"GTO is unexploitable" is a big myth!

what did i tell you about posting on gto forums before taking your meds grandpa

Ok you sound stupid

what's wrong with you guys, do facts sting your eyes?)

Here is an example. Final table of regular MTT, two big stacks(60bb eff) confronting, HJ minr, BB 3bets very polar to 12bb:
HJ reaction is 72% fold, while shoving any hand is +EV, considering BB folds 75% to 4bet shove.
Am i wrong?
can't upload images
https://ibb.co/7tzm0hL
https://ibb.co/5kGv1KQ

What matters in a MTT is $EV, not bbEV. the solver is trying to maximise $EV, it doesn't matter that one play wins big blinds or not if it loses money on average

aner0 You're right. It's incredible, that by Chip EV you need only 9,5% eq and considering ICM it's nearly 30%, huge difference. In another example: 15bb HU SNG, SB limp-folds to jam 78% and while we can shove any two, solver says that free play has more EV with bottom part of our range, but let's be honest- you must be very good to play that garbage postflop profitably. So in total i admit, that i was wrong, in situation where it can gain more EV by just checking behind on BB- it prefers it vs shove. But if it folds-you can be 100% sure it's a fold, for example if SB minr and BB want to shove, having 62% FE, T5o is just fold.
wait a second.. yes, it's true that shoving whole range in a first example, we make a big mistake considering icm, but doesn't player on BB himself make minus EV with his overfolding to shove, isn't by definition any action performed on the GTO should be at 0 in the worst case? I meant not overfolding, but facing 4bet shove not in 5%, but in 50%

Both players are trying to maximise $EV by making bbEV sub optimal plays. Having said that, in an MTT you could deviate to hurt your opponent, for instance, you could hurt your opponent's "ICM folds" by jamming way too much, but you would not be gaining anything in return. In fact, you would both be losing $EV and gifting it to the rest of the final table that's not punting around.

This is

https://app.gtowizard.com/solutions?custree_id=e97f18d3-5b2f-4671-a299-90364952b50a&cussol_id=3c92778d-915d-4073-b8a9-722100c7956f&solution_type=gwiz&gmfs_solution_tab=ai_sols&gametype=MTTGeneral_ICM6m1000PTFT&depth=20.125&stacks=20.125-50.125-10.125-30.125-40.125-60.125&gmff_depth=100&gmff_rake=NL500&gmfft_sort_key=2&gmfft_sort_order=asc&history_spot=6&gmff_favorite=false&soltab=strategy&preflop_actions=F-R2-F-F-F-R12&stratab=strategy

GTO Wizard

GTO Wizard

MTT Final Table, HJ opens, BB 3-bets, HJ?

BB 3-bet

HJ vs 3-bet:

BB vs Shove:

Take a look at what calls you. BB is so polar, you are risking your tournament life to fold out hands you dominate, while getting called by AA/KK and AKs.

BB isn't making a mistake. They are playing the least exploitable ICM strategy. Try this exercise to better understand the spot. Calculate HJ and BB's $EV when:

• BB folds to shove
  
• BB calls shove and loses
  
• BB calls shove and wins
  

Once you put it into perspective like this, the ICM strategy will make way more sense.

tombos21 Hi, could you please explain one thing: HU SNG, 15bb eff, what if SB limp-folds to shove not 78,4% according to GTO, but 60%, do i understand correctly that BB shove range will change(screenshot of range would be great, ty), but default GTO range of BB will still be at least zero EV?
So by default it's:

Sure. I'll use GTO Wizard's HU preflop solver. We want to solve 4 pairs of strategies:

• GTO vs GTO
  
• GTO vs Station
  
• Exploit vs Station
  
• Exploit vs Counter-Exploit
  

Step 1: Solve the spot, lock every node (to prevent exploits), and measure EV at the root node

GTO vs GTO

It's important to measure EV at the first decision point (root node) because this represents the EV of the entire strategy.
Here EV is measured in big blinds. BBs EV is 0.97. Since they posted a blind of 1bb, their default GTO winrate is -3bb/100.

Step 2: Nodelock SB to call shove to wide

GTO vs Station

BB's new EV (again measuring at root node) goes up to 0.98, or -2bb/100.

Here's BB's GTO EV against the calling station:

Step 3: Unlock BB and let them exploit SB.

Exploit vs Station

Here we can see BB uses a very linear shoving range:

And EV at the root node increases. Now BB's EV is +6 bb/100

Step 4: Lock BB exploit shove, Unlock SB strategies and let them counter.

