Single betsizing vs Multiple Betsizing
Posted by Mates.
Posted by
Mates.
posted in
Mid Stakes
Single betsizing vs Multiple Betsizing
"There's a misconception when it comes to bet sizing. As it turns out, when playing GTO, all sizes will show almost exactly the same performance. You can check this for yourself, by building the same tree, but for different sizes. After that, check the overall performance for OOP. As you'll find, all EVs will be nearly the same, and any bet size should basically work.
What is really important when it comes to your performance is your quality of play; the bet sizing has almost no influence whatsoever; for any size there's a strategy that performs almost exactly the same as any other size. A consequence of this is that the approach of using multiple sizes (to see which one occurs most) doesn't really work. Within a given dEV there's several different configurations of betting frequencies that all lead to the same result. You're basically looking at static. Bet sizing is really not something to be overly obsessed with; there's almost nothing to be gained here as opposed to focussing on your quality of play."
This has been spinning me for quite some time ..
I don't have much experience with solvers, but trying to improve I try to find ways to simplify strategies.
How true is this?
if it were true, what only bet would OTT and OTR choose in the following sequences? (assuming we can also push all OTR)
At 100bb
Pot 5bb
* SRP IP
* SRP OOP
At 100bb
Pot 24bb
* 3bet OOP
At 100bb
Pot16.5-19bb
* 3bet IP
Loading 11 Comments...
This can't be right. If you give a solver more options in betsizes your EV will increase. It isn't doable in practice but given he is talking about GTO then more betsizing options generally is better.
I did some work previously with simplification strategies where I compared 2 betsize strategy against 1 size strategy. The 1 size strategy always loses a small amount of EV. It's not that big but it's there.
I also saw a RIO video on simplification where the EV loss of a simplification wasn't big. Think it was Saulo Riberio's but can't find the one. The EV loss is still there.
Betsizing is already relevant from preflop and then flop onward. A very small increase preflop will make the pot much bigger on the river and set up cleaner betsizing on the way.
Imagine a 2.25x open or a 2.5x open. Doesn't matter right?
2.25x: 5.5 preflop vs BB. Pot pot pot bets gets us 5.5 flop, 16.5 turn, 49.5 river.
2.5x: 6 preflop vs BB. gets us 6, 18, 54.
So a quarter of a blind difference preflop gets us a 4.5 difference by the river with the exact same line.
Citanul ... Thanks for answering! :)
When I meant having only one bet size, I meant TURN and RIVER.
And to some extent what he says is correct, it is practically impossible to execute a "perfect" strategy if we overcomplicate it. As for the EV, it would be more expensive to have structure errors than bets size errors (having several various bet sizes wins VERY little EV vs a GTO player)
This is taking the exploitative game out of context.
What he proposes is
create subsets with 1 single bet on the flop (25,50,75, range bet, checkin range, etc),
We would select the one that retains the most EV.
From there, we started creating the turn and river strategy.
This flop single size is a simplification a lot of people are doing. It's very common these days. So it's good advice.
Not as simple of just saying SRP IP will play this line with this size. The actual cards coming will be the decider of the betsize and how often we are betting. So it's not possible to generate an all round strategy.
What most people are doing is small cbet with range in flops. Turn onward big bets with a checking range. In general this should work.
Not much to add.
In general, the difference between a "GTO" solution with multiple betsizes (close to as many as realistically possible) and a one-size solution is +/- 1 bb on flop and +/- 0.5 bb on river. That is for a 3-bet-scenario with 18bb in the pot and 90bb rem stacks.
I'd expect a SRP scenario to fall in comparable bandwiths. That makes a difference of up to 100bb/100. Not "nothing", right? :)
Obviously it's not possible to mimic the solver solution for us humans, that is clear. That means, we have to simplify. On the other hand - our "non-GTO-aware" opponents do simplify as well AND additionally they make more mistakes (hopefully), which outperforms our EV-loss due to not sticking to the exact solver solution. Which makes +/ 100 bb/100 realistic again.
Additionally it gives us room for exploits. And those exploits - again - heavily depend on betsizing, at least in certain situations. It would be criminal just not to care about that and take the short route.
