Inside PokerSnowie's brain reveals the work of the Snowie AI Team. It explores first hand how the brain of PokerSnowie evolves and learns advanced strategic concepts, on its own.
PokerSnowie's ultimate aim is to produce the perfectly balanced game, find the ultimate un-exploitable equilibrium for all No Limit Hold'em configurations. Join us on this fascinating journey, which is just starting, into the future of poker.
PokerSnowie's pre-flop strategy: Opening the pot
The easiest decision in Poker is about opening the pot. Nobody has shown aggression yet, the pot size is as small as it can be, so no costly error can be made.
On the other hand, you better get this part right! There are several reasons why opening the pot correctly is so important:
1) This situation happens all the time. In a big portion of your hands you will face this decision. Even if you only make small errors, they'll add up quickly.
2) If your opening range is out of balance, your opponents may notice that and play back exploitatively. Your errors are small against balanced players that don't exploit them. However, once your opponents start exploiting your weakness, your errors potentially become huge.
The first concept when opening the pot is positional awareness. Intuitively it is clear, that the more players still left to act behind you, the more careful you have to select your hands to open the pot and the tighter you have to play.
Let's have a look at PokerSnowie's opening percentage in a 6-max game ($5/$10, 100BB stacks):
1st player: 17%
2nd player: 23%
Cutoff player: 26%
Button player: 42%
PokerSnowie shows a high positional awareness, playing looser closer to the button. It also makes quite a big difference between the cutoff position and the button position. Only on the button are you sure to have a positional advantage for the rest of the hand, which is reason enough to loosen up considerably in that position.
If you compare PokerSnowie's opening percentages to online professionals, they seem quite tight. Why is there a difference? Should online pros not play close to PokerSnowie's numbers? There are two differences between how the game was learned by poker pros and PokerSnowie:
- Poker pros are the favourites in the game, they are by definition more skilful than their average opponents. That's why they can afford to play a bit looser. PokerSnowie, however, has trained in a tough environment, its opponents were always as skilful as itself. It has to stick to playing optimally.
- In the typical online games, the players in the blinds play too tight. They often don't realize the good pot odds they get in the blinds and fold too much. In return that makes it more attractive to steal the blinds.
Constructing the range
The next question is, how should you construct the opening range? Which hands are selected for raising? As an example, let's take the cutoff player. He should only raise or fold. Calling would give very good pot odds for the small blind player and a free ride to the big blind player. Better is to raise and give yourself a chance to win the pot right away.
If we know that in cutoff position we should raise 26% of the time, how do we decide which hands to use?
The first step is very easy. The strongest hands will be put into the raising range. All value hands like medium to high pocket pairs, big cards like AK, AQ and KQ and high suited connectors will clearly be raised.
The first part of the second step is also easy. To fill up the range with bluffs, it is also clear that from the remaining hands, the top hands should be used. It is clearly better to raise 87o and fold 32o than to raise 32o and fold 87o. But what are the top hands?
Which hands to use as bluffs?
The top hands are not necessarily the hands with the highest showdown win rate. Much more important is how well the hands play after the flop. But even if we knew this, we'd have to think one step further: what happens to our range after a re-raise? If, for example, everybody folds to our opening raise, but the big blind player re-raises pot, we have to come up with a correct folding range. If we have to fold the majority of our hands, we have constructed the opening range wrong, and the big blind player can profitably raise back with weak hands.
As a concrete example, let's compare using A8o and 65s for opening the pot in cutoff position. At first glance A8o seems to be the superior hand. A8 can make top pair, a pair of aces, on the flop or a pair of 8s with top kicker. Even without hitting a pair, A8 may hold up and win a showdown on the river with high card.
But if we think of the points just mentioned above, we also notice that 65s has a much better potential to extract money from the opponents. With A8 you don't know if your pair of aces is the best hand, because of your bad kicker. If you make a pair of 8s, most likely there are over-cards on the flop and you cannot exert pressure either. If, however, you hit a flush or a straight with your 65s, you may be able to build up a nice big pot for you. Even if you only hit a draw, you can play your hand aggressively and win the pot without ever making your hand.
Additionally, what happens if the big blind player re-reraises? According to PokerSnowie your A8o has an EV for call of -1.53 big blinds, so you clearly have to fold your hand. With 65s, however, you have a positive EV for call of +0.30 big blinds. The 65s is not only better in terms of EV but it also helps construct a broad calling range.
Due to the nature of the learning algorithm, PokerSnowie was forced to construct the most optimal ranges. They should give the highest expected value possible and provide solid resistance against re-raises at the same time. In the example above, PokerSnowie opens with 65s but not with A8o. PokerSnowie has gained a huge amount of experience and knows exactly which hands play well post-flop. At the same time, during training it had to defend against exploiting agents, so the ranges are constructed in such a way that its ranges even after a re-raise are also nicely balanced.
Inside PokerSnowie's brain
PokerSnowie's ultimate aim is to find the ultimate un-exploitable equilibrium for all No Limit Hold'em configurations. Join us on this fascinating journey, which is just starting, into the future of poker.MORE LESSONS
Oliver Heuler (not verified)
Sun, 01/26/2014 - 10:21
AJ (not verified)
Wed, 02/05/2014 - 02:04