OK, I have no interest in a strawman argument and don't care about the travellers dilemma.

Here is what solver wants to do.

Check flop - unless you have the nut spades then bet (AA75)

We mess up and bet the flop, it says, ok, fine wtf. On this turn, check behind unless you have spade or clubs, then pot it.

We get it right. Villain pots it, representing the nuts, or air, we have AA75. We have four combos it wants to call with at approximately 0 EV. The rest it wants to split between 75% folding and 25% raising. So, with this action, it says, hey, you want to raise the nuts you better have some bluffs, here is a combo that is a bluff 22% of the time you get to this spot. It makes your overall EV higher. Solver says you have better bluff catchers here. It wants to call with a lot of the AA hands, but not so much with this one... probably because it has so many combos it is calling with that it decides not to use this one. I'd look at the output and say calling probably isn't that bad of a play and is approximately 0EV. Which means here is where we can look at solver and turn this into exploitable play. Calling with AAxx is approximately 0 ev if villain is bluffing witht eh right frequency... but if he isn't bluffing enough we have an easy fold and if he bluffs a lot we have an easy call and if he is right.

You disagree, ok, fine, with what? If you aren't bluff raising this then what are you bluff raising and if you aren't bluff raising then fine - that's a different game.

Sorry, what strawman

That one.

A little combative gentlemen but I feel like it’s a pretty productive thread. ❤️

My goal is not to piss you off. If I'm doing so inadvertently, I can't fix that unless you be a bit more specific. Are you objecting to my claim that

'Disagree with solvers? /shrug - ok'

Is a fairly casual but obvious way of placing solver output as above my questioning?

Just to repeat myself, it's not that I object to solvers or game theory, I'm a huge fan of game theory, I'm just worried that there's an undue faith in the idea that the solver output is always correct, even before we get to the question of whether it could be useful for our games.

MarkD i simmed something like CO open vs SB flat with 100bb stacks, since SB flat shows a strong but capped hand, consistent with Hero read on villain. Using FlopHero.

I assumed villain could call some 1p hands, especially KK, vs bluff-raise but I didn't review that particular node for villain after Hero's low-frequency pot-raise.

Yes. You are saying you "disagree", but you didn't say at all what you disagreed with but you are very clear you disagree with solvers.

Assuming no bugs in the software, and that they are doing some form of counterfactual regret minimization, they are correct, for the inputs given them. That doesn't mean they are best for a given game type that can be exploited. I firmly believe they provide an intelligent insight into how situations can / should be played if our opponents were playing optimal and thus we can learn a lot from that. Not limited to the fact that we can look at what an opponents optimal strategy would look like for a given situation and then we can ask questions like - does the population act like this? What about this specific opponent? No, the opponent will always take x action... oh, so then what? With software like FlopHero, you are severely limited in these types of questions as you can't really node lock a strategy for the villain in quite the same way.


I'm not upset or angry, but I also don't have time to go tit for tat over whether or not solvers are the cats' pyjamas. You have an opinion on them, which I respect, and that's it. I will be unlikely to engage in discourse like this in the future as it feels unproductive.

I said several times that I specifically disagree with the given output. Which means that either the inputs were wrong, the solver is not doing what it's supposed to be doing, or I'm wrong.

'Assuming no bugs in the software' is a mighty big assumption, and bugs aren't just missing a ; here and there, it's not doing what it's supposed to be doing. That is my claim: we don't know for sure that it's doing what it's supposed to be doing. When it gives highly counterintuitive results, such as jamming this river with one pair for value, that gives us an opportunity to find out: are we just bad at using it, is it wrong, am I wrong? I think that's highly productive and gives us the best chance to improve our game by finding out how solver works.

If you don't find this productive, perhaps next time don't just throw out accusations of strawman and shrug when I question an output.

It’s not raising one pair for value. I’ve said this many times. The results here are not at all counterintuitive as I’ve also shown. They are very straightforward and intuitive but you are choosing to focus on one comment in this thread because that comment supports your your believe system.

Belief system? Really? Now who's strawmanning? My belief system that we should be careful using technology whose workings are opaque, and that we should pay attention to when it comes up with counterintuitive results? If you're saying that you've used better inputs than Monikrazy, and thus we're never expecting to be called by a worse hand, and that therefore this is a pure bluff, cool, this feels resolved to me, but I don't feel like I made an error in highlighting that this might be wrong. Do you want me to stop doing that?

I don't know if this helps or makes the discourse worse, but here's another data point. The hyper-aggressive solver I train with is Omaha Poker Training.

If it's hero in this hand, it will turn AAxx into a raise-bluff fairly often on the river. It tends to raise-bluff AAxx a lot in other situations too. It doesn't like to fold AAxx.

If it's villain in this hand, it's going value bet a lot of two pair hands on the river because the turn was checked through. It will either bet or check-raise sets and straights.

As villain, it will bluff almost everything else. Every one pair hand except maybe KKxx will be turned into a bluff. This includes AQxx.

As villain facing our raise-bluff, it'll never fold T8xx or sets. I've seen it call super light with one pair and blockers sometimes, but not always.

