Optimal sizing on wet flops: texture based betting - Page 6

If its letting you tinker with different bet sizing tree's, its not a GTO solution. The GTO solution only has one bet sizing tree.

I think you are correct in saying that. Hypothetically, there can be more than one decision tree that is GTO, but there doesn't necessarily have to be (and I'm pretty sure in No-Limit holdem we don't know enough to be certain). We certainly can't just arbitrarily pick a bet size and construct a 'GTO' solution using it, that is complete nonsense.

I think the sizing of bets is one of the things that make NL-holdem so complicated, and despite its name, this GTO site isn't actually giving you something truly GTO. I think what it is doing is taking inputs you give it (such as bet-sizing and handranges), constructing a mixed strategy that is unexploitable and balanced with the assumption that all preset inputs are fixed, then calculating the expected value of the scenario. But coming up with an unexploitable line is much easier than coming up with THE unexploitable line that is GTO, which this program doesn't come close to doing. Thankfully we are still a long time away from that for holdem games with any decent stack depth.

On February 04 2015 18:35 Minsk wrote: I dont understand how GTORangeBuilder claims to have a GTO solution to turn and river play if you can change bet sizes within it. Wouldn't a true GTO solution only have one betsize that cant be altered?

you dont change betsizes or anything in the gtorb answer; you just set up the variables before the turn and it spits out the turn and river tree for you to study. two of those variables are preflop and flop betsizing.

Also there may be multiple cooptimal solutions to HUNLHE we aren't sure yet

No, theres only one optimal solution, for sure. It's when you can make no more adjustments, so its a singularity by definition. It's like the concept of the enlightned one, there is only one enlightened one.

It's like there can be multiple blackholes, but they are never even and one always eats the other one. If you look at GTO, it would be like the biggest blackhole that exists, GTO is not all blackholes.

I don't believe this can be it since the process of reaching GTO is made by adjustments, until you cannot adjust anymore. It's like you cannot know GTO until you know GTO -1 move, and nobody knows that.


So the question is what is this software doing? It seems to pick a range, then find a Nash Equilibrium that breaks even with that range?

On February 04 2015 23:27 Minsk wrote: No, theres only one optimal solution, for sure.

That's 100% bs. There are tons of examples of games with multiple nash equilibria, and I don't think we know enough about hold'em to say for sure there is only one solution.

We know for a fact there is more than one solution.

Example:
Board is 3333A in holdem. Both players automatically have the nuts. As long as neither one of them folds, either player could bet 5, 10, 20, or even zero bb against the other (checking the nuts), and they won't be able to improve their expectation against the other.

On February 07 2015 06:07 NMcNasty wrote: We know for a fact there is more than one solution. Example: Board is 3333A in holdem. Both players automatically have the nuts. As long as neither one of them folds, either player could bet 5, 10, 20, or even zero bb against the other (checking the nuts), and they won't be able to improve their expectation against the other.

I think when most people here are talking about hold'em solutions we're only considering non-trivial information sets.

On February 07 2015 06:07 NMcNasty wrote: We know for a fact there is more than one solution. Example: Board is 3333A in holdem. Both players automatically have the nuts. As long as neither one of them folds, either player could bet 5, 10, 20, or even zero bb against the other (checking the nuts), and they won't be able to improve their expectation against the other.

