Exploiting GTO: Bet Sizing
I’d played against the best players in the world for many years before solvers were invented. I’d succeeded in large part due to my ability to hand-read.
After solvers, I found myself in a difficult spot – not able to trust the very same skills that got me to the top.
Between running a business and my general resistance to boring things, I played for many years against solver-studied players before I dove into studying myself.
Sure, I watched GTO experts make training videos. I could understand the concepts involved, and they provided me with some “ah-ha” moments.
But they also often left me with the feeling that these players were infallible – that they’d memorized optimal strategy and there wasn’t much I could do against them.
The Wrong Approach
When I played against these players, who I decided were too good due to their knowledge of GTO play, I stopped playing my game. Much like I talked about last week as it related to live tells, I didn’t even try to make reads.
Instead, assuming they were ultra-balanced, I attempted to play my own approximation of balanced, GTO play.
They potted river? Where am I in my range? I’ll try to call the top 50% of hands.
They bet small on the turn in a spot where I’d never have done so? Ummm… Let me just take an entirely uneducated guess at what a GTO response would be.
This was the worst adjustment I could’ve made. I was trying to beat them in the exact area they’re obviously better than me!
The truth is, I was afraid to guess against someone so “perfectly balanced.”
I was afraid to feel stupid.
So instead, I actually was stupid. Instead, I guessed in a way that I was less qualified to.
It wasn’t until I combined learning from videos and solvers that I truly understood how hand reading is still very much a thing in the post-solver era of poker.
There are many ways in which players trying to emulate solvers get things wrong and, as a result, are exploitable.
Today, I want to talk about just one of them: Bet sizing.
How GTO Really Works
Once I combined videos with solver study and play, I began to truly get it.
I began to understand the difference between GTO and what my opponents – even the best in the world – were doing at the table.
And poker became fun and exciting for me again!
True GTO strategies are unbeatable.
Scary, right?
But GTO strategies are also extremely fragile.
What do I mean by that?
If you play a strategy that is pretty close to GTO, but I know which way it’s off, I’ll take your money.
Nobody is capable of playing GTO. The definition of “close” is subjective, but in my view, nobody is capable of getting close*.
*assuming we aren’t playing heads-up 10 big blind stack poker.
And since nobody is close, all you have to do (not that it’s easy!) is understand where they are falling off from optimal strategy and how you should combat it.
Bet Sizing
One of the many ways people can mess up optimal play is with bet sizing.
Solvers have taught us that optimal poker involves a lot of different bet sizes. If we were to look at truly optimal poker – which our computers can’t quite conquer yet for reasonably sized game trees – it would involve more sizings than you could remember.
Now, there’s no real reason to solve to that level of granularity. You don’t lose much at all by using “only” three sizings in one spot instead of 15. And humans have more than enough trouble splitting into three bet sizes.
Why is juggling multiple sizings so hard?
Let’s just imagine a simplified scenario, where the out-of-position player can’t bet, and IP (In Position) has to decide to bet or check each street.
With those two decisions across three streets, there are 2 x 2 x 2, aka two cubed, aka eight potential lines of action for IP.
That’s eight separate ranges to keep balanced.
Now, let’s take the same example and let IP use two bet sizes on each street. So now, they can decide to check, bet small, or bet big, at each decision point.
With three options at each node of the game tree, they now have three cubed, or 27 ranges to balance.
Remember, if those ranges are pretty close but not quite perfect, they are going to be exploitable.
Now in the real world, there are more factors. You need to handle bets. You need to handle check-raises.
If you’re out of position, you need your betting ranges and a check-raise range, a check-call range, and a check-fold range!
And you need to be prepared to navigate drastically different turns and rivers with the ranges that you’ve arrived with.
If it sounds impossible, that’s because it is.
But this isn’t bad news. It’s great news.
It means that nobody is capable of playing perfectly. It means that everyone is exploitable and beatable if you know what you’re doing.
But How Can I Exploit Their Bet Sizing?
Let me offer you a simple and concrete example, and one that I think is representative of the way most players mess up.
You’re playing a great player in your $5/$10 game.
You raise to $25 in the HJ with A♠️Q♠️, they 3-bet from the SB to $110, and you call.
Your opponent started the hand with $1k in their stack and you cover.
Flop: K❤️Q♣️5❤️ (pot: $230)
Your opponent bets $170.
What should you do?
You flopped 2nd pair top kicker, which often feels worth calling. And in theory, AQ is a call.
Let’s look into our opponent's strategy – specifically, their bet-sizing – and decide if we should exploit them by doing something different.
Solver Outputs
If we let a solver use two bet-sizes (25% and 75%), here is how the SB is supposed to play:
Bet 100% of their range. Never check.
(Coincidentally) Bet 25% pot 75% of the time, and 75% pot 25% of the time.
Mix both sizings for most hands, with some preferences:
AA, AK, KQ, 5x like to bet big (75%) more than other hands
Sets, KJ, KT, Qx, and underpairs like to bet small (25%)
Bluffs have some preferences, but let’s ignore that for now
It might feel intuitive to you that the strong (non-set) value hands prefer to bet big and get stacks in by the turn.
It probably feels intuitive to you that QJ and JJ prefer betting small.
What won’t feel intuitive to many readers is that all of these hands use each sizing a lot.
To truly play GTO, you need to bet big with QJ and TT somewhat often. You need to bet small with AK and AA and KQ frequently.
GTO play involves that level of balance.
GTO-Studied Humans
What often happens, is that well-studied players will do one of two things:
Simplify into one sizing. Bet 33% pot 100% of the time, for example.
Use two sizings, and remember which hands prefer what.
Since your opponent bet ~75% pot, we can assume they’re using strategy #2 – attempting to balance two bet-sizes.
If this opponent still:
Successfully bets 100% of their range
Plays every unpaired hand with a perfect mix
Knows which hands prefer what sizing
…but makes the simplification of always betting the sizing each hand wants to bet, that leaves them always betting:
AA, AK, KQ, 5x for 75% pot
Sets, KJ, KT, Qx, and underpairs for 25% pot
Even though they’re bluffing perfectly, this throws them out of balance.
…and now, you should fold your AQ!
See, the solver puts every hand in its place for a reason. And it mixes bet sizings on so many hands that “don’t feel right” because the alternative is getting exploited.
It likes to bet big with 77 to get calls from 65 but fold out 99.
It likes to bet big with TT to get calls from A5, so that you don’t have an easy fold with AQ, and so that it can bluff jam on the J❤️ turn.
Your opponent is playing really close to an optimal strategy, but not close enough to avoid being exploitable.
How to Make These Reads
The type of simplification (from your opponent) described above is a common one:
Roughly remembering which hands like to do what, but not remembering to mix enough.
Playing hands in a more comfortable way – the way that feels more intuitive and is validated by solvers.
In lower-stakes games, your opponents will usually not get anywhere close. You can figure out the types of mistakes people are making based on common card sense and experience with them and other players in your player pools.
And you can use logic to work backwards from there to the exploit.
Once you’re playing against players who are well-studied, this is going to be much more challenging if you’re not well-studied yourself.
This is the struggle I had for several years.
I could make observations. I could logically deduce exploits.
But I didn’t know what they were studying. I didn’t know what they’d think an optimal strategy would look like.
To understand and beat them, I had to change my approach to studying.
P.S.
As the owner of a poker training site, I’m incentivized to tell you that if you want to win consistently, you have to start investing in your education.
But you don’t have to take my word for it!
Ask someone who has actually earned a side income playing poker for years this:
How much would investing $150 into learning save me over the next year?
