I don't know if anyone has had this idea before, but I've come up with a mode of play which completely removes game theory from poker, called Strategy Free, All Tactical Poker, or SFAToker. Sfatoker creates a completely static tactical situation, in which the only information available is contained within the hand, as it progresses: each hand is totally unconnected with past or future events.

This has a distorting effect on the way the game is played, which is something it has in common with the Prisoner's Dilemma. In both cases the difference is iteration: if the Prisoner's Dilemma is to be presented to the same players again and again, then past and future events may influence the outcome, but if it's a single event, the logic of the situation is entirely different.

Rules: Sfatoker can be played with any poker variation, in a multi-table online cash game with a large number of players. All players are completely anonymous and are moved to a new table, with new players, and a new stack, at the start of each hand. So each hand is the first, and the last you will be playing against those players.
.........................................

It's not an attractive proposition to actually play the game, but it's an interesting hypothetical. Have a look at what happens to game theory, and bluffing frequencies, and the need to vary your play... I mean, before you read on.

DDDDDDDDDDDDDDDDDDD

Let's look at what happens in a standard bluffing situation, where you would normally bluff in a suitably randomised fashion, at some prescribed rate. What should you do in that situation in Sfatoker? For one thing, there is absolutely no need to randomize your play, because no one knows what you did last time, or cares what you are going to do next time. So randomization is unnecessary. And it is also pointless bluffing at a particular rate, because again, no one is watching what you are doing, over time and you can bluff 100% of the time, and no one will know you are doing so.

Similarly, there is no point in calling to discourage your opponent from bluffing in the future, because you have no possible connection with those future actions. If the call has a negative expectation, then you should not call.

In fact, the only time you should bluff or call in Sfatoker is when you have a positive expectation in that situation, and all similar situations. And if that is the case, you should make the same bluff or call 100% of the time, in all tactically identical situations.

Conversely, if a bluff or call has a negative expectation in a particular situation, then you should never bluff or call in all similar situations.

It's a head-spin, but in Sfatoker, there are no situations where you should bluff or call only sometimes: it's all or nothing.

In actual poker, the potential value of many situations can only be extracted by game theory type moves, which have effect over time. In the static world of Sfatoker, the inherent value of many tactical situations may be insufficient to generate a positive expectation for a bluff, and therefore that value is completely lost from the game, leaving the actual value of a player's concealed cards as the only factor under consideration, in most situations. In other words, by making game-theory-type actions pointless, Sfatoker not only makes strategy impossible, it also destroys a lot of what we think of as tactical value, much of which turns out to be unavailable in the absence of game theory.

I'm going out on a limb here, based on some skim-reading and a wild guess, but Sfatoker might create a well-known game theory situation (which is new to me) known as a "Pareto-efficiency" which, to quote the wiki entry for the Prisoner's dilemma, means that:
"The unique equilibrium for this game is a Pareto-suboptimal solution, that is, rational choice leads the two players to both play defect, even though each player's individual reward would be greater if they both played cooperatively."

In Sfatoker, the equivalent of "defection" or "betrayal" could be the practice of bluffing 100% in certain situations, and never bluffing in all other situations, which is a betrayal of game theory principles. Game theory has a kind of co-operative, or responsive element to it, which is lost completely in Sfatoker. But I'm way out of my comfort zone here, so I'll mark this as "speculation".

How Sfatoker would work out in practise is an interesting question. Sfatoker could perhaps be reduced to a mechanical game, where the correct play in all situations could be read off a chart, as it can in blackjack. Blackjack is also a SFAT game, in which all similar tactical situations must be treated in exactly the same way.

But then, if everyone knew what to do in every possible tactical situation, where would that leave us? Perhaps it would not matter, because many players would try to second-guess the situation - as they so often do in Blackjack - and would not in fact bluff 100% of the time in a particular situation, which would encourage other players - potential callers - to also second-guess. In which case the optimal play, as read off the chart, would still be the optimal play.

But I'm really not sure. It is likely that the tactical complexities of Poker mean that different actions may have little difference in value, and that some players will take the options with less variance, but lower expectations, and others will take the high-risk path, with greater long-term expectations. So there would still be room for different styles of play in Sfatoker.

Unpleasant and one-dimensional as the play of Sfatoker would be, it would at least be incredibly fast, because as soon as you fold your hand, you are sent to a new table and dealt a new one. Perhaps in the online world it would be possible to sustain a game of Sfatoker for the same reason it is possible to keep a game of Badugi going: that is, even if players only try it once before giving it up as a bad joke, there is a constant flow of curious newbies, so the game keeps going indefinitely.


