My girlfriend Becky (not her real name) returned tonight from Tennessee, and brought back some really nice moonshine, which I must admit I have been sampling to excess. I don’t see anything in the “Community Playbook” about posting while drunk, so screw it, let’s talk about the pipe dream of game theory optimal poker.

Ain’t gonna happen, ever, period.

It’s like asking someone to design an optimal race car. Optimal for what? Optimal for driving 1/4 mile in a straight line? Optimal for a road course? Optimal under NASCAR rules? Pro crews set up a race car for a race type, for a specific track, for specific race conditions. There’s no “one size fits all,” and there never will be… they have evolved past all that.

Eh, it would be easy for me to ramble on for 6.000 lines on how I feel about this subject. As someone pointed out in another thread, NLHE is a long way from being “solved,” and as far as I’m concerned, it never will be. Math and game theory will only take one so far. Yes, you should have a solid mathematical and theoretical foundation… no doubt about that, but in the end, you’re playing other people.

In the interest of brevity, let’s jump start this with a look at a game that IS solved… rock, paper, scissors. The GTO “solution” would be to throw rock 1/3 of the time, paper 1/3 of the time, and scissors 1/3 or the time, and to do so with no discernible pattern. Such a strategy will be unexploitable.

If both players used this strategy they would, on average, tie. They would be at Nash equilibrium, and any wins would be purely the result of luck. But what if only 1 player was using this GTO strategy?

Let’s say player 2 was throwing rock 50%, paper 25%, and scissors 25%, and look at it over 1200 throws…

The 600 times player 2 throws rock, player 1 wins 200, loses 200, and ties 200, for a net gain of exactly 0. The 300 times player 2 throws paper, player wins 100, loses 100, and ties 100, again for a net gain of 0. Same when player 2 throws scissors. One size fits all GTO gets you nowhere. It’s like Mohammed Ali’s “rope a dope,” a defensive stance.

The theoretical “exploitative optimal” strategy here would be to throw paper 100% of the time. You would win all 600 times your opponent throws rock. You would lose the 300 times they throw scissors, and tie the 300 times they throw also throw paper, putting you 300 units ahead over the 1200 throw series.

But c’mon, nobody is gonna let you throw paper 1200 times without making an adjustment to their own game. From this we can learn at least 2 things…

Undisguised exploitation strategies can themselves be easily exploited.

“Unexploitable” strategies are basically defensive.

For me, the take-away is to develop a solid base-line strategy, but be willing to deviate from that strategy as needed in order to benefit from the situational opportunities that arise from the human factors, tournament stage, and a myriad of other factors.

Most mistakes are frequency-based. You fold too much or call too much, or raise too much or whatever. Learn to spot your opponent’s frequency mistakes while minimizing your own, and you will be on the way to better poker.

Yeah, math and game theory are the foundation, but as Doyle Brunson says, “You ain’t playing solitaire kid.”

4 Likes

4 Likes

I think he should drink moonshine and post more often Hahahaha.

This is where my request for a chat room 24/7 comes in handy

1 Like

SPG -

A ton of what you said makes perfect sense, however, the player pool is changing and doing so at a rapid pace. I think this statement is fairly widely held as a truism at this point: The last successful pro without a solid GTO base has already been playing for a few years. As more and more new players learn the game, all of them who wish to make their living at it are using this as the foundation of their games. Since more new players will play more total hands this year than most pros played in their entire lives, GTO theory is quickly becoming ubiquitous. Therefore, the argument that GTO isn’t playing the players is quickly becoming outdated as all the new players are in fact playing based off GTO theory.

Just like every other game has done, poker is going through one of the periodic revolutions in the way it is understood and played. Pure feel players simply don’t exist any longer other than in Hollywood. Even the most flexible / adaptable players today have the math down. That was 1 evolution. Now a next step is being made until a new equilibrium is established and the game settles in for a while.

Without rambling on forever about it, my thoughts are that whether we like it or not, this is the way the game is progressing. Online games for any stakes of 200NL or more are flat tough already. Live games are some of the last places where you can find soft games and players at stakes decent enough to be worthwhile. Since there is money to be made, eventually this gap will close.

I look at it as an information arbitrage. For a while, there will be profit to be made for as long as the information gap exists. Because there is profit to be made, that gap will absolutely close over time to zero. That’s just the nature of all things with profit motives and poker is no different. Players standing on one side of the information gap will eventually be shut out by those who have crossed it.

None of what I am saying discounts the importance of flexibility and the ability to adapt to situations. That will always be a key element. However, just as math-based players crowded out feel-only player a long time ago, GTO based players are crowding out the basic math players today.

