10/26/2005

10-26-05 - 4

10-26-05

Dan and I have been playing a lot of gin lately. I now see gin as a lot like poker. Yes, there's a lot of randomness, but with correct play you can improve your chance of winning. Many of the decisions in gin are "no brainers", but in a typical match you encounter a few very difficult decisions (assuming that you find the basic decisions not difficult). I've never read about proper gin strategy, I'm sure it's an exactly solved game and you could learn to play perfectly, but I'm having fun trying to figure it out myself.

The most basic ideas are these : 1) you want to use your 10 cards in a 3-3-4 split if possible, because this is far more likely than any other way of using them up. 2) you want to at all times maximize your outs of improving. 3) for a set of 3 to become a set of 4, it's far better to have runs in sequence (2 outs) than sets of the same value (1 out).

The more general strategy is to maximize your outs (the number of cards that help you), and to try to maximize the benefit of each action. Picking up a random card has some average benefit to your hand (which partly depends on what you have in your hand and what's been discarded). Think of it this way - every 10-card hand we give some value based on how good it is (fully matched up is maximum value). Your goal is to take the biggest value step with each move. The value of picking from the deck is simply the average of the value delta over all the unknown cards. What you do is compare this to the value of picking up the face up card in the discard pile.

The cases where this become difficult are in marginal cases. For example, I'll never pick up from the face up pile just to make a 2-set. That is, if you have 7c and the discard pile shows 7s, there's no need to pick that, picking from the deck will have a better average value. But, what if the discard pile shows 7s and you have the 7c and also 8s. Now picking that card would give you a 778, which is one of those nice things to have because it has 4 outs to become a triple. That's a tough decision that I'm not sure about and is partly dependent on the situation.

Obviously you must be aware of trivial things like avoiding straights with the ace because you lose an out that way, you have to watch what your opponent is collecting and try to guess what cards they have. You need to avoid having cards that take each others outs. For example if you have a TTT and a JQK , you only have 2 outs to quad those trips, rather than the 3 outs you would have with 777JQK or something like that where they don't run into each other.

Another thing people often get wrong is to think about their trips and quads as being "locked up", and also to think that all "junk" cards are equal. Any set can be broken up if it makes sense. For example, I'll usually have things like the 778. Now when you get a 7, you have a 7778, and you think "that 8 is junk" and you should drop it. Not so. If you have some other card which is even less useful, you should drop it instead. The 8 is not very useful, but what if you get the 9? Now you have a 77789 and you might discard a 7, because the 789 is better than the 777 (more outs to quad up). Even then if you have some other card which is even more useless, you might just keep that 77789 going, because it gives you 2 outs to make two trips (the last 7 or a T). Another interesting situation occurs if you get something like the 8c when you have the 7s7c7d8d9d. If you could keep that 8 you would have 6 outs to make two trips, which is mighty nice.

On some poker show once I heard that Stu Ungar was considered by many the best gin player alive (in his day, when not too high or hung over). At the time I wasn't thinking about gin, so I thought "what the fuck, gin is just luck, how can you be the best? it's like being the best War player". Now I know different; in addition to strategy, you have the aspect of trying to disguise your hand with your discards, reading your opponent's hand, and also remembering every card in the discard pile so you know the dead outs.

No comments:

old rants