If you don't want to talk about game theory, then don't. I would kindly ask you to just stay out of threads where people are discussing aspects of it.

So the bit about behavioral psych/economics as a method of finding methods of exploitation is valid. While different from game theory its still a science based theory and worth exploring

but this bit:

makes no sense.

Options can be mathematically analyzed (imperfectly) only because there is market of (theoretically) perfect pricing of the underlying security, volatility, etc. I don't see at all how that has anything to do with behavioral economics, or games of imperfect information.

What was that, "if I don't want to promote GTO, don't post on this site?"

Good sir, it is game theory OPTIMAL that you meant just then, game theory is different. Game theory is basically a mathematic explanation for the way in which all beings strategize, it covers both exploitation and GTO. Game theory optimal is just one strategy that can be applied using game theory.

There are many threads and forums on this site, you're welcome to post in many of them. Your schtick of coming into threads where people are discussing a topic and taking a dump in them is getting really old. Stop doing it.

I'm correcting you... I wont do it if you don't want. Bye!

utk, I think your posts are excellent, and distinctly thought provoking, btw. I own a copy of Mike Caro's tells (book and DVD), and his section on chip stacking also has some of the risk aversion in them. The way he was able to synthesize what he knew about poker, and adjust what he knew about psychology to it, made him a distinct theorist (along with his ability to communicate to people with little-to-no poker understanding). I'd recommend his dvd for this specific topic.

This is all I am going to say about this as I don't want to derail what seems like a perfectly good thread but our minds are terrible at modeling anything of significance especially things that can be modeled perfectly with cold hard data like poker. See examples below.

https://en.wikipedia.org/wiki/Cognitive_bias

https://www.scientificamerican.com/a...probabilities/

Most of this thread is well above my head. I'm not gonna try to pick apart the stuff I can't follow.

I forget where I heard about this, but about 10 or so years ago someone won a Nobel Prize for showing that the average person will do a satisfactory job at whichever task they are given. If the average person did a better job, then the level of consistency that would be considered satisfactory would be raised to a higher standard.

A very simple example of this is batting average in baseball. If the average hitter had an average of .275 hits per at bat, then anyone that has a better average than that over a decent sample of at bats would therefore be considered a good hitter. Conversely, anyone with an average under .275 would be considered an unsatisfactory hitter.

What does this mean for poker? As the average player gets better at poker, the standard of play is raised to a higher standard. The strategies of yesterday become obsolete and the strategies of tomorrow become the present standard as time passes.

I tried to put this concept into poker terms in this thread a few years ago:

https://forumserver.twoplustwo.com/1...27/?highlight=

I agree in the sense that I don't think poker fits nicely into existing option pricing models for the reasons you say. However, there are some common characteristics between a not-yet-made poker hand and an option, such as asymmetric payoffs, in particular payoffs which are zero a lot of the time, but disproportionately large relative to the initial cost otherwise. There is no underlying per se but you could compute probabilities of different payoffs, just as in option pricing. Maybe this is what the OP is thinking. But yeah you would not use EV to price an option, you would use arbitrage / hedging arguments, and those dynamics are definitely quite different.

A bowl of cashews and a game of poker.

In the book “Misbehaving: The Making of Behavioral Economics” Richard H. Thaler tells the stories behind the creation of behavioral economics as well as covers the basics of the science and how it differs from fundamental economics.

Way back in the beginning, Thaler had a list of strange behaviors that had no basis in fundamental economics, yet were common among humans.

Most of these examples dealt with similar themes such as loss aversion. However, one on the list was different. One was more subtle. It was a bowl of cashews.

Thaler retells the story, that he and some friends were waiting for a dinner to be ready. Everyone was hungry, so a bowl of cashews was presented, and everyone had a snack. Then the bowl was removed, lest everyone run the risk of spoiled appetites.

This may make perfect sense to a human, however this is not economically sound behavior. This is misbehaving. Every economics professor will point out that more options are better, and no econ would ever choose fewer options on purpose.

Basically, if you decide to eat cashews or not, they should remain on the table. This is optimal.

So why does it seem so logical to remove the bowl of cashews? What is the dynamic here?

This is the human behavior of self-control.

The very need for self-control is human. Humans can plan ahead and make decisions in the present that influence decisions in the future. When a human limits a present choice to prevent a perceived future mistake, this is self-control in behavioral economics.

In a game where sophistication is the edge, why is the careful practice of irrational choice avoidance considered a skill and not a weakness? Well, this is because poker is a game amongst humans, not bots.

Self-control and bank roll management.

I am of the opinion that no self respecting econ-bot would ever have less than five percent of liquid net worth in front of them on a poker table. This number is a rough guess, but could be more precisely figured by the odds the bot expects to encounter in a poker game.