without rake that is true, with rake, the only play is check


edit, except if rake is already maxed at higher stakes

On January 28 2015 22:15 Smuft wrote: Show nested quote + On January 28 2015 17:23 dogmeat wrote:   On January 28 2015 12:02 Romm3l wrote:   On January 27 2015 18:43 dogmeat wrote: one does not bet small on the flop im 3b pot b/c he worries about effectiveness of c/r's but b/c theory suggests betting equal fractions of the pot on all streets is the best strategy w/ polarized range that's an interesting result as well and i'd love to hear where you got it from. however your argument is circular and empty: you are saying 'x is suboptimal not for a reason y, but because z is optimal and x is not z' which simplifies to 'x is suboptimal because it is not optimal' - no shit. remember optimal strategy is defined as the strategy that minimises the expectation of a perfectly adapting opponent. try to think why minbet, and why 4x pot shove are each suboptimal betsizes in our spot (4x spr, ip on flop, villain checks). why does overbet shoving, or betting min make life easier for our opponent and fail to minimise his expectation? once you arrive at better answers to these questions than 'it is suboptimal because it is suboptimal', think again about why fullpot might not be optimal and consider again my argument that the crai option becomes too good (he gets good odds on it, and it forces you to bluff cb/fold less often and miss profitable spots, for example). mathematics of poker by bill chen anyway optimal strategy maximases your ev and your argument is completely invalid, you can have balanced PSB range in your example, but its not gto b/c gto play maximases your ev, which is achieved by betting equal fractions of the pot over 3 streets (which allows highest bluff:value ratio), c/r is not an issue at all I think what Romm3l is trying to say is your argument is "this play is the best because it is" instead of giving some solid reasoning. I have to agree with him. Your posts are often just random statements which is too bad because you probably have some decent reasoning / thoughts for how you came up with your ideas and could benefit yourself and others if you voiced them. This wouldn't be so bad if your random statement was at least a correct random statement but it's not. I will try my best to explain below.   On January 27 2015 18:43 dogmeat wrote: one does not bet small on the flop im 3b pot b/c he worries about effectiveness of c/r's but b/c theory suggests betting equal fractions of the pot on all streets is the best strategy w/ polarized range The theory you're referring to from "Mathematics of Poker" is illustrated in toy games that are somewhat similar to NLH but in the end are not the same at all. The board canges in NLH but it does not change in the toy games used in MOP where geometric betting (betting equal fraction of the pot on all streets) is proven to be best for the polarized range. The author suggests that geometric bet sizing should be used in some situations in NLH but he does not prove it. It's just his 2006 opinion. It was a very advanced opinion for it's time but poker has been much more deeply studied and analyzed now. I have my own ideas for why I think this opinion is out-dated and mostly wrong but it's not organized and way beyond the scope of a simple forum post so I'll give you a much cheaper but probably more credible argument: Look at the best NLH players in the highest stakes games today, the standard cbet size on most flop textures in 3b pots is 1/3 pot. This is 9 years after MOP was written and most of these guys have read it, expanded on it, read more about game theory and done intense quantitative analysis on the game. Many of them will play stakes up to 200/400 against pretty much anyone in the world. Why aren't they betting equal fractions of the pot in 3b pots?

Your statements about bet sizing in 3bet pots used by the very best HS players online is incorrect. There is only 1 person who has done this at the very highest stakes: Sauce. And his is not 1/3, it varies from 35-74%; closer to 100bb is 35-40% but can still be as high as 71% near 100bb

Thanks for the rest of the discussion.

edit I am talking about HU

edit Looking at 3+ people, my statement stands except for Kanu

edit Kanu is the only HS person who consistently uses 1/3 in HU/3+ with the least consideration for board texture.

On February 07 2015 16:25 Highcard wrote: Show nested quote + On January 28 2015 22:15 Smuft wrote:   On January 28 2015 17:23 dogmeat wrote:   On January 28 2015 12:02 Romm3l wrote:   On January 27 2015 18:43 dogmeat wrote: one does not bet small on the flop im 3b pot b/c he worries about effectiveness of c/r's but b/c theory suggests betting equal fractions of the pot on all streets is the best strategy w/ polarized range that's an interesting result as well and i'd love to hear where you got it from. however your argument is circular and empty: you are saying 'x is suboptimal not for a reason y, but because z is optimal and x is not z' which simplifies to 'x is suboptimal because it is not optimal' - no shit. remember optimal strategy is defined as the strategy that minimises the expectation of a perfectly adapting opponent. try to think why minbet, and why 4x pot shove are each suboptimal betsizes in our spot (4x spr, ip on flop, villain checks). why does overbet shoving, or betting min make life easier for our opponent and fail to minimise his expectation? once you arrive at better answers to these questions than 'it is suboptimal because it is suboptimal', think again about why fullpot might not be optimal and consider again my argument that the crai option becomes too good (he gets good odds on it, and it forces you to bluff cb/fold less often and miss profitable spots, for example). mathematics of poker by bill chen anyway optimal strategy maximases your ev and your argument is completely invalid, you can have balanced PSB range in your example, but its not gto b/c gto play maximases your ev, which is achieved by betting equal fractions of the pot over 3 streets (which allows highest bluff:value ratio), c/r is not an issue at all I think what Romm3l is trying to say is your argument is "this play is the best because it is" instead of giving some solid reasoning. I have to agree with him. Your posts are often just random statements which is too bad because you probably have some decent reasoning / thoughts for how you came up with your ideas and could benefit yourself and others if you voiced them. This wouldn't be so bad if your random statement was at least a correct random statement but it's not. I will try my best to explain below.   On January 27 2015 18:43 dogmeat wrote: one does not bet small on the flop im 3b pot b/c he worries about effectiveness of c/r's but b/c theory suggests betting equal fractions of the pot on all streets is the best strategy w/ polarized range The theory you're referring to from "Mathematics of Poker" is illustrated in toy games that are somewhat similar to NLH but in the end are not the same at all. The board canges in NLH but it does not change in the toy games used in MOP where geometric betting (betting equal fraction of the pot on all streets) is proven to be best for the polarized range. The author suggests that geometric bet sizing should be used in some situations in NLH but he does not prove it. It's just his 2006 opinion. It was a very advanced opinion for it's time but poker has been much more deeply studied and analyzed now. I have my own ideas for why I think this opinion is out-dated and mostly wrong but it's not organized and way beyond the scope of a simple forum post so I'll give you a much cheaper but probably more credible argument: Look at the best NLH players in the highest stakes games today, the standard cbet size on most flop textures in 3b pots is 1/3 pot. This is 9 years after MOP was written and most of these guys have read it, expanded on it, read more about game theory and done intense quantitative analysis on the game. Many of them will play stakes up to 200/400 against pretty much anyone in the world. Why aren't they betting equal fractions of the pot in 3b pots? Your statements about bet sizing in 3bet pots used by the very best HS players online is incorrect. There is only 1 person who has done this at the very highest stakes: Sauce. And his is not 1/3, it varies from 35-74%; closer to 100bb is 35-40% but can still be as high as 71% near 100bb Thanks for the rest of the discussion. edit I am talking about HU edit Looking at 3+ people, my statement stands except for Kanu edit Kanu is the only HS person who consistently uses 1/3 in HU/3+ with the least consideration for board texture.