Thinking about Sfatoker has been useful for me, as it has focussed my attention on what tactical strength really is, and how self-defeating it is to surrender real tactical value for possibly illusory strategic reasons. I guess that my half-baked notions about changing gears have been costing me money, because tactically strong play is taking advantage of opportunities as they present, and if you stop bluffing for a while after you have been caught a couple of time, you are not only behaving predictably, you are surrendering all the value that those situations contain, which can only be accessed by bluffing. There is also the fact that each subsequent bluff has a greater strategic value, due to the psychological impact: if you have been caught bluffing twice in a row, then a third bluff has much more strategic/psychological value than a bluff half an hour later. But that's wandering OT so I;ll stop here.

Quit stealing my ideas:

http://forumserver.twoplustwo.com/15...2/#post8312358

sdfasdf32 say "Quit stealing my ideas" heh, thanks for the heads-up, It was a difficult subject to search for, so I thought I'd just go ahead.

I don't think there is much chance of me stealing your analysis, which as far as I can see consists of:


"My thoughts:

1. You would have to play more straight forward, no?
2. Because of having to play straight forward, this hypothetical would favor TAGs?
3. You should be 3-betting a lot more?...because the first 3-bet usually gets respect and every hand would be the first hand of a table."

Yes, until you get to high stakes where most people know what they're doing. Then start mixing up your play, since otherwise everyone will be playing the same game and there would be no edge.

Depends on the playing field. If everyone's a TAG, play like a LAG. If most people are LAGging it up, go LP.

Again, depends on the playing field and especially on the stakes.

So much confusion.

Ouch, harsh ... Just pointing out that I had the idea a while ago ... I do think this is an interesting topic, however it does take away the most fun part of poker, that is, adjusting to your opponent. Maybe special 'random' tables could be made, but I don't think they would work on the whole.

By the way: "In Sfatoker, the equivalent of "defection" or "betrayal" could be the practice of bluffing 100% in certain situations, and never bluffing in all other situations, which is a betrayal of game theory principles." Go argue with Bobbofitos about this one, I think he'll give you a run for your money.

There is no point in mixing your play, because you are invisible, and no one would notice. And also because if it's profitable to play one way, you should keep playing that way, because in Sfatoker the value of all identical tactical situations is identical, exactly as in Blackjack. You could, hypothetically, and probably should read your plays off a chart, exactly as in Blackjack. It would be a big chart, but not infinitely large.

The only reason you bluff in Sfatoker is if you have a +ev expectation for a bluff in that situation, and all others like it. It cannot be correct to bluff in four identical, +ev situations, and not to bluff in the fifth, or hundredth.

I don't understand all the implications of the play of Sfatoker (no surprise there) but I suspect that it would come down to poker heart: can you extract the maximum possible tactical value out of a particular situation by doing something scary and dangerous?

If we look at a really strong play, such asy Tom Dwan's unforgettable bluff against Peter Eastlake and Barry Goldstein on HSP, convincing them to lay down trip 2's, and pocket aces, on a flop of 10, 2,2, holding only a 10Q, I wonder, would that play be possible in Sfatoker?

Could Dwan have played like that without the strategic information he had regarding his opponents? That is, if all players were anonymous, would that play have had a +ev? If the answer is yes, then whenever you are in that situation, you should play just as Dwan did, every time.

My point is, extracting the maximum from every possible situation requires real courage, and in sfatoker there would be the additional burden of playing the same incredibly rash way every single time, and taking the huge hits of variance that would entail. So it would be a deep test of poker heart.

Remember, it's effectively an infinitely large game, and no one knows who you are.

Hey sdfasdf32, thanks, and you were definitely ahead with the idea. I couldn't see any name for it, so maybe sfatoker could do, but then, we may find someone else got there ahead of us, so who knows. It's a challenging concept, for me anyway.

Is Bobbofitos a game-theory guru? If so, maybe he would be interested.

I don't think I've ever been called a game-theory guru, although that's mighty fine praise.

With regard to SFAToker, more or less that's how I approach real poker, where I do feel every decision:
Randomization is unnecessary, in a nutshell.

Good luck in this endeavor.

Game theory is a complex animal. Read mathematics of poker and try and understand the nemesis. Without any way to understand player tendencies game theory is likely the _best_ way to approach strategy in your game.

Otherwise you would just assign player tendencies to the whole population of players you play against, but I'm not sure that this would work out.

That's what I think it would probably come down to. Treat the entire opponent pool like one opponent with a random strategy (with a distribution that matches the opponent population as a whole)

Hey there BobboFitos: nice to know that I am not completely alone in my
wonderings/wanderings, which are still very new to me. Could you point me at some posts which might help me avoid re-inventing the wheel on this?

