BigFiszh's avatar



2277 points

Comment | BigFiszh commented on GTO+ v PIO RAM needed

It depends on the tree size (obviously). A smal river tree with shallow stack sizes can be solved with 4 GB. A huge flop tree with broad ranges, deep stacks and multiple betsizes can consume 60 GB and more.

July 30, 2021 | 6:53 a.m.

July 18, 2021 | 8:11 p.m.

Comment | BigFiszh commented on GTO+ solver

How many sizes do you use? On the flop, the tree size grows over-exponentially with each additional size, but 2 sizes should be okay with 16 GB.

I've just tested it with Pio, large preflop ranges, two sizes on the flop results in ~4.5 GB, three sizes lead to ~12 GB (which is probably too large as a good portion of the RAM is not available due to system occupation.

June 30, 2021 | 4:52 p.m.

a) What is that?! It makes no sense to calculate an EV that is higher than the pot.

b) Ignoring the pot, the EV-difference between the options might be small enough that at the current iteration it's still treated as equal. A solver does not switch to the highest EV-option but it slightly increases the frequency because otherwise it would never reach the Equlibrium but toggle between extremes.

June 27, 2021 | 7:21 p.m.

A-high-boards simply are very special case as almost half of Villain's 3-betting-range contains an Ace.

Combined with the fact that there will be 2 more betting rounds, it's pretty likely that Villain runs out of bluffs on the flop.

Summarized there's no need to bluffcatch with a hand that only has 2 outs and is severely threatened by getting drawn out on later streets.

June 25, 2021 | 11:28 a.m.

I guess, he checked.

As for the action, only reason for betting would be protection. "Protection" in and off itself is not existent as a reason to bet, so it should be +EV (compared to xb). It would be +EV, (for instance) if Villain would always fold (no x/r, no river-bluffing).

In all other cases (Villain is waiting for a x/r, Villain x/c with a fraction of made hands, Villain would bluff the river if a brick / overcard comes), it's most likely better to xb. We preserve our EQ against the threat of a x/r, Villain only has ~25% EQ (with overcards) and we have a great bluffcatcher for the river. Summarized there's too little value in betting.

But in reality it's likely close. If Villain is prone to cbetting (with most of his missed overcard range) and is one-and-done after getting called, it's probably better again to bet (for the reasons mentioned).

June 24, 2021 | 7:31 a.m.

* moved to mid stakes *

20k is a rather small sample, especially for the less frequently happening spots on turn and river. Gap between VPIP and PFR seems quite wide, WSD is a bit low compared with low WTSD, but probably caused by VPIP/PFR-gap.

June 11, 2021 | 12:50 p.m.

0.35% is overall, it could well be that the delta EV for this special combo is still > 2%, meaning, if Hero called 100%, it could be way better for VIllain not to bluff as much, which would put Villain to an advantage again. I'd try to go even further down, to 0.1% and see if there's a tendency (of either EVs getting closer together or the call% going up).

May 30, 2021 | 7:26 a.m.

Where's the overbet on the river? And yeah, post pictures (use the little picture icon on the very right of the comment bar), that makes no sense.

May 29, 2021 | 11:54 a.m.

Regardless of the poker stuff, you publicly admit to have planned to lie to the government? Who - do you think - will pay that money for you? The people around here (located in your country) you're asking for help?!


May 27, 2021 | 7:48 a.m.

I have no problem with the flop line, but the motivation "to simplify my turn decision" is just plain bad.

The turn jam is atrocious. Even against the range you described it made no sense, it just sounds like you were scared of the upcoming second flush draw and wanted to end the hand. That is no sound strategy. ;-)

May 25, 2021 | 7:51 a.m.

Comment | BigFiszh commented on Having nightmares

If his range were acctually narrowed down to 33, QJs, you can fold. It's easy as well to define a model where calling down (and folding on A, 9) is +EV as one where calling is -EV. "Unfortunately" neither scenario shows a big margin (in winning or losing), that means, there's no real favorite; even if we give him a higher chance of tending to the tight range, it's still a close decision.

That said, as the winning margin is very small either, I'd tend to give him credits and just fold - instead of playing guessing games. There's really nothing to win big in this situation, you did not invest that much and the situation is so special so there's no real issue with folding.

May 21, 2021 | 4:05 p.m.

Nothing strange about it, it's just the concept of multistreet-bluffing. Obviously, his bluff-range should be (or is allowed to be) much wider than 3:1. How much exactly can be calculated, I once created a video on that topic (different platform) where I explained that in great detail. Would lead a bit to far here though ...

May 21, 2021 | 1:31 p.m.

Yeah, sorry, I mixed it up. :D It might be the other way round, AQ (that dominates us) and KQ make a bigger fraction of his range when we have KQhh, so less value for betting it (hence being used as a bluffcatcher after xb).

May 7, 2021 | 6:51 a.m.

When we have KQss or KQdd we block the potential Qx-combos he was calling the flop with (that we could now get value from).

May 6, 2021 | 2:23 p.m.

The answer should be quite obvious, at least because as he can't know about your expenses. The bankroll is purely due to the risk of ruin (while playing), not for keeping you alive while you have no income.

May 4, 2021 | 7:34 a.m.

Comment | BigFiszh commented on Pokersnowie

enough of your hands have enough clean equity vs range

Obviously I talk about situations where we are indifferent. ;-)

April 24, 2021 | 12:09 p.m.

