GameTheory's avatar


3949 points

Feb. 8, 2018 | 2:28 a.m.

1) No
1) It is not typical calculus, it is a different kind of math. Closer to logic. If you fear math and logic you will need to work hard to understand everything. You will likely understand a bit the first time you read the book, and more the second time etc.

Jan. 13, 2018 | 1:53 a.m.

Jan. 4, 2018 | 12:39 a.m.

@6:49 you get checkraised large by jts80. His reasoning for the big checkraise sizing is:

Let's have a raise and see what this fishhead wants to do, he never ever folds to a checkraise and I have a good hand. We're probably gonna win this hand somehow.

GitLOL at his russian finest v's jts80

Dec. 5, 2017 | 11:10 p.m.

We will have to agree to disagree that .5% frequencies on river nodes are worth spending resources on to see if they're real, and then if they are real that they're worthwhile to implement relative to the added complexity involved.

You already spent half an hour on this spot in your video, and further discussion in this thread including another poker player with a nodelocking approach. That is a lot of resources already, and you didn't cover the biggest decision in the hand: you were put to the test for all your money and you didn't knew what to do.

I don't see what would be problematic about running a good simulation and looking at river ranges, even if it was just quickly.

If I'm able to avoid 0 reach spots with higher accuracy (when I'm not consciously exploiting), then I'll avoid the problem you raise while bringing my game nearer to GTO and improving my performance vs tough opponents.

So you try to avoid this exact situation by way of playing closer to what PIO tells you to do on flop and turn - be more aggressive - and to simplify the river - always check even if having a small leading range is optimal (according to PIO). That works to some extent, but I see two problems with this approach:

  1. You learn a little bit, but not very much. On the river you didn't have a strategy, you decided ad hoc to bet without a plan how to react versus a shove. Because of such ad hoc decision making you will be less likely to figure out in the future that you've entered a similar situation where checking should be your high frequency play. Also you didn't figure out on what type of board structures this occurs, does it still occur on 667dd, 338cc, TTJhh?
  2. You think you can avoid this situation in the future "reach 0". This is a fallacy. You can be in the CO, play the hand in the exact same way, get to the river and have your opponent lead into you. And you will be just as clueless as you were before studying this hand for half an hour in PIO.

I believe this highlights some of the differences between a more academic approach versus a more practical or professional approach.

An academic issue with your approach is that by game theory definitions you don't even have a strategy if you don't know what to do outside your PIO abstraction.

A practical approach would be to have strong fundamentals on how to play this spot based on range vs range equity, blockers, stack to pot ratio etc. I got the impression that those fundamentals were lacking during live play.

You could study the PIO abstraction of this spot in detail and be able to get your exploitability below 5bb/100 within this abstraction. But you could be 500bb/100 exploitable if you are clueless outside of this abstraction, especially when leading is optimal or near optimal. In such a case leading rivers has to come up very infrequently to make it not worth your time studying this situation.

Dec. 5, 2017 | 7:58 p.m.

Do you regret teaching Cumicon PLO for "too little"?

Dec. 4, 2017 | 11:12 p.m.

The .5% leading frequency shown in the sim is likely noise because I ran a flop sim and some of the river nodes will remain poorly converged.

0.50% leading frequency doesn't have to be noise. If you suspect that your simulation is not accurate enough to decide if you should certainly never lead or lead with a low frequency then you should run a better simulation. You can still observe their range for shoving over your river lead and see if it is balanced and how well Qs6s performs against that range.

I don't think it's a particularly useful exercise for me to node lock flop/turn and then see how my river range changes because in general I'm attempting to play a basic strategy that's near GTO

(Even) when the previous streets should be played more as checkraises and the river should never be lead in your shoes, playing a 'basic strategy that's near GTO' is not a well defined objective. Before you ran this hand through PIO, what was your near GTO estimation of what your river leading range and your river leadcalling range should look like?

A reasonable goal that translates to attempting to play near GTO and that also covers 'mistakes' is that you try to minimize your own exploitability given that your opponent knows your complete strategy. And when you assume that leading the river is never played in GTO by you and that your opponent plays a GTO strategy, that means that your opponent has freedom of responding to incorrect river leads. With this freedom he could for instance choose to play a little more on the loose passive (more calling and less raising) side or to be more on the tight aggresive (less calling and more raising) side. Because you deviate from GTO, his response does not have to be unique.

It could be that
EV(checkcall) > EV(betfold) > EV(betcall) for for the tight aggresive GTO strategy, and
EV(checkcall) > EV(betcall) > EV(betfold) for the loose passive GTO strategy.

But when EV(betcall) > EV(betfold) for all GTO strategies then folding Qs6s after betting is a fundamental mistake. By definition, such fundamental mistakes make you more exploitable when there is a positive chance that you lead rivers (as was observed during this hand).

I'm not appreciative of the part of your comment that uncritically assumed what "proper" analysis should be and then criticized me for not having the same goals as you happen to have.

I hope this comments helps you towards understanding "proper" analysis.

One last thing, players like wilhasha and bajskorven exploited their opponents by way of getting them into (uncommon) situations that they did not understand as good as them. A player of your caliber should be doing that to 500 NL zoom regs, and not the other way around.

Dec. 4, 2017 | 8:08 p.m.

The Q6 analysis @ 40:10 you say "we do not lead", which is incorrect. There are combos that lead, and your refusal to look at how they respond to a shove makes it impossible to see if your river fold was correct.

Your PIO simulation leads you to conclude that you would raise less on previous streets with draws, value and marginal/bluffing hands than you actually did. So that makes your river range different and most likely narrower and stronger than what PIO would arrive with. And still, given all your incorrect assumptions and fundamentals, you don't know what your opponents river raising range would look like or how you should respond to it.