Having “the nuts” and getting all in and called is an obvious example. To be limited in such a scenario to a fraction of net worth less than five percent would be inferior bank roll management, to a bot.

A bot would sit down at a stake where the maximum buy-in was less than five percent net worth, and reload exactly five percent net worth each time unless the bot could no longer meet the minimum buy in with five percent, at which point it would drop down a stake and start over the same process.

I could be wrong, but I don’t think most human poker players operate this way. Playing NLHE with five percent chunks of bankroll is pretty aggressive, and moving freely up and down stakes is usually not what we humans like to do. We plan ahead, foresee variance, and bring smaller bankroll chunks to the table and linger at the stake we feel comfortable with.

We know we are not perfect, and do not want to allow our future self to be unwise. Maybe future self could go on “tilt” and spew some precious bankroll against tougher opposition. We like to save up enough “extra” bankroll to “take a shot” at the next higher stake. Econ-bots would never do such a thing.

Proper bank roll management is an essential skill in poker, and yet is entirely irrational from a fundamental economic perspective. If a bot has an edge, the bot will be ready to maximize it. A bot will never limit its present option to avoid a future mistake.

There is obviously something going on inside the head of a human being, a sort of dichotomy of purpose.

This is old hat to psychologists, but goes to the heart of behavioral economics. Thaler calls it the “planner” versus the “doer”. The doer will consume resources immediately up to the point of zero utility in the consumption. A planner will deny present utility to preserve future utility. The two are always present in human decisions. Whether we know it or not, we have both and the varying strength of each is one way to measure self control.

A poker player who makes a plan and sticks to it, can be formidable on the table, so long as the original plan is +EV versus having no plan at all.

Since GTO should never need a plan, other than a mathematically perfect amount of money on the table, any +EV implementation of self-control must be a human exploit. We sit down at a table because we expect to outplay our opponents by implementing our plan, and exerting more self-control.

What about within a single hand of poker. Do we see tactical implementation of self control? I think so, and it is always, by definition, an exploit.

I am not a GTO poker player. I know this. Fortunately for me, I only play poker for fun, at stakes that are small enough so that the money is JND, so that I can concentrate and practice while still feeling the pain of losing a buy-in.

But even down at my stakes, there are players much more balanced than myself. I try to keep an eye out for such players, and respect them differently than the other players. What does this mean for a hand of poker? It means checking or folding on some flops, with thin value that should be continued with, optimally. If the flop is tailor-made for GTO balanced bluffing, and my holding is not strong enough to get involved with pot sized bets on later streets, I will just bail on the hand without further investment. Conversely, if my holding is much higher in value range, such that I can improve to the nuts especially, then I am much more likely to continue in the hand and hopefully reap some benefit from optimal bluffing from an opponent that may not notice I do this, and adjust frequency accordingly.

What about live pros. Do we see some live high stakes professional poker players limiting options to attempt to outwit opponents?

One example comes to mind, and I am sure it is an exploit, but not sure if it is +EV.

When a live poker player is “out of position” meaning they have to act first, they can do a maneuver called “checking in the dark”. After the conclusion of preflop betting, but before the dealer reveals the flop cards, the first player to act can announce “check”. Such a player has given up the option to bet out the flop and attempt to win the hand first. What they get in return is to see what the opponent decides to do with regard to betting the flop and reserve the right to act until after doing so. This is en exploit, and no bot would ever do this.

As a matter of fact, I do not believe that a bot would even avoid a “donk bet”, which is a first to act bet that precedes an opponent who was the preflop aggressor. Surely such bets would happen from time to time in a perfect implementation of GTO.

However, at least in No Limit Hold’em, humans tend to avoid these bets, as part of poker negotiations. So much so, that “checking dark” is simply acknowledging that a player did not intend to ever make such a bet, regardless of the flop. Such a player is saying:

“ I don’t need or want my option to bet first. I would rather give up this option than make the mistake of such a bet. You will not get information about my hand by any reaction from me to this flop. However, I am watching you and intend to outplay you after you act.”

Such a player is highly exploitive. A GTO player would have to have at least SOME kind of hand that would merit a donk bet some percentage of the time. A GTO player would simply look at each flop, and decide accordingly.

So, does the dark check really give an edge to the player out of position? Only if it influences the player in position. If the player in position can stick to the plan they brought to the table, and view the dark check as simply a check, then the dark check is just a failed attempt to exploit.

Do most GTO simulations and calculations allow for checking in the dark (I'd guess not)? I.e. is there a line of code which distinguishes "check" from "check in the dark". If not then the simulations are actually technically being run on a different game than a live one, where the option to forfeit your turn before seeing the flop actually does exist.