When the solver yells at me to raise-bluff my AAxx in training, I ignore it in situations like this. That's because solver is a stubborn idiot.

But if I'm playing against me live, I (as villain) would value bet two pair on the river (I'd bet two-thirds pot with my whole range). Knowing that, I (as hero) would raise-bluff with AAxx sometimes (but not too often, lest villain has an easy call with two pair).

As someone who is working with monkersolver for more than a 5 years (since it came out basically) and have gone through hundreds (maybe even thousands) of sims, I dont think I have ever seen results like that on the river unless I nodelocked or fckd up the sim in someway - bad ranges, bad settings, bad gametree etc.. (talking mainly about the river).

My guess how solver would play the hand - even tho its very hard to come up with preflop ranges as its a very awkawrd spot for a solver with a CO limp - is that we would bet the flop with like 30-35% range IF the OOP player has no leads in his range. I think if solver would play this spot he would have like 20-25% leading range in this spot and therefore IP can bet like 40-50% of range maybe? This hand in question has only one real component to be betting and thats a 5d (maaaybe Ac could be argument for low freq betting) so my guess is with this hand having no spades looks like pretty clear check with this shallow SPR. Jc improves OOP calling range a ton so wouldnt be surprised if he should have some leads on this card so check with your hand is very reasonable. And the river spot - unless you have insane read on your opponent I would just instafold and wouldnt think about it that hard - your range defends quite well on this river as you still have bunch of Q5xx, Q3xx, J5xx, J3xx that checked the turn, some straights and all the other AA with good blocker properties.

And now the river in solverland. If you would simplify the river to 2 sizings for OOP (33% and 100% - which are usually the highest EV options when you start simplifying) and you give IP player option to raise pot only (which is very reasonable) and IP has option to bet 50% and 100% (once again usually the highest EV betsizes when you simplify). In this game my guess is that solver would bet big part of his range for 33% with weak 2 pairs some straights blocking twopairs etc and ofc some bluffs with no SD value. And the stronger hands that dont block bluffcatchers for full pot (my guess is like 15-20% range bets full pot here). And my guess is that IP against a full pot polarized bet (set/strong 2p+) raises all in quite infrequently like maximum 10% of the time and I am quite sure this hand is just not there. My guess as a bluff you need to have a K, T or an 8 and my guess is that you need atleast 2 blockers to raise and not only one - so your hand AA57 has literally 0 blockers for a bluffraise and would just be burning money. These are my assumptions how solver would play the spot. But let the assumptions be assumptions I am gonna solve it and probably will have the results by tommorow/this evening.

Solving this spot is quite tricky because of the preflop so I used 50bb btn open (36.4%) and SB coldcall (5.5%) with PLO500 rake structure that I think could be close enough (only thing I dont like about these 2 ranges is the wider btn preflop range but lets work with that).

First set of results are here (flop sim) and there is one surprise - for me atelast - OOP is supposed to lead around 37.7% (with these sizings in play).



IP player should bet around 40% when OOP checks and have leads in his strategy.



And like I thought AA57 with no spades have no bussiness in betting - even some combos with As checks ( mostly with bd fd as its kinda standard - you want to realize that turn equity).



So since its probably not realistic that in real world that people are leading over 30% of range on these boards (unless OP can say more about his games) I am gonna run it once again with no leading range and see what exploits can we make against these players - my guess is its gonna be close to range check. And then I am gonna proceed on turn probably with the sim that has leads because if we are range checking (or close to) there is no point using these ranges.

Alright got the final results (was faster than expected because of the shallow stacks).

On the flop if OOP doesnt have leading range. IP player can exploit him by checking back significantly more - goes to 80% check and 20% bet - and by doing that IP player gains 1.5% of the pot which is massive. Therefore I used the sim where OOP leads and included your hand.

On the turn as I thought Jc smashes OOP range - he actually gains range advantage (56.3% vs 43.7%) and therefore as expected should lead quite often - 37% of the time and IP should bet 43.9% of the time. On this turn players want to just use pot size bets everything else was just lower EV - which make sense on such dynamic board with these SPRs. AA57 as expected is an absolute punt bet losing heaps.



On the river. As I expected doing anything else than folding vs a pot size bet with AA57 is a huge mistake (its a mistake to not fold even against 33% bet). I ran sim with 33% and 100% for OOP and pot betting range is tighter than expected. Overall freqencies are 24.1% bet with 1/3rd and 9.2% pot and its potting for value just T8+ (there are like 1 or 2 combo of sets and thats it). IP player responds with bluffing exclusively every single combo of TT that is <AA and sometimes bluffs with combo like KKQ2 but its more likely to call with those, all KK8x combos pure call and AATx is 100% pure call, AAKx mostly folds and sometimes calls (folds 62% and calls the rest). Interesting part is that OOP player when he bets pot and gets raised allin he even folds some T8 combos (and no OOP is not calling worse than T8). Also IP is folding 58.8% calling 28.9% and going allin 12.3% against pot size bet and against this OOP strategy.





