When to Avoid a Solver’s “GTO” Plays

I was just listening to a podcast with Doug Polk and Andres “Educa-p0ker” Artinano. This quote from Andres stood out (paraphrased):


Thursday January 01, 1970

Solvers are very good for studying the preflop and flop strategy, but not so useful for studying the turn and the river.


Thursday January 01, 1970

Andres is basically saying that it’s less important to play theoretically sound poker (or at least a solver’s version of theoretically sound poker) on the later streets, which got me thinking — why is this the case?


Thursday January 01, 1970

What is “GTO”?

GTO stands for game theory optimal. Solvers have helped us learn more about the GTO poker strategy, but the true GTO strategy is not yet known (which is why GTO is in quotations in the title).
You can learn more about what is known about poker game theory here.
 

Why is it More Important to Adhere to Theory on the Early Streets?
It may seem counter intuitive, but the optimal strategy is more complex the earlier you are in a hand.
This increased complexity is derived from the immense amount of outcomes possible at the start of the game tree. 1,326 preflop starting hands, 19,600 different flops, each flop with 47 possible turns, each turn with 46 possible rivers, over 50 different preflop formations (BB vs BTN, SB vs CO, etc.), a huge number of possible stack sizes (especially in tournament poker), antes or no antes, any bet size up to all-in can be used at all points, and more.
In other words, there are millions of different “roads” that we can take but only a handful of them will lead us to win. This plethora of options makes it extremely hard for us, as mere humans, to determine the correct preflop strategy. Thankfully, some tools have been developed to help us approximate correct preflop and flop ranges (PIOSolver and PokerSnowie).
Going back to the question, it’s most important to have theoretically sound ranges on early streets because otherwise we may find ourselves in losing situations — on a road peppered with landmines.
Additionally, our opponents will naturally have more experience playing early streets than the later ones (simply because they occur more often), so we should expect them to make fewer fundamental mistakes on the earlier streets, and thus it is more important for our strategy to be theoretically sound.
I think a couple examples will drive this point home. Shall we?
Preflop Example
Consider the hand QTo. If you are just started out in poker, this might seem like a good hand to play from middle or even early position. Two broadway cards, it can make straights — why not?
Well, after mass database analysis and, later on, PokerSnowie simulations, we know that this hand is not a profitable open-raise because doing so leads to many bad outcomes in the later parts of the game tree (such as losing a bunch of money with a dominated top pair).
(NOTE: Want to take the guesswork out of your preflop game? Get 259 preflop charts for almost every situation when you join the Upswing Lab. Learn more now!)

Flop Example
There are so many combinations of hands in each player’s range, so many different turns and rivers that can come, and so many different possible stack-to-pot ratios that it’s extremely hard for to fully grasp the implications of betting or checking certain hands on the flop. We as humans cannot see that far ahead.
This is where software can help us because they can simulate the situation 40 times per second and, if you leave it for a few minutes, it will have played out that battle 10,000+ times for every possible turn and river card. They know which line is preferable for each hand, sometimes for reasons that are effectively impossible for us to deduce, because they have seen what happens at the end of the game tree over and over and over again.
We need to be humble enough to accept our limitations and try to understand why the solver’s solutions look the way they do.
Here’s a flop example that immediately came to mind because of how surprised I was at the solver solution: suppose you raise preflop with J♠ J♦ from the button, the big blind calls, and the flop comes 8♥ 5♥ 4♦ (100bb deep). When the big blind checks to you, what would you do?
I am going to make a wild guess and say you’d c-bet, because that’s what I would have done too.
So, let’s plug this situation into PIOSolver to see its solution (I’ll summarize the solution below the image for those of you who don’t use solvers):
JJ and our specific combo of JJ are highlighted in red. Click here to enlarge the image so you can see the details.

Click here to see a simplified breakdown of the solver’s strategy

C-bet 28% of the time and check back 72% of the time.
High frequency value bets include 99, strong top pairs, sets, and straights.
Most bluffing hands contain a 7, 6, or backdoor straight draw, but even those should be checked back at some frequency.
High overpairs (JJ+) and weak top pairs are very high frequency check backs.