2 Likes

Another thought on the subject: without knowing another thing about how it came to be, the demand to create an unexploitable strategy in response to how the game had progressed was inevitable. When poker went online and analytical software was developed, a players habits and leaks became very easy to spot once anyone had enough data on them. It was no longer a matter of SPG sitting at a table and after a few orbits picking up on something to exploit. Instead it was taking all the hands and tossing them into analytical software and being told what the exploitable plays were against this player. Since online players were producing massive amounts of data, all exploitable behavior stood out like a sore thumb and changes in behavior by the player were quickly discovered and adapted to.

So, now players had to come up with a better way of not giving tells than simply wearing sunglasses or covering their necks with scarves. There is nowhere to hide from the data except through changing the data you are producing. You mask everything you are doing in a balanced approach so that the data doesn’t reveal anything exploitable at all. Well, that is the goal of it anyway and you can see why something like GTO was desired.

For 99% of players, I don’t think this matters at all right now. Almost everyone who plays is doing so recreationally. However, for people trying to make a living at it, I don’t see any way around learning this approach as the foundation of your game. For people like me, it is relevant because I’d like to get better at a hobby and because I find it interesting even if I never played another hand of poker in my life. At some point in the future, even the recreational player will learn the game based off this approach though or will never be able to even think about entering a tournament of any size or stakes.

3 Likes

I’m no expert on GTO play itself, but from my understanding it works for exactly the reason you stated, poker hasn’t been solved yet. Players are continually making mistakes, so by playing a GTO optimal strategy you will not break even (as you would against another GTO optimal stategy), you will in fact profit off of the non-optimal strategies of others while not being exploitable yourself.

Exploitative strategies can be more effective against inflexible opponents, but are at risk of being exploited themselves. For example, the optimal strategy against most players on replay may be to bet for value with a tighter range to exploit passive callers, but that strategy would lose against a more GTO optimal opponent in the long run.

The caveat to this is that playing a GTO style incorrectly can make you even more exploitable and lose a lot of value, but that doesn’t mean the GTO approach itself does not work.

2 Likes

Good article on the state of AI and the quest for GTO. It is a continuum of course and the fact that we are not near perfectly solving the game does not negate how far we have come to approximating an equilibrium. As with everything, there is a law of diminishing returns. The advantage gained by attempting to play a GTO style is huge over those who aren’t playing this style at all.

The incremental gains of playing more closely to the equilibrium become smaller and smaller as we approach the solution. Therefore, the advantage is greatest while there are still players not utilizing this theory at all and will be reduced over time as all players eventually take on a GTO base. This is clearly shown in the progression of AI vs humans to AI vs AI in the relatively recent past. Each version steps closer to the solution and outperforms the previous version but by smaller and smaller increments each time.

First of all, I’m not posting here in order to school the top professional players. The last time I looked, there weren’t a lot of pros earning a living on Replay.

Note also that I said, " …math and game theory are the foundation…" and this is increasingly true as was pointed out by Warlock. I’m also not suggesting that “pure feel” is even a valid concept, let alone an effective strategy.

So what is the perfect GTO strategy?

Nobody knows.

What I do know is that many casual players find the subject to be so intimidating that they don’t bother to learn the basic principals involved, and that’s a pity. It’s really a fairly simple concept, though its application can be much more complex.

Game theory in poker basically means randomizing your actions so you don’t present patterns that can be exploited.

There was another thread started by JoeDirk about a quiz question. If you answered this question “fold” or “call” or “raise,” you can be exploited. My answer would be something like, “fold 60%, call 30%, raise 10%.” To me, that’s a game theory approach.

The real question here is, “are those percentages optimal?” They didn’t provide enough information to make that determination. Against a nit, you would fold more, against a LAG, you would raise more and fold less. The optimal solution is about the environment… It’s all about frequency.

So no, there will probably never be a single “one size fits all” GTO strategy.

So let’s talk about some of the basics and how the more casual players can bring them into their game.

2 Likes

One more point…

There are huge differences between live poker, online poker in general, and Replay poker specifically.

In live poker, you have plenty of time usually, but not much data.

In online poker in general, you have access to much more data, but not as much time.

On Replay, you have little time and virtually no data.

So, for Replay, that last fact should and MUST be part of the process. You simply don’t have the time to get in the weeds and do a lot of number crunching. Even if you did, you won’t have enough data to make it all that reliable.

So to me as a Replay MTT player, the essential skill is to be able to quickly and accurately profile the other players. Ranges, betting patterns, and other tendencies have to be internalized in real time on the fly. The math and game theory aspects of your game have to become second nature… you don’t have time for anything less.