Maybe forfeiting is always sub-optimal, or maybe it just reduces from a GTO standpoint to the same as seeing the flop and checking. It would seem likely, as more optionality is better, but can it be proven?

My two-cents:

In low-to-midstakes it's perfectly fine to tight-up a bit after losing a few hands, bluff a little less, and if you've been caught bluffing, value-bet a little more, and other small adjustments, because your opponents will tend to over-adjust in this circumstance. At one point, I use to "play a rush" when up, and bluff more, and try to win more pots. Now I think that's bad, and it's also common.

The main idea though is that you have an idea of what "perfect poker" looks like, though your picture should be incomplete and condensed so you can implement it, and then you can make small adjustments. Taken to the extreme, the detractors of game theory and math would need to consult a psychologist before raising AK under-the-gun in a six max game. Why? Because they have no frame of reference. This is why you should have some ideas about game theory independent of the live play psychology component no matter what game you're in (back to my point earlier), and that element is primary. And this is also why you make adjustments from it, and you make adjustments to opponents who error according to it.

If you're truly playing close to GTO you are the smartest man in the room, but for the most part, we're all talk when it comes to GTO, and we implement some strategy concepts that we've picked-up along the way, which is fine for mortals.

Add: dark checking is not GTO. This is known, because there should be some bet-folds worked into almost every situation heads-up from the blinds. The way to answer questions like these is to look at simplified versions of poker (aka. model poker), and then apply the knowledge to real poker (played at the casino). There *might* be some super scary boards where you are checking 100% anyway, but checking light loosely dominates checking dark, because there's no reason to let your opponent know you don't have any bet-folds on a very specific board. There is no advantage checking dark has to waiting a few seconds and checking light.

I disagree that econ-bots would not take a shot. The bankroll management techniques you describe are solid, even if aggressive. Though an econ-bot is not a finance bot and doesn't know what else you can do with the money, whether they should be tied into stocks, etc.

To elaborate. Suppose Game B plays 5x as big as Game A, but you're over rolled for Game A, and under rolled for Game B. The way to solve this paradox is to sometimes play Game B, and sometimes play Game A, and have it work-out on average according to your role, and your preferences regarding role size in the future. The ratio is adjusted until you're in action for Game B. Correct me if I'm wrong.

Checking dark isn't bad as an exploit where you're enticing your opponent to continuation bet. If out of friendliness, your opponent will continuation-bet 100% of the time to your dark-check, then you should always dark-check. This can be reasoned because the pot has one more bet in it (from the continuation bet), it's still your turn to act, and the ranges are roughly the same (you've gained information if bet-sizes vary regardless). So situation one where you do not dark check has a smaller pot than situation two where you do dark-check and your opponent reciprocates with a bet.

GTO solutions do not add a dark-check in the range as it's superfluous. With computational limitations, it's a little much to throw-in this kind of sauce that a human can do on their own in the casino. Game trees are already too large to explore as it is. Artificial intelligence poker is differentiated from live casino poker. There's nothing saying you can't add a line of code for dark-checking, but that added complexity explodes the complexity of the game. Adding a dark-check range is presently not a worthwhile question for machines.

If a “check dark” button could preselected online such that you are closing the preflop action and first to act postflop, would a bot ever use such a button? Would there be any computational advantage by giving up a future choice?

No.

Such choices are irrational and human.

If I were to be proven wrong on this point, then a check in the dark could be proven to be a fundamental strategy in poker, such that mixing it in against random opponents is +EV.

This is splitting hairs, but goes back to the “lens” through which we categorize theories as behavioral or fundamental.

Something that strikes me is that we are talking about exploiting deviations from rational play by non-rational actors.

How can we say with certainty we know how non-rational actors will be have to our adjustments? Is it basically we are gathering data and using statistical analysis to classify and adjust? If so how do we avoid our own biases when playing live that come from our own non-rational econ behavior?

Interesting thread btw. Just had these questions pop in my head so wanted to discuss.

I guess more succintly:

Why do we believe we can manipulate situations with non-rational actors in a predictable way?

and

If that is based on imperfect data analysis (say because of our own heuristic biases in live play) then how do we expect to predictably deviate from there.

Or much more succintly

Since our data can be crap and our targets make irrational decisions, how do we know these deviations from fundemental play will ever be effective?

Hey just_grindin, welcome.

Studies show that humans lean predictably toward some biases, and these biases are the basis of behavioral economics. Of course there is no certainty here, just likely biases that vary in severity with the level of sophistication or naiveté of opponents.