The solver wants to check with our specific combination (J♠ J♦) 96% of the time. We can speculate why:

J♠ J♦ doesn’t benefit much from protection as the only overcards are A, K, and Q. This is why you see the solver betting more often with the lower overpairs and less often with the higher overpairs.
J♠ J♦ doesn’t get to extract three streets of value on most run outs.
Getting raised sucks because we’ll be up against a range with many strong hands and there will be a lot of bad turn cards for our hand (any heart, 7, 6, 3, Ace, King or Queen)

Note that the JJ combinations that have a backdoor flush draw are bet at a higher frequency. This is likely because they have slightly higher equity and it sucks less to get raised while holding a heart.
Why is it Less Important to Adhere to Theory on the Turn and River?
If we use the same logic as in the previous section, it’s easy to see why we can rely more on our intuition on later streets:

Ranges are tighter, so it’s easier to accurately assess our opponent’s range.
The stack-to-pot ratio is lower, which decreases the complexity of the situation.
When on the turn, it is far easier to estimate what will happen on the river since there are “only” 46 cards that can come (as opposed to 2,162 combinations of turns and rivers).
Players strategies are not as sharp or balanced as the far more frequently played preflop and flop spots.

Let’s take a river example where you defended from the big blind against your opponent’s under-the-gun raise. Your opponent goes on to fire three barrels on A♥ Q♠ T♠ 6♦ 8♠.
In this river spot, it’s extremely difficult for your opponent to be bluffing because almost all of his possible semi-bluffs (flush draws, J9, etc.) got there. Unless he is a very high-level player who planned for such a situation on earlier streets, he will very rarely show up with a bluffing hand.
Of course, the solver will reach this river with enough hands to bluff with, as you can see in the picture below:
The hands highlighted in blue are bluffing hands that most humans would never reach the river with. Click here to enlarge the image so you can see the details.

Click here to see a simplified breakdown of the solver’s strategy

Bet 40% of the time and check 60% of the time.
Value bet with flushes, straights, and some sets.
Bluff at varying frequencies with KTs, JTs, T9s, 99, 77, and 55.

Due to its extremely developed visibility across the game tree, the solver was able to reach this river with relatively many combinations of hands that can profitably bluff. You can see that it bet twice with some pocket pairs (55, 77, 99) and some flopped bottom pairs (KTs, JTs, T9s) which it now bets at some frequency.
We, as humans, do not understand poker this way and would likely not reach this river with so many potential bluffing hands. That’s not to say we should try to play exactly like the solver (that’s basically impossible), but we can follow the solver’s lead by mixing in some early street bets with hands that will become profitable bluffs on certain run outs.
So, if you were playing against a solver (or solver-like player) on that A♥ Q♠ T♠ 6♦ 8♠ board, you could call on the river with some bluff-catchers. Against the vast majority of players, however, you’re better off exploitatively folding all bluff-catches.

Editor’s Note
An additional point is that the solver’s strategy for the turn and river is built on the strategy used on previous streets, and it is fairly fragile.
This means if our human opponent deviates from the solver’s strategy on a previous street (which is more than likely in most situations), the solver’s river strategy will probably be a losing one because it is essentially relying on incorrect assumptions about how our opponent would have played his range up to that point.

So, Should You Completely Disregard Theory on the Turn and River?
Most definitely not. If your opponent is a good player, you’ll need theory in your back pocket to stand a chance at beating them in the long term.
Against weaker players, you still need to understand and consider theory to accurately assess how they are deviating from what is theoretically correct. Armed with that knowledge, you will easily be able to implement counter-strategies that win.
On the other hand, if you don’t know what is correct, you will have a tough time identifying and countering your opponent’s mistakes.
That’s all for this article, guys and gals! I hope it lit up some light bulbs in your mind. If you have any questions or feedback leave a comment down below and I will answer as soon as I can. (Which might be a little while because the comments have been buggy lately — we’re working on a fix.)
If you want to read/hear Doug Polk’s take on “GTO” vs exploitative play, check out this article.
Till’ next time, good luck, grinders!
NOTE: Want to learn high-level poker strategies without having to work with solvers? Let our world-class pros do the work for you in the Upswing Lab training course. Learn more now!

.fb_iframe_widget_fluid_desktop iframe {
width: 100% !important;
}

The post When to Avoid a Solver’s “GTO” Plays appeared first on Upswing Poker.

Comments are closed, but trackbacks and pingbacks are open.