Which Game Requires Most Talent?

[Rudbaeck]

One way to think about it is that the bot plays perfect defense. It's not going to lose. It will however only win what your ineptitude gives it.

Phil Ivey plays damn near perfect defense, so the bot is going to beat him very slowly.

But the bot is also going to beat you very slowly, because while you play worse defense than Ivey it's still hopefully not really attrocious.

Ivey also plays damn near perfect offense. So he is going to capitalize on your mistakes majorly. While you both lose to the bot it will take Phil longer to lose to the bot. And Phil is going to bust you much, much faster than the bot does.

Get the metaphor?

[vypremik]

I love to hear all of this talk about the perfect Bot and how it could beat any Hold-Em player by playing defensively.

For those of you who are on the border - don't believe it.

First of all Hold-em is not only a game of skill, but also a game of luck. If what these poeple are claiming is true, that there is a perfect bot that will always win, and that players like Phil Ivey nearly mimic the bot, then Phil should ALMOST NEVER lose!

I give Phil a lot of credit, he is absolutely one of the top five players in the world at what he does (in my opinion), but it NOT beacuase of his perfect defensive play (although he is good at defense), it is because of his varied aggressive attack.

The perfect defensive bot would probably lose to the majority of the players on this forum, amateurs included. If the bot holds KQ and there is an ace on the board, the bot will ALWAYS fold to a big bet. Why? Because it is the right thing to do!

Give me that player anytime!

This talk about a bot reminds of of all of those sci-fi movies of the seventies that showed intelligent computers taking over the world. Computers and bots alike only do what they are told. That is all they can do. That is all they will ever do.

BRING ON THE BOTS!

Actually, I think this worth a post of its own. So I will.

[Rudbaeck]

[ QUOTE ]
The perfect defensive bot would probably lose to the majority of the players on this forum, amateurs included.

[/ QUOTE ]

You extrapolated how the bot would play from a metaphor I used to discuss Game Theory in general. And you reached an utterly incorrect conclusion.

I used 'defensive' because the math used actually attempts to calculate the optimal strategy for both parties, without any attempts to abuse the weakness of the opponent.

It's not even a theoretical discussion about GT bots for HU play, you can buy one off the shelf. The AI department at University of Alberta built one as part of their research. (Which also includes a HU bot that _does_ try to attack your weakness.)

For a more thorough overview of GT in general read the Wikipedia article: http://en.wikipedia.org/wiki/Game_theory

[ QUOTE ]
If what these poeple are claiming is true, that there is a perfect bot that will always win, and that players like Phil Ivey nearly mimic the bot, then Phil should ALMOST NEVER lose!

[/ QUOTE ]

This part you actually did get right. If Ivey plays HU against someone, and they can both reload if broke, then after a few thousand reloads Ivey is going to be ahead against almost everyone. He will only be behind or even on a very rare fluke, or against equally skilled opponents.

Obviously I didn't mean 'never loses' as in 'never loses a single buy-in', but as in 'given an infinite number of hands it's won't be behind'.


For every other poker scenario I totally agree with you. Bring on the bots. At a full table I have no doubts I can beat the best bots in the world right now, and I am not very good at poker. But I am just as certain that no living human can beat the best bots HU. Just make it three-handed and the humans have the advantage again. Because the HU game can be, and has been, fully solved, which the three-handed game can't.

Even HU I'd actually rather play the bot than Ivey. I'd lose to both, but I'd lose faster to Ivey, as he can take advantage of my leaks. Ivey wouldn't lose to me, but he'd lose to the GT bot as well. This isn't actually strange at all if know what you are talking about.

[Rudbaeck]

[ QUOTE ]
One way to think about it is that the bot plays perfect defense. It's not going to lose. It will however only win what your ineptitude gives it.

[/ QUOTE ]

This metaphor was no good. It causes confusion.

I'll instead state what actually happens. The bot knows the perfect play for both parties. It assumes you will make the perfect play every time. None of us will make perfect plays every time, and some will wander pretty far from perfect play. Whenever a human makes mistakes the bot won't charge extra for that.

Lets take an example from limit hold'em. You are UTG with A9s, the by far worst player at the table is David Sklansky. You fold. This is the play the bot would make every time.

Now, even at a table with 9 horrible, horrible calling stations the bot would fold, when the correct play actually is to raise. It's unable to account for their weakness.

It can't take advantage of the specific weakness of your style. This will soemtimes lower the EV of a play, as it assumes and plans for a perfect response. But no bad play the opponent makes can turn a move that would be +EV against a perfect player into a -EV move. (HU any deviation on the part of the opponent will make the play more +EV actually.)


Ps. Please don't turn this into a debate on whether it's right to fold A9s against 9 clones of David Sklansky. It's just an example pulled of my arse. Ds.

[SkyRocker]

[ QUOTE ]
But I am just as certain that no living human can beat the best bots HU.

[/ QUOTE ]

I wouldn't be so sure about that. Poki's Poker Academy has the best comercial HU limit bot in the world and there exist many players who can beat that. But then there also exist some non-comercial bots people have made. But I have never heard anyone claim that their bot would beat any human.

