don’t think this is end times for poker as long as the industry treats it seriously enough, which I think it is doing. It will be hard to stop people trying to solve the game in real time, so my suggestion is to make the game harder to solve.

How would you limit the impact of real time assistance software? Let us know in the comments:

Barry Carter

Barry Carter

Barry Carter is the editor of and the co-author of The Mental Game of Poker 1 & 2, Poker Satellite Strategy and PKO Poker Strategy

Latest posts