Your guess "maybe this is just a really easy call too" @ 30:30 remains a guess.
Calling with a straight combo like A5 @ 30:58 while folding a Q6 flush is debatable, especially when you don't back this up. Having A5 would just block 1 out of 3 combos of A3s and none of the other boat combos, while a high flush blocks a lot more value combos. Obviously, capping your leading range at K/Q high flushes and folding all Q high flushes makes the Ks an extremely profitable blocker for your opponent.

After watching this video I'm left with more questions than answers. I can't like a video when half the video is analysis of just one hand that is not done properly.

Dec. 3, 2017 | 1:53 a.m.

Congratulations everyone!

Dec. 2, 2017 | 2:55 p.m.

Impressive, very nice

Nov. 30, 2017 | 1:01 a.m.

5-bet size is too small if you believe his 4-bet range is soo wide that you want to reraise AKo for value as a high frequency play.

Nov. 8, 2017 | 4:37 p.m.

The river completes KQs, which is terrible for me.

Nov. 3, 2017 | 1:59 a.m.

I remember when I was in my youth, I used to make the Monthly top 5 like all the time...

Nov. 2, 2017 | 3:17 a.m.

He needs to be called around 40% of the time. If you raise here with 12% of hands and do all the calling by yourself that means ATs+,AK,99+.
And more like AQs,AK,JJ+ if the other callers are capable of calling with some TT-JJ,AK type hands that they might flat here.

Nov. 1, 2017 | 2:37 a.m.


Oct. 28, 2017 | 5:18 p.m.

What a spewtard!

Seat 1: takechip ($27407.65 in chips)
Seat 4: Katya18 ($37045.02 in chips)
Seat 6: OhHeyCindy ($28764.39 in chips)
OhHeyCindy: posts small blind $50
takechip: posts big blind $100
takechip: posts the ante $20
18: posts the ante $20
OhHeyCindy: posts the ante $20
Katya_18: raises $175 to $275
OhHeyCindy: raises $925 to $1200
takechip: folds
Katya_18: calls $925
FLOP [9h 9s 5s]
OhHeyCindy: checks
Katya_18: bets $843.81
OhHeyCindy: raises $2292.09 to $3135.90
Katya_18: raises $2292.09 to $5427.99
OhHeyCindy: calls $2292.09
TURN [9h 9s 5s] [6c]
OhHeyCindy: checks
Katya_18: checks
RIVER [9h 9s 5s 6c] [7c]
OhHeyCindy: bets $4426.28
Katya_18: calls $4426.28
OhHeyCindy: shows [Qd Jc] (a pair of Nines)
Katya18: shows [As 4s] (a pair of Nines - Ace kicker)
18 collected $22265.54 from pot

Oct. 17, 2017 | 9:56 p.m.

Oct. 8, 2017 | 10:02 a.m.

Comment | GameTheory commented on A Poker Site Should

He is saying even the beta won't come out in a couple of months. Can't give hard estimates. Phil doesn't want to come out with disappointing software, it has to be polished till they have it how they want it.

Looks like this thread will continue in the way it has for months to come.

Sept. 29, 2017 | 10:51 p.m.

Comment | GameTheory commented on A Poker Site Should

The update everyone has been waiting for, the domain is up:

Sept. 28, 2017 | 3:03 a.m.

Sept. 14, 2017 | 5:34 p.m.

Sept. 2, 2017 | 7:18 a.m.

Comment | GameTheory commented on Solve this game

Yeah sure. Isn't it harder to do a full proof that you've solved a game with just CREV?

Aug. 27, 2017 | 10:31 p.m.

So you base these plays on experience. I was wondering if you had done simulations or database work, but that seems to be not the case. Thanks for answering.

Aug. 27, 2017 | 10:11 p.m.

@GT: the "one measure" is that KQJ9 makes better wraps and more and nuttier straights.

I'm not sure if "nutty wraps and straights" are relevant enough to qualify as a relevant measure since it intentionally caps the handstrength at straights. It would make more sense to me to talk about handstrengths that are both at least a straight and also the nuts. That would make them truly "nutty".

For starters, AJ98 makes the nuts by the river more often than KQJ9, 8.54% vs 7.69%:

On the flop AJ98 has the nuts that is at least a straight more often than KQJ9, 3.23% vs 2.89%;
On the turn AJ98 has the nuts that is at least a straight more often than KQJ9, 7.26% vs 6.52%;
On the river finally AJ98 has the nuts that is at least a straight more often than KQJ9, 8.64% vs 7.70%:

Lastly AJ98 has more equity against 2222 than KQJ9 because KQJ9 blocks its own outs:

Aug. 27, 2017 | 8:36 p.m.

Comment | GameTheory commented on Solve this game

The leading for less than 1% of the pot should be done with not just bluffcatcher I presume.
It is also unlikely that an exact size of 125% is still optimal after OOP starts leading.
Can you show your math?

Aug. 27, 2017 | 6:31 a.m.

Comment | GameTheory commented on Solve this game

Would be interesting to see if OOP can improve by way of leading instead of checking. And if not, why not.

Aug. 27, 2017 | 2:01 a.m.

Comment | GameTheory commented on Solve this game

Then you should've the solution by now. What is it?

Aug. 26, 2017 | 8:37 p.m.

Comment | GameTheory commented on Solve this game

Nobody did this? Shouldn't be that hard....

Aug. 26, 2017 | 8:16 p.m.

Comment | GameTheory commented on Boxing vs. MMA

What a smokeshow!

Aug. 26, 2017 | 11:33 a.m.

Load more