I think one of the fundamental things about poker which is often missed is that it is in large part a guessing game, like rock-paper-scissors, or one of those pub-games where two players hide matches, and guess the total. Perfect randomized play cannot be beaten, over time, but nor can it win: expert players of such games do not play randomly, they find patterns of play, and exploit them. Poker, as a game of counter-strategy, is that kind of game, if played well. The rest of us suckers use a faux-game-theory approach, which is in reality a lot of guess-work and hope, and exactly the kind of approach that a real expert can crush.

My long-held feelings of unease at some of the assertions and aims of games theorists I now realize come down to the use of the words "optimal strategy": the optimal strategy in GT is actually a break-even, strategically neutral one, which relies on other players making mistakes; so how is that a strategy? It is no different from playing randomly in RPS, that is, it's what you do when you have no strategy, or counter-strategy to follow.

Hyper-aggressive players like Tom Dwan are not using game theory,they are scaring the hell out of their opponents and taking their money, in other words, profiting directly from bluffing too often by GTO standards. That also means that value bets on extremely thin values can be profitable, and that strong betting hands are paid off far more often.

Now that's a strategy, and it has nothing to do with game theory math, as far as I can tell. Randomization is sooo over-rated, and is a methodical way of destroying tactical value, IMO, so we seem are on the same page on that, BobboFitos. Like, as if Tom Dwan would have thrown dice to decide whether or not to bluff in the Eastgate/Greenstein play on HSP? Uh uh. He would not waste that opportunity, ever, and it doesn't matter if his opponents know that, because they don't actually know what he has in his hand, THIS time, and can usually find reasons to fold.

Hi mickb70 from Sydney,and RustyBrooks: the problem is that GT calculations can play no part in Sfatoker. It's hard to grasp, and makes my head spin, but that is the situation.

I'm thinking that "assigning player tendencies to the [field]" would be playing from a hypothetical chart of optimal plays? That is, if you have the historical data for every tactical situation, you can form an optimal strategy, and play it 100% of the time? It sounds weird, but that seems to be the situation, that sfatoker could be played off a chart, like blackjack: in practice, very few players would follow the optimal play at all times, because it would almost certainly involve a huge amount of variance, and courage.

I agree that you could essentially use a chart. You'd be playing against the aggregate player base, rather than any particular player. "Random" is an easily misunderstood word in this context - I don't mean that players act uniformly randomly, betting/checking/folding/etc at random. Rather, I mean, whenever there is a choice that can be made, say a bet has been made HU and your opponent can raise, call or fold. If 30% of the field would raise, 20% would call and 50% would fold, then in a way it's like playing against a single player would would raise 30% of the time (randomly), call 20% of the time, or fold 50% of the time. Operating correctly would require being pretty sure what the distribution of opponent actions would be.

If the player pool is infinite and varied this might not be possible or useful. But in truth, some choices would be much more common than others. Probably when you first introduced the game you'd see a lot of fluctuation until long-term regs settled on strategies, winning or not, and then the player pool would probably settle into a relatively unchanging distribution.

The only strategy GTO can't win against is its GTO counter strategy. It wins against all other strategies. I agree that the best players exploit weaknesses but they could not do that against GTO as it has no exploitable weaknesses. Now while a GTO is always a winning or break even strategy it is not necessarily the best strategy. And I'd agree that Durr and others would probably fare better than someone who magically could play a GTO strategy.

again its only strategically nuetral against other GTOs.

check out around the 1.00 mark when DURR talks about his river calling strategy with the under full against sahamies...

http://www.youtube.com/watch?v=nxLojsI1XK0

He talks about calling 10% of times with KK and about 90% of the time with his 7. No I'm not saying Tom is a game theorist but that is a very game theoryish thought process.

I agree its a head **** but I would argue that the opposite is true. Consider this ... in your game you can't assign a strategy to your opponent because you know nothing about him. For mine against a random strategy it would seem that an unexploitable strategy that would win or break even over any strategy it encounters would be a pretty good option.

That's a lot of words that, if I understand correctly, are explaining maximally exploitive strategies, instead of perfectly balanced strategies (that you call game theory).

Although game theory doesn't only concern optimal solutions for both players. If one player has a consistent leak game theory can help you exploit it. Anyways, this is a misconception about game theory. I'll use it as you did, referring to game theory optimal solutions (GTO).

From here on I'll also refer to SFAT poker as being the maximally exploitive strategy. I like typing 4 letters better and I believe it's the same thing anyways.



SFAT poker needs the conclusions from GTO solutions as well.

For example if our opponent folds just than his GTO solution, the maximally exploitive strategy is to bluff 100% of the time we have a bluffing hand, but when he doesn't it is never to bluff.

Still we need to know whether and how he deviates from GTO to make our decision, we just don't care about balancing our own range.

