I'm an experienced poker player and yes I am fully aware that "betting to see where you're at" is no good... But sometimes I question that truth. Maybe there actually is sound theory behind doing just that, betting for information?

As I've progressed as a poker player, I have been tending to try to size my flop bets so that I will maximally split my villains' ranges (between fold, call, and raise), and therefore provide the most information about their hand for later streets, especially in case of a call. Or sometimes, I think the villain is very rarely folding on a flop, but will give off more information if I check and allow them to bet (especially if there's a lot of variation in their bet-sizing).

Obviously there's quite a lot more to bet-sizings and poker decisions than using them *just* to split up villains' ranges... But I'm wondering if there's any proof that in multi-street games, especially NLHE, that a significant part of our EV is earned on earlier streets by acquiring information about villain's hand to use on later streets -- as opposed to simply stating that bets should always be purely for value or purely as a bluff (or some value-bluff merge combination).

The problem with "betting for information" is actually quantifying in EV what that information gives you and comparing it to the EV of other options with "less information".

The only information that is reliable is either information that can be tracked directly by a software program over a large sample size or information you can obtain by seeing a showdown.

Even then showdown information is typically biased and often doesn't reveal more than you can already assume based on the known cards in the deck and some assumptions/generalizations about the players in your game.

I am curious what maximally split villain's range means. Could you elaborate?


I think a lot of players should be and are moving away from the idea of "pure value" or "pure bluff" or even using terms like value-bluff merge and should/are considering what betting sequence is most profitable with my hand in this situation?

Basically, I mean I've tended to size down bets on flops, both heads up and multiway, both IP and OOP, but mostly OOP and multiway... There's many ways to look at the motivation for doing this, but one thing that I think about is how can I size my bet so that they're as equally likely to fold, call, and raise as possible. If they take each of these actions exactly 33% of the time, then they are providing the maximum amount of information about their range. If their raise sizes are variable, they are tending to provide even more information about their range... Now obviously when they fold, their range on later streets doesn't matter, and most players usually don't raise without at least an overpair - but still, this is one framework I've used to think about my bet-sizings... As I progress to the turn, I think less in terms of information and splitting their range and more in terms of betting to bluff or betting for value, and then the river is 100% about bluffs vs value... (somewhat as an aside, in a live setting I also like to raise flops and turns often, or donk turn after PFR cbet just to pick up some sort of live tell. Because I raise very often, I get called a lot, and I know I'm betting/raising with hands in spots I theoretically shouldn't be, but I figure I can recoup whatever losses by using my reads to turn hands into bluffs, make very thin value bets on the river, or just check when I sense I'm beat and they're not folding).

I recognize that no bet or raise should purely be "for information", but it would be interesting to figure out if there is some quantifiable value for the information we do acquire throughout a hand.

Could you elaborate more why you believe this to be true?

Specifically why you believe your bet sizing can force an opponent to split her range in a certain fashion other than solely on perceived equity and why splitting her range equally among 3 actions maximizes information about an her range?

This. I'd recommend OP reads Matt Janda's books and the threads about them on 2+2, because the reasons for betting are too complicated for me to answer his questions succinctly. A bet is rarely made purely for value or as a bluff. Arguably more important is the protection or denial of equity. That said, I pointed out in one of the Janda threads that some of the smaller sizes used today (as recommended by solvers) look like the stereotypical bets and raises for information that are beloved of weak/live players. Things like donking quarter pot on a monotone flop, or check-minraising top pair on a wet flop (moves that used to be considered "fishy" plays for "finding out where I'm at") are apparently part of optimal play. So-called "blocking bets" (where you bet small for value, but fold to a raise) also seem to fit into this category. It's just that with optimal play, some of your small bets are "fake" blocking bets that are actually trying to induce a raise or a light call. There's no tangible/quantifiable value of the information, though. Optimal play just uses the sizes that generate the highest profit for the strategy as a whole in the long run.

Interesting... I'll definitely have to read Matt Janda's books and the threads you mentioned. I'm guessing that some of those bets/strategy would also be classified as "fancy play syndrome."