Exploit vs Counterexploit

Lastly, it's good form to always measure the counter-exploitability. This gives us a worst-case scenario and helps gauge the risk/reward of deviating from GTO.

SB limps more often and more polar, and only calls nutted hands vs BBs shove. Now BBs EV has dropped to -7 bb/100

Summary

So in conclusion, GTO gains EV against mistakes (as you should have expected).
Exploiting potentially gains a lot more, but you risk getting countered if you're wrong.

If we assume SB is either a station or will counter us, then you need about 38% confidence in order to make this exploit. So it's a pretty safe exploit since in reality they won't fully counter even if they know what you're doing.

That’s a very nice walkthrough!

What’s the underlying logic of the required confidence part? It’s the ratio of the potential gain to the overall interval of bb-deviations. But why can I interpret that as a required confidence in my read?

Thanks!

The confidence thing is just the implied probability of an 8:5 bet. So if you bet $5 against my $8, you need to be right 5/13 = 38% of the time to break even.

Or alternatively, imagine you bluff $5 into an $8 pot. You win $8 if I fold and lose $5 if I don't. So you need me to fold at least 5/13 of the time.

Similarly in this case, I'm modeling it as a two outcome problem: either they station and you gain $8 (relative to playing GTO), or they counter and you lose $5. So you need to be right at least 5/13 times.

If you 38% confident he is a station and 62 he is the counter, then it would make sense to model BB strategy as linear combination of the two strategies. Which would result in less extrem SB exploit and that would also be harder to detect.

If someone wants to scream from the mountain tops that GTO doesn't work and your a for profit player wouldn't you be better off to encourage them? The more math and theory is for nerds in the games the softer the games are.

From a game theoretical viewpoint it is in nobody's interest for anyone to get better at anything.

If you crash on a desert island and you're the only survivor who can swim you don't tell the others about it. That's your superpower and God wanted you to have it.

Its not that I'm against people helping others just saying the more GTO is for Dweebs types the better

Unless theyre right and were trying to trick them heheheheheh

tombos21 thx man, great post!
I would like to ask about that exploitative 30% push vs calling station. Did Solver removed non-allin isolates itself or did you to simplify the picture? Why it's so much difference between gto and exploit ranges vs station, why even monsters don't slowplay? Is there a gradual transition in BB reaction, when SB's limp-fold decreases from 78% to 60%?
According to my calculations, plus EV shove will be with a much wider range, but as I have already found out for myself, solver often considers freeplay with bottom more profitable. Again - will it be like this in reality? According to my observations, at micro limits, many regs bet flop at nearly 100% rate even OOP in limp pots. It becomes a bigger problem, when you play 10 tables and you have to make a decision in a couple of seconds. It's a pity I deleted old database, it would be interesting to look at winrates of groups of weak hands in check on BB. But what I was able to compare - was winrates of most garbage hands in check and bottom of the shove, they turned out to be about the same. So most likely shove hands would have shown better results in check, as GTO claims.

I think OP could be right in MTTs, but never in 2-player chipEV

Say there’s an extreme bubble spot where SB is shorter than the BB and should only open shove TT+

Because their shoving range is so tight, BB has to call off really tight, maybe QQ+

Then it might be +EV for SB to shove ATC because they get such high folding freq. It can’t be the equilibrium because as soon as you start doing it BB adjusts and calls wide. But against a BB player who blindly follows the chart calling QQ+ shoving ATC is higher EV

charlesChickens I'm confused..
Let's take TT+ shove and QQ+ call as equilibrium.
1)if SB shoves ATC it's +chip EV(if BB doesn't adjust) or zero Chip EV(if BB adjusting accordingly), but minus in both cases $ EV for him.
2)if BB doesn't adjust - it will gain $EV? Or, as aner0 mentioned, both give their $EV to other players? But the latter is contradicting with definition of GTO no(playing GTO equals at least zero EV, no matter if it's Chip or $)?Wait.. he makes minus ChipEV, but +$EV right?
3)if BB adjusts - he gains $EV and more than, if he's not adjusting? He will be zero Chip EV?

No, that's not what that means. You have to choose your utility function. If you value chips according to ICM, then all the solver cares about is $EV, and ChipEV is no longer a consideration. If you value chips according to Chip EV, then all the solver cares about is maximizing the accumulation of chips.

Well, GTO is never guaranteed to win or breakeven multiway, and MTTs are a very exaggerated version of multiway where one play affects everyone else even if they already folded.
So truthfully, GTO is exploitable, but not for the exploiter to gain anything in return, more so you could get "griefed" by another player immolating himself