Also, our own selves are vulnerable to these biases, and our own sophistication is our only defense. I suppose an inward dive of self-improvement could root these out, but there is no guarantee.

Probably just practice at observing these behaviors in others might make us less likely to make biased irrational choices. It is unlikely that I could calculate my own Nt number, which could represent my maximum tolerance for losing before I go on tilt. Also, simply acknowledging that tilt can happen, does not make it less likely to happen.

Hopefully I would notice that something is wrong, mentally review the last hour or so, see that I am down and not processing events correctly and sit out from the table.
But, all of that is the last concern.

First, is to be fundamentally sound.
Second, is to be aware of behavioral biases of opponents, and make adjustments to the fundamental game against those players.
Third, is to attempt to eliminate our own biases, or institute rules which prevent our biases from costing us money.

If behavioral theory is correct, such that it predicts human decisions involving strategy and money BETTER than the most precise fundamental theory, then it must be present on a poker table, and it must offer a more refined and more accurate prediction of decisions of human participants.

So, no matter how crap our data is, adding in the principles of behavioral economics will make that data less crappy.

My hunch here is something about every hand of poker being a financial market.

The market starts small and humans take turns investing. Hole cards are “showdown options” and revealed hole cards are “showdown stock”.

Investors vary in their intent towards the market. Some investors are trying to increase the value of their options, while others are trying to crash the market entirely and are investing to short holdings of competitors.

The market participants vary with sophistication at this task. A perfect investor will get the maximum overall equity from the investments and usually profit from the imperfect investments of competitors.

The more naive investors will make predictable mistakes across multiple markets, where the fundamental investors only use data from within each market to make decisions.

A perfectly sophisticated investor will notice these tendencies and actualize extra profit from the markets by predicting the mistakes of the naive investors, instead of treating those mistakes as purely random.

Yes, that's correct you do not know, even as a human what your opponent is going to do. But, you can hypothesize, and deviate based on the strength of your hypothesis. So, if you hypothesize that they'll bet if you check dark, you can start by doing it rarely as to not be exploited too much, and then if the hypothesis gets stronger, you can do it more. If it gets weaker, you can back-off or stop the exploitive play.

To me the timing tells are more conclusive, but you have to be somewhat weary of accusing someone of being a bot by playing certain streets in a seemingly robotic way, if that makes sense. Of course, someone could play on camera for a few hours and generate a few graphs, and then play off camera, and if there's a difference, and it appears robotic off camera, but not on camera, then that's compelling.

Of course, you do not have to be 100% sure to ban someone. Given the circumstances banning based on the graphs seems acceptable to me.

Be weary though because collusion bots might not be detected. That would take an entirely different set of tools. High stakes collusion bots are a very real threat.

Sorry for the derail.

Yeah I think I was just conflating irrational and unpredictable, even though you stated clearly in this context and in behavioral economics in general irrational actors deviate in predictable ways from some idealized model where actors are assumed to be rational.

I think my concern about data collection still stands since we as humans are bad at collecting and using empirical data on our own (i.e. without tools to aid us).

I do like your 3 step process answer though I would switch step 2 and 3. It is probably easier to recognize and resolve our own behavioral shortcomings than recognize and take action against others mostly due to sample size issues.

@robert_utk I recently read a book that I thought you might find interesting:

https://www.amazon.com/End-Theory-Fi.../dp/0691169012

It is about how financial models and modern economics have failed to model the human side of things and thus have become essentially useless in setting economic policy and/or forecasting crises. Although specifically geared towards the financial field there are a lot of human biases and behaviors discussed which are relevant to some of your posts. It is well written and rather erudite. However it's not a behavioral economics book per se and also I found the conclusions / last half of the book rather light / disappointing. Anyway, take or leave at your pleasure.

With a title and a synopsis like that, I simply must read that book. Does sound ambitious though, so I can probably forgive the author if the end result is light.

The portion of “Misbehaving” that I am reading is dealing with the whole disconnect between rationality and theories of humans.

Basically, rationality is not required for a theory to be valid. Theories can be constructed however the theorist chooses, and then proven or disproven by thorough experiments with replicated independent results.

If rationality is unnecessary, and leads to flawed results, then why the heck are theorists obsessed with fundamentally rational theories?

In poker, what is driving the supposed inefficient exploitive players to behave more like econ bots?

Is the movement caused by diminished results of exploiters in a tougher field of players? Or is the landscape of players changing because every book and video about poker is pushing GTO, and exploiters must adapt or go broke.

Ugh, this one is expensive, but would appear to be THE textbook of behavioral game theory. I bet I can find older editions on eBay though. Got to read this one.

https://www.amazon.com/Behavioral-Ga...dp/0691090394/