Okay thanks for correcting me - looks like you have a healthy amount of data on HSNL to draw from and I do not so I'll take your word for it.

I monitored the HSNL threads on 2+2 late last year and it seemed like every 3b pot was being bet at 1/3 for a couple of months; guess it was just a phase the metagame went through because obviously that's not the case anymore.

It still seems a lot of 3b pot cbets are <50% pot though, can you comment anymore on 3b pot cbet sizes of either specific players or general HSNL population?

I did not look at the other HS people like forhaley, baron, etc but I assume ike/sauce/wcg/jungle have better data than those guys when 3betting out of position.

edit quick look at others with same results

I don't have much info on the games at the very end of 2014 for 6max if some people decided to randomly start 1/3

Most players it varies by board texture and stack sizes, but the stack size correlation is weaker than board textures. It takes 200bb+ to see more consistent sizings, similar to none 3b pots. I won't type out a bunch of sizing here for players, generally, it averages around 50%, 45-55%, but can be 63%, 66%, 68%, 71%, 74% and some textures 29%-32%

The 6max HS pros seems to stick closer to a single bet size and closer to 50%, sometimes ike sticks with 66% other times it varies widely. Kanu was stuck on 33%

looking at King10, his avg is 50% (feb 2015)

There is only one solution to the 3333A example. It's to check, both because betting could cause rake, and because betting wastes extra energy. The simplest and most direct and most optimal solution, that cannot be countered is to check.

Considering betting is wasting energy, as you have to decide how much to bet then.

Rake messes everything up.

One weird spot is when the board is A3333 or something similar in a $1 pot on the river hu, but in a 6max game at 50NL (so rake is capped at $3). In that case, calling a shove ($50) from your opponent is -EV since what you're paying in rake is greater than your share of the pot. So if the first player in shoves, and the second player folds, we are actually are at a Nash equilibrium since neither player can singularly improve their expectation.

On February 07 2015 21:13 Minsk wrote: There is only one solution to the 3333A example. It's to check, both because betting could cause rake, and because betting wastes extra energy. The simplest and most direct and most optimal solution, that cannot be countered is to check. Considering betting is wasting energy, as you have to decide how much to bet then.

The concept of 'wasted energy' isn't really something that we worry about when talking about game theory. Also, I would say the opposite of this is true, assuming further betting doesn't increase rake (either it's capped or players are just playing at home with no rake or something). If you go all in, the exact same outcome happens if your opponent plays correctly, but by betting you give opponent the chance to make a mistake. So jamming dominates checking (because of the one in 10^16 times your opponent accidentally folds)

GTO strategy doesn't think like that.

It's opposite, its like fuckery, GTO wants the least ammount of fuckery, because its only concerned with beating a perfect opponent, it never thinks the 10^16 times they fold.

If we compare it to a computer program, if we don't check now we have to write code for the ammount to bet, which is more code, and our opponent now has to write code for what to do against a bet. Both of these are completely null in EV terms, but its slowing both of us down from theoretical perfection. This slowdown or mistake from our two programs does not matter against each other, but it brings both of our programs rating down against everyone else. It's fuckery, were fucking around with each other forcing each other to write that extra little peice, does nothing against each other loses a little bit of edge to everyone else.

On February 09 2015 02:06 Minsk wrote: GTO strategy doesn't think like that. It's opposite, its like fuckery, GTO wants the least ammount of fuckery, because its only concerned with beating a perfect opponent, it never thinks the 10^16 times they fold.

That is entirely incorrect

GTO is the perfect solution of ranges/bet sizes/pot sizes. If player A plays GTO and player B counter by playing anything except GTO, player A crushes player B. The further from GTO player B goes, the more Player A wins.

If Player A knows the GTO play and knows player B will not counter with the correct GTO play, then Player A can make a Maximally exploitative play based on what Player B is actually doing. The maximally exploitative play is more profitable than GTO only because player B is so far from GTO. But any mistake by Player A deviating from GTO incorrectly can be exceptionally negative EV.