An example with rock paper scissors:

Villain picks rock 1/2 the time and the others 1/4 of the time?
GTO says we should randomly pick each option 1/3 of the time.
SFAT says pick paper 100% of the time, because 1/2 > 1/3 (the GTO solution).



Say you have an opponent that you exploit by bluffing 100% of the time, the SFAT way. Now he plays perfect SFAT poker as well. He knows your strategy and he can beat you by playing SFAT and calling you with 100% of his bluff catchers.

But say for some reason he doesn't know or learn your strategy, so he cannot exploit you for exploiting him. He can assure your perfect SFAT strategy earns the least by playing in a GTO manner.

So the only way for this to make sense is if your opponent has leaks and doesn't ever change his strategy (sounds like real poker ).

Now in reality people do adjust, so what we should to is play in a balanced way, but with small deviations to what SFAT would dictate so we can be more profitable without being detected or exploited for it.

I disagree that playing hands in a vacuum means that there can be no mixed solutions.

There is one glaring problem with your SFAT poker concept. You say the game has totally random opponents and then you talk about exploitative play. How can you exploit players when you cannot observe their tendencies. The game describes is actually a better game to play from a game theory point of view than normal poker.

FYP

A "situation" includes what you know or think you know about your opponent. A GTO strategy assumes that the "situation" includes a GTO opponent.

Given that, a GTO strategy never makes -EV calls or bluffs (compared to other actions).

Also, a GTO strategy does not really "mix it up" in that the strategy itself never changes. It may play a given situation a mixed way, e.g. bluff x% and call (1-x%) but only when bluffing and calling are equal EV. But since the x% never changes I would not call that "mixing it up".

The only way to improve on a GTO strategy is to exploit a non optimal opponent but in the scenario you describe we don't know anything about our opponent.

Best contribution ever! Why and how do you disagree?

Obviously playing one hand against a fixed, mixed strategy is going to have one optimal line, unless more lines have identical value, but we don't have to mix those.

I play this game on the first few hands of many of my husngs. You could also call it 'Bayesian poker' because your opponents are not 'random', but are drawn from a demographic about which you will have some experiential knowledge. I think it's in MoP that they show how you can apply mathematics to compute the likelihood that a player fits the maniac sterotype after raising the first x hands of a limit holdem ring game.

I agree that river or all-in call/fold decisions can be pure based on which way you think the demographic leans, but you will probably do better using mixed strategies on a lot of check/bet/raise decisions.

SquirrelsUnite said "So much confusion", and I'm not sure I can add to what I've already said without making things worse.

I can only say that the one thing I am sure of is that GT and randomisation cannot play a meaningful in sfatoker, which seems to be in doubt for some, which surprises me, because it's the central feature of the game.

What that might mean for real poker, if anything, is another question.

This is all new to me, and I'm obviously no expert, on playing poker, or playing Sfatoker, so I might leave this for the enjoyment and/or confusion of others for a while. Thanks for the interesting, though sometimes confusing and confused responses, and apologies for kind of opting out by not replying in detail.

Sorry about the double post, but I can't find the edit button for my prev post.

Couple of things: mvdgaag says
"From here on I'll also refer to SFAT poker as being the maximally exploitive strategy. I like typing 4 letters better and I believe it's the same thing anyways."
I don't agree: Sfatoker is a game, or a mode of play, not a strategy, and you should be clear about that; confusion is raining, lol. And I cannot always tell if you are talking about real poker or SFAToker in the analysis, so I'm reluctant to comment on what you say.

zalmkleurig says, correcting me and adding "non" to the original
"Game theory has a kind of "non"-cooperative element to it" FYP

No, I really meant to say co-operative, or responsive, in the sense that in real poker, players who bluff too much are "betraying" the notion that you should bluff according to some GT rate. And in Sfatoker, if it is a Pareto-efficiency like the Prisoners' dilemma, the players might be forced into strategies which are sub-optimal when compared with real poker, such as bluffing 100% of the time, or not at all, in each situation, ie total betrayal of GT principles. I may be very wrong, but please let me be wrong in my own way.

Thanks.

.

I suggest you stop using random terms from game theory if you have no idea what they mean. There's no "betrayal" in 2 player zero sum games and Pareto optima don't play any meaningful role either.

Bluffing too much isn't betrayal any more than playing scissors too much in rock-paper-scissors would be.

Say you and I play modified RSP. Loser pays $1.00 but ties I win $.10.

If we play 3 million games with no memory between games, I'm still going to randomize and take my sure ~$100,000 rather than risk losing it all with a fixed strategy. Maybe you're rich and like to gamble and choose the right fixed counter strategy all the time.