Comment | BigFiszh commented on Pokersnowie

defend at least MDF

Beware! :D Don't get carried away by the term "minimum". We should defend EXACTLY MDF, so it should better be called "ODF" or something like that. More than MDF is as bad as less.

April 24, 2021 | 11:35 a.m.

Comment | BigFiszh commented on Pokersnowie

BTN can raise any 2 cards [...]

This is neither (100%) correct, nor it's not relevant. There are many, many more factors that impact the defense frequency, partly positive, partly negative. For example you got to look at the entire BTN-strategy, he paid money to get into that exact spot, so being able to "raise any2 in one isolated situation" does not help BTN at all. Furthermore, MP will have a calling-range which means, his defense frequency has to increase ... and so on and so forth.

By the way, nobody claimed that Snowie "is" GTO, I just stated, that they use a method that is able to approximate GTO. Obviously - and that should be self-explanatory - Snowie HAS to be far away from GTO, otherwise the game would have been solved (at least for standard situations / sizes), which is not even close.

And when you deviate from the Snowie-defaults, things will get really bad, so that's no real senseful idea anyways.

April 24, 2021 | 3:56 a.m.

Comment | BigFiszh commented on Pokersnowie

Algorithms like CFR do have guarantees on this kind of thing [...]

As far as I was informed, Snowie was based on CFR as well - at least when they started, maybe they changed over the time. In that case I might be wrong ...

April 23, 2021 | 7:16 a.m.

"loose-passive" and "get bluffed in that instance a good amount" [in a 3-way pot!!] somewhat contradicts each other ... :)

April 23, 2021 | 7:12 a.m.

Ah ... okay. Yes, indeed I use the postflop check-down-option with EQR limitation to speed things up.

It would be possible to create the entire tree from preflop to the river (as Pio does it in the background) and then solve with CREV, but in this case I simply cut down the tree for the call-options (90% EQR for the small 4bet-call, due to the low SPR and 75% EQR for the call of the 3-bet). But even with 100% assumed EQR for SB when 4betting small and getting called does not change the picture, so pretty sure that EQR is not the dominating factor.

April 22, 2021 | 11:11 a.m.

Against "loose-passive" guys I'm no real fan of cbetting - given it's 3-way. Kd is no game-changer here ...

April 22, 2021 | 6:30 a.m.

It's like a manual version of a solver ...

CREV has an implemented "full-functional" solver (that works for preflop as well). Pio is faster, because it's designed simply for solving (incl. brutal memory-optimizations and such), but especially for small models (like this), CREV is lightning-fast as well.

How do you get the equity realization?

What do you mean?

April 22, 2021 | 6:28 a.m.

What preflop solver are you using BTW?

Actually it's the CREV solver ...

April 21, 2021 | 6:26 p.m.

You are saying it's difficult for him to design a strat that is +ev once we use a 38bb 4b and he calls, given about a PSB going to the flop?

Even worse - I found it difficult to design a +EV strategy for Villain right once we 4bet!!!

Are you suggesting that it is better to implement a 4b/fold a ton of BB strategy rather than playing jam/call/fold?

In theory, every simplified strategy loses EV (with very rare exceptions).

I have to admit though that I did not put much effort into that study. I just quickly constructed a model and gave it a try. I now designed a model where I gave Hero the option of either shoving / calling or 4b / call and let it solve by the GTO engine. Now, it seems that the solver does not really like the 4b at all, he simply converts to shove/call/fold, which contradicts my previous observation.

That said, now I'm sparked! :D I will have a deeper look to see where I was / am wrong. I'll let you know.

Sorry for the confusion all around, obviously (and completely surprisingly) I am not perfect. :-D

April 21, 2021 | 3:51 p.m.

Comment | BigFiszh commented on Pokersnowie

Snowie is not mimicking human neurons, it actually works similar to a solver. Only difference is that it did not solve single situations (in short time), but it solved "the entire game" - which obviously is impossible, so it approximates the "solution" over years.

Technically there's no difference.

April 21, 2021 | 7:26 a.m.

Meh, please forget what I said. :D I completely messed up the numbers ...

Given that - and with the precondition that I did not mix things up again :D - even though Villain gets very good odds it's pretty complicated to design a strategy that is +EV for him.

Even if we told him that we would open-shove ANY2 on any flop once he calls, it's better for him to fold all but the premiums (and not call, but shove those).

April 20, 2021 | 1:54 p.m.

Count how often that situation occurs, calculate the loss in winrate and multiply. Then you know how much of your winrate you sacrifice.

PS: Obviously you got to do that for every situation you'll simplify.

April 20, 2021 | 10:15 a.m.

Technically spoken you would not be committed with a small 4-bet, you could (better should) still fold A5s against a shove (when we assume KK+, AK).

But you offer your opponents of 5:1 - which makes almost ANY2 a call, even if he would expect you to blindly shove any flop (and V. calls with any pair+).

I modeled it in CREV and it was actually close to indifferent for Hero between 4b-shoving (a wider range) and 4b/folding + calling in combination with shoving the flop.

Summarized it comes down to where you think Villain will make the bigger mistake: will he fold too often against your shove? Or will he refrain from calling any pair on the flop? I'd say, making a mistake against the small 4bet is almost impossible, it's just "button clicking" (or in live game "throwing chips") without any thinking.

From my gut I'd stick to a shove/call-range.

April 20, 2021 | 7:05 a.m.

Load more uses cookies to give you the best experience. Learn more about our Cookie Policy