Lets move onto the next river sim I ran and that is with only pot size bet for the OOP player or check (EV difference is 0.2% - so its like whatever). If the OOP player is allowed to bet only pot or check he bets 24.6% of the time for full pot and checks 75.4%. Now doing anything else than folding AA57 is even bigger mistake than in previous strategy. The big difference is now that OOP gets to bet some 2pairs and some sets for full pot because he wants to extract value and he cant bet smaller. IP defending range changed a little bit - AATx is now calling only 50% of the time instead of 100% and other combos that are blocking some 2pair hands are calling instead. IP player is still bluffraising basically every single combo of TT expect when he has 2p (thats calling) only difference is now you are also bluff raising all AATT instead of calling.

So my doubts on the river were correct that you just cannot call 1 pair after being raise (probably something messed with the tree/settings - would check if the OOP has a fold action available). And that raising or even calling a hand like AA57 is absolutely horrendous. If I made a mistake or made wrong assumptions feel free to point it out and I can check it out. Only thing that is still sketchy is the preflop range but I was trying to select two preflop ranges that I would see myself playing with these shallow stacks.

Necromadx thanks for sharing.

I was using 100bb for my sim which is probably wrong given the straddle - let me look at FlopHero again and see if I can do shallower.


I agree with giving villain a SB flat calling range, though maybe giving Hero BTN range is a little too wide - not sure how much solver solution will change if it is a MP or EP open vs SB instead.



Edit: ok re-ran the sim at 50bb stack and flophero is now folding 73% of the time and all-in 27% of the time with this AA combo. But still supposed to checkback the flop.

Sick thread

I posted last night but I guess I didn’t send it. The only explanation I can think of is that FlopHero is using an actual range, where AA57 never gets to the river this way, and a range with 100% of hands and within this second range (which is sort of useless) we are seeing this hand as a bluff range.

I’d be curious to see the river spot analyzed with moniker assuming a 100% range.

As monikrazy suggested and I was also curious, I did sim it again with much tighter iso raising range - I used CO 40bb 18% openraising range in low rake (PLO5k) enviroment. I think this range might reflect it more given the fact that there was also a BB complete with only CO openlimp I think raising about 35% should be fine tho.

Flop is very very similar to the previous solution - ranges are still pretty much 50:50. OOP player leads around 35% of his range, IP player bets 43% instead of 40% of his range and responses to bets are almost the same.

Turn plays the same - OOP leads around 38% and IP bets around 44%. Only difference is that now OOP player has only 54% range advantage instead of 56% on the turn.

River is now significantly worse for the OOP player - only 53.7% range advantage instead of 59%. But in a strategy where OOP has two sizins (33% and 100%) - solver likes to pot a little bit more 18.2% instead of 9.2% and 1/3rd range is just the same - some hands. Now solver likes to trap a little bit less - bets more of KT, T8 and more sets with blocker combos for full pot + more bluffs so some of his checks went to potting range. Response to this pot betting range is very similar - our particular hand is still a huge punt if not folded (call loses $36 instead of $55 and allin is losing $100 instead of $140). Also OOP bets 30% of JJ and 99 (that doesnt have KT or T8) instead of only 1% and he folds only 50% of those to allin and when he bets T8 and get raised he also folds only 20% of those instead of 57% (plus he bets twice as often with T8 - 30% vs 60%).

River with a strategy of only potting or checking for OOP player the same principles are applied. Solver likes to bet a little bit more 32.1% instead of 24.5% and its just less trapping and more merge betting with hands like sets, two pair with excellent blockers and T8 + some bluffs to balance it out. For example 99 without a straight bets 50% of the time instead of 12%. Response to this bet is the same - bluffing exclusively TT and value raising only KT. Bluffcatching with almost all AAT hands for example (folding most AAK) and bluffcatching a lot of naked KKxx etc..


EDIT:

Monker is not on the river with AA57 no spade either given our line - checks all of those combos on the flop. But I just added our combo manually to see if there is some way we do anything else than what the human logic says and that is just check the turn because we have bad hand and fold the river because we dont have relevant blockers and monkersolver just agrees.

Not sure what you mean by 100% of range. Like if we ignore flop and turn action? Even then AA57 is just a bad hand on this board so unless you have some sick liveread there is no way this hand does anything else than folding.

I said it poorly, what I mean is a range of 100%. 100% of hands.

You added single hand to a refined range and monker said - I don't care, this is still bottom of my range and I'm going to fold it. FlopHero is suggesting a bluff raise some percentage of the time, but I think it is acting based on a range of 100% when it says this.

Not sure how that helps analyze this spot. But if I gave IP player literally 100% hands (every single combo in the game) - AA57 is still bad hand and a very minus EV call and performs three times worse as a raise. Whats kinda interesting is, that now we are not bluffraising every single combo of TT because we would be overbluffing but we are just calling combos like TT42 with three spades and no club (high EV call actually). Also AA57 is just a purefold even against 33% betsize. I guess this just shows how important in PLO is to have good blockers when bluffcatching rivers and not just go by hand strength.

It doesn't - it was only to help understand the suggested output from FlopHero.

on boards like this flop if we bet pot instead on flop do we basically surrender the hand? i mean to ask what turn cards are you 80 percent or higher betting here? thanks alot