[Snoogins47]

"
I love to hear all of this talk about the perfect Bot and how it could beat any Hold-Em player by playing defensively.

For those of you who are on the border - don't believe it.

First of all Hold-em is not only a game of skill, but also a game of luck. If what these poeple are claiming is true, that there is a perfect bot that will always win, and that players like Phil Ivey nearly mimic the bot, then Phil should ALMOST NEVER lose!"

When used in a discussion like this, the only reasonable definition of "unbeatable" or "won't lose" is that there will be no edge against this bot, regardless of how you play. That's the most skill can do for anybody in a game with an element of randomness.

This, of course, all amounts to how close HU hold'em is to being solved. I don't know, but from what little I have seen/read, it seems that a HU game with no betting after the flop could probably be solved, or very close to it. The game tree branches out to an astounding number of branches if we make any considerations after the flop. I can't comment further, because I don't know how developed bots are in.

Another way to think about how a GT approach would work, but wouldn't be perfect, is to think about a simple bluffing play.

An optimum bluffing frequency against one player would be such that regardless of the frequency that your opponent calls you, your results do not change. It's a strategy that is unexploitable.

But would this strategy win the most money against a player who makes fundamental errors? Like say, an opponent that will call every single time? I think this example shows both the power of an unexploitable strategy, and the very, very prominent shortcomings.

[SkyRocker]

The largest poker game completely solved is Rhode Island Hold'em which has about O(10^9) leaf nodes in the game tree. Hold'em has about O(10^18) leaf nodes so we are not even close to solving this. But it is possible to simplify hold'em and solve this new game with less nodes. This is how it is possible to create "pseudo optimal" poker bots.

[Dromar]

[ QUOTE ]
[ QUOTE ]
[ QUOTE ]
Anything that can easily be taught requires less talent than something where you need to be intuitive.

[/ QUOTE ]
So rock paper scissors requires more "talent" than chess?

[/ QUOTE ]

Are you saying that chess can be easily taught?

[/ QUOTE ]

I think chess can be easily taught. Whether the person is GOOD at chess is another matter. I don't think there's much difference between "skill" and "talent" though. Maybe, as you play chess you get better at it, defining it as a skill that you acquire. I think that in the same way, people acquire poker skills as they play and get better. The difference is that chess has no luck factor, whereas rock/paper/scissors does (IMO, although I'm sure that's debatable). If a luck factor makes proficiency be referred to as "talent," then all types of poker proficiency are made up of both skill and talent. Although, this assumes "talent" to be nothing more than luck.

On a side note, I don't consider rock/paper/scissors to have any skill factor. I'm assuming that a game of rpc has a 1/3 probability of player A winning, a 1/3 probability of player B winning, and a 1/3 probability of a tie.

So, from what I said above, it seems the definition between the words "skill" and "talent" are blurred, but I believe the OP meant skill to be explicit things (such as hand selection, reading the board, betting correctly, etc.) and talent to be implicit things (such as bluffing, reading, making moves, etc.)

Which game requires most talent then?
My guess would have to be tournament style NLH, not necessarily because it's the most talent-intensive at a mastery level, but I think because the learning curve for explicit NLH skills such as hand selection is higher for most players than that of other games. In other words, I think people know more in general about how to play NLH because of it's overwhelming popularity in books and TV, and as a result, more talent is required to win at these games. Of course, a case could be made about TV making people WORSE at NLH, but I think that's probably a minor effect.

[Atropos]

[ QUOTE ]

The perfect defensive bot would probably lose to the majority of the players on this forum, amateurs included. If the bot holds KQ and there is an ace on the board, the bot will ALWAYS fold to a big bet. Why? Because it is the right thing to do!


[/ QUOTE ]

You would lose soooo horribly against PokiBot. Why should a bot fold KQ on an A-high board ALWAYS? Because he is "scared" of the ace? Bots dont play that way. In a heads-up game he will analyze how many hands you are playing preflop, raising preflop, betting the flop with etc... Obviously your range in heads-up games is so big, that it would be very correct to call the flop with KQ, and the bot will do it. If you would have ever played against PokerAcademy, you would know that sparbot often calls flop bets with K-high or Q-high even... defensive doesnt mean "only play the nuts".

[Snoogins47]

To elaborate: defensive in this context means "unexploitable." We usually think of defensive as a tight, risk averse style. That isn't what we're talking about here.

A decently illustrative example, while it is far from optimal play, is to imagine you're horribly out-classed in a heads-up NLHE battle, with medium sized stacks. Like, awfully out-classed. You don't even belong in the same country as this opponent, let alone the same ballpark. Let's say you realize this, and have decided your goal is to do everything in your power so that your opponent can't bitchslap you as he rightfully should. You're trying to be defensive, and to prevent him from exploiting the overwhelming skill difference. What do you do? You push all in, before the flop, a LOT of the time.

The "defensive" play outlined above is much more aggressive than a "normal" strategy.