However, I'm still not so sure there's no quantifiable value of information. But this is something that I'll have to think more about.

It could be argued that smaller bets allow for more room to maneuver, creating a more complex decision tree for both you and your opponent, which may be beneficial if you are better at navigating a web of possibilities. Of course, it may be possible that this fancy play is +EV but a less fancy style is more +EV.

Trying to split someone's range so they are equally likely to fold, call, and raise seems silly because it's hard to come up with a scenario where they have that many raising hands. Take a reasonable hand range for calling your preflop range, pick a sample flop like A87 with two spade, and figure out where they change from folding to calling and from calling to raising within that range if all three actions are equally likely. Symmetry can be pretty sometimes, but how do you get more information from splitting their range into thirds than if you get them to raise the top 10% of their range, call with the next 50%, and fold the bottom 50%?

Yeah, there aren't many scenarios in which someone is more likely to raise than call or fold. That being said, you often read people in LLSNL say something like, "you have to bet more than that or you're going to be bluff-raised here a lot." And my attitude is I like to bring that part of their range along - it gives me an opportunity to make decisions for more money, and therefore hopefully a higher EV. And if they are induced to raise, it also gives me more information when they just call as I can become more confident they're not slow-playing a big hand. This seems to allow my range to be much more polarized than their's on the following street (I.e., I can bomb turn with a lot of value hands and a lot of bluffs). Of course there are exceptions to everything, and sometimes the last thing you want to do is induce your opponents to bluff as it is equivalent to getting them to be more balanced in spots they'd otherwise remain unbalanced.

I'd tell people who bet too small that they should consider betting more so that they don't level themselves into calling a raise with a bluff-catcher. Great players can eke out extra EV my making correct decisions in tough spots. Most players aren't great and would benefit from playing a style that simplifies some decisions, even if it is exploitable, because they won't run into too many players who exploit it well. So, my general advice is to pick a bet size where you don't know what to do if raised. If you can't figure out a bet size for that, then maybe you should check.

Wow. Awesome thread! I hate to be that guy, but aren't all bets a trade off between trying to win the hand versus revealing the info that you can win the hand?

In my opinion this is the most recent thread that deals with information directly and turns it into something close to value (in this case only equity). After thinking about this thread for awhile I figured there really isn't a direct way to map information to money, but obviously I am by no means in expert in that area and would invite others to discuss!

https://forumserver.twoplustwo.com/1...tions-1691093/

Imagine you were running a bot with access to the complete GTO solution to poker in a database the size of the moon. You already have all the pre-calculated information you'll ever need, because you've got the entire decision tree mapped out. If you bet 33% of pot on the flop, there are certainly action sequences that lead to higher or lower EVs for particular hands (e.g. you won't have many combos that want to face a 3x pot raise), but the response of your opponent (the "information" he gives you by calling, folding or raising) doesn't really have any direct value, as such. If your bet size was optimal in the first place, it kind of doesn't matter how villain reacts, because you have an optimal response to whatever he does.
Some parts of your range will have easy calls facing a raise, some parts are folding, and some are fist-pump jamming or whatever. To re-iterate, it doesn't matter what information villain gives you, because all of his possible actions were assigned a probability before you made the first bet. In short, you already know what you'll do if villain raises or calls, because you considered those possibilities when you made the initial bet.

I understand what you're saying, but consider this scenario. We are playing a GTO bot, except for at each street we choose two actions: (1) our real action, and (2) a hypothetical action. The hand proceeds with our real action, but the bot must also tell us what it would have done if we took our hypothetical action (in the case that the bot is playing a mixed strategy, it must choose a response at the correct probability).

In this scenario, I think it's obvious we gain a significant advantage. However, the EV we gained was (a) solely because we have additional information about villain's range and (b) the information provided was based purely on the bot's actual strategy/decisions.

Now of course we can think of poker (and a GTO strategy) as simply a game where we have this enormous table of all possible situations and EV values for each possible action, and we just choose the action with the highest EV... But poker still remains a multi-street game and the EV of each possible action should in part be "because" of the information about our opponent's range passed down to future streets.