Poker is not easy. There's no shortcut.
BigFiszh
PS: Who gave that "advice"?!
This information is from the creator or administrator of GTO + (Solver) / Flopzilla / CEV +.
So it makes me rethink.
These are the comments I found:
Quote:
Originally Posted by p0ker_n00b View Post
So if bet sizing does not matter much in a GTO solution, why would PIO pick one size preferably over another?"
You're basically looking at static. If you slightly change ranges, or the tree, or even just solve to a different Nash distance, then different frequencies may be reached. Even if you were to pick the line that the GTO solution least seems to prefer, then the overall EV would at worst be only slightly below the overall EV for the "preferred" line. There's very little to gain here.
This entire approach doesn't work. All bet sizes perform nearly identical to one another. Typically the difference in performance between two bet sizes is within 1% of each other. Quality of play is far more important. Versus an optimal opponent there's not much to be gained from bet sizing, but far more from playing well within the bet sizes that you've chosen.
"When it comes to turn play, I would recommend just using single bet sizes. Turn and river play is only needed to estimate the EVs of hands in the followup lines. It's not necessary to use overly complex play here. Your trees will become smaller, and solve faster.
instead I would really recommend just using single bet sizes on later streets (turn/river). Using multiple bet sizes will not result in any significant better quality of the solution, given that they perform only slightly better than single bet sizes. Also, even with single bet sizes it can already be very challenging to make sense of the solutions from a human perspective. When just sticking to single bet sizes on the turn/river your trees will be smaller and solve faster, while being easier to interpret.
The above applies to GTO strategies versus perfect players. However, in practical play, people (particularly poorer players) may react sub-optimally to certain bet sizings. So in practice it makes perfect sense to adapt your bet sizing to your opponent's leaks. However, as the quality of opponent increases, the added value of bet sizing will likely decrease; it's the quality of play what is most important. If having to choose between high quality of play and low quality of bet sizing, or vice versa, then the former will outperform the latter by a wide margin."
What Scylla is saying here is that the EV difference for different betsizes isn't going to be big. So when you want to solve and look at flop ideas then your turn and river doesn't need to be overly complex. The EV won't differ much but the solve will be much faster.
And this is a great idea if you are solving to look at the bigger picture from solves, such as the EV. But let's say we remove complexity from turn and river. Only small sizing, no big sizing.
You will notice that the flop play will change a lot from the OOP player as IP is not putting a lot of pressure on their weaker hands on future streets. So GTO+ is calling more than it would when turn and river has big bet options.
This won't affect the EV a lot. But if you then model your own play to those flop defense frequencies you'll be arriving on the turn too wide.
Strange he'd say this, there are definitely bet sizes that are vital (especially on turns and rivers) to providing EV to a strategy. Imagine if we couldn't go allin on river and could only bet 25% pot?
I understand, what he means, but his explanation is indeed somewhat strange. What is "quality of play"? What does a "strategy" actually consist of? We cannot do anything else than choosing betsizes, right? Even checking is a "betsize", namely 0. That means, there is nothing left than choosing between betsizing.
And what Scylla now explains is to reduce betsizes to two (0, X), which seems pretty arbitrary. Why not one? Or three? It might be that the additional value of betsizes decreases with any further added size, i.e. the EV-gain of 20 betsizes vs. 19 is likely not measureable anymore, but it's there - and reducing from three or four to only two seems a bit over the top.
Finally, restricting ourselves to only one betsize (besides checking) not only affects our (potential) EV, it makes us pretty easy to play against, which a) reduces mistakes of our opponents and b) makes us vulnerable against counterexploitation.
https://forumserver.twoplustwo.com/167/commercial-software/cardrunnersev-155966/index403.html?
Here is the link.-
on this page and on the previous one I ask again and it answers the same thing ... if you have a user you can search on the thread "multiple bet sizing" and you will find that this always responds.
I respect Scylla, but I disagree here.
What if I chose "all-in" as my preferred betsize - because it's the most easy sizing? Or "check" 100%? If one says "meh, that's nonsense" - who tells? What is the range in which the actual betsize is irrelevant?
I'd love to see what he answers that question <3
Be the first to add a comment