2 Likes

Thought this was an intellectual exercise brought up by a quiz from a GTO training site. I was also a bit confused by the thread title “The GTO thread”. I think it would be great to start a thread specifically tailored to nano/micro stakes poker strategy that could be applied here by a large percentage of players.

I will stipulate to everything you said about casual players but simply take a different view of what is going on here. If we take your example of NASCAR, then no there isn’t an optimal car for every track or race. However, you also don’t see many designs resembling panel trucks on the track. Somewhere along the way, people started looking for optimal designs to reduce drag or to boost power to weight ratios or any number of other factors that go into the car.

Same thing happens in every single other sport or game or activity. I have the Senior PGA Championships going on about a mile from my house today. Guys in their 50’s+ are hitting the ball farther than they did when they were in their 20’s. They benefitted from tons of research and experimentation looking for the optimal designs of clubs and balls, along with training and so on. None of them are out there trying to compete with the equipment used 20-30 years ago.

All of the efforts in improving race cars or clubs or rackets or whatever were started to give the top players / teams in their fields the edge. That’s where the money was and where edges mattered most. Eventually however, all of these improvements filtered down to even the least skilled recreational players just starting out in the game. Now no one expects the average 30-handicap golfer to understand launch angles and club head speeds and force transference but he benefits from all the work others have done in those fields regardless. The family sedan that gets 30mpg today that used to get 15 benefits from advances in racing technology whether they know it or not. How any people know or care about why they have something called Gatorade? They wouldn’t had U. Florida football not spent time trying to figure out how to give their team an edge in rehydration and electrolyte balances.

So, some of us find it interesting to see where the game is going and to try and learn about it. I think that’s great. Some people don’t want to hear about it. That’s fine too. Some people want to know and understand the odds of filling an open ended straight flush draw on the turn. Some people are content to know that its possible and that is enough. Everyone gets to pick and choose what they wish to study, if anything at all.

Always appreciate your insight and practical approach, especially when it comes to playing the game as structured here. The object is to win the chips from the people you are playing at the time. I don’t see any harm in exploring what players at the very top are trying to do in order to get the chips from each other, especially as the WSOP main event is underway.

Just to add a bit - people thought chess was dead when Kasparov was beaten by Deep Blue. It was solved and too complex and what was the point. Turned out that chess became revitalized with more people playing than ever. Sometimes the new and complex can attract as well as repel. Different things do it for different people.

1 Like

Take it down one more step - at the most basic, the concept is to “Stop telling the people trying to take your chips what you are doing”.

The more information you give to your opponent, the more you expose yourself to being taken advantage of. Say you are playing 1 person and only raise 1 hand, AA You limp everything else. Well, once your opponent knows that you only raise 1 hand, how do you think you will be able to profit from that play? You are going to be exploited. Expand the possible situations and number of players and you get to the issues GTO is trying to address.

Exactly!

And that’s what I am trying to do here… demystify basic game theory in order to make it more accessible to the Replay community at large.

I’ve been toying with this for at least 14 years. I’ve tried a lot of things that didn’t work and a few things that do. I’m not against GTO. It’s the future of poker, and as you pointed out, it will trickle down to everyone eventually.

One of the things I struggled with early on was how exactly to randomize my actions. One of the ideas I had was to let the cards do the heavy lifting. To this day, there are specific card combinations that I will bluff with from any position, and I’m not talking about premium holdings here. Since the specific combos change from time to time, it doesn’t get more random than that.

It’s possible to sum things up with 3 words… “mix it up.”

3 Likes

Familliarity breeds contempt, while Watson could do reasonably well… the power of Watson dwarfs what big blue did to kasperov … There was no big money in chess, poker is a different story… Anyone with a AI good enough to consistantly pull 5-10% profit, can exploit online gambling just as high frequency trading revolutionized the stock market.

It only took 1-3% shift in odds for Gambling to be profitable in general. Many ppl play Hold’em against other ppl cause that way they aren’t play’n the house, but once the house has bots on every table then no longer are you playing against ppl and it ruins Poker. The house takes a rake or %, and that should fund the house, but we all know ppl are greedy… One of the next revolutions will be forcing players to give face-time, so then all know noone are bots.

Then instead of bitching about the algorythm for RND, we’ll be debating thier algorythm for thier bots. Then you’ll have the App that gives us all the power of Watson, displaying realtime data on players, and watching out for bots…

As in olden times, live games will be sought after all the more… due to the fact noone will “trust” online anymore…( what is GTO ? ) Poker to me is as much about psychology as it is about odds/math… balance the 2 and don’t beat urself, and usually you’ll do ok.

1 Like

1 Like

LOL Craig!

1 Like

I’m still laughing at the moonshine thing. SPG makes me laugh, good wit Lol

hi

hi