They are not for eating
The Genetic algorithm is working well. It can come up with a buy strategy and test different players with that strategy to see which one is more fit to survive.
*I found out recently that a flaw in the code was making it actually search for weaker players after a while instead of stringer players, so don't take this dominion strategy to be that good :)
after about 5 hours (100 generations into the simulation) it came up with the following strategy:
Name | Priority | Min | Max |
---|---|---|---|
Province | 9.46 | 2 | 9 |
Witch | 6.44 | 2 | 5 |
Silver | 6.55 | 6 | ∞ |
Duchy | 5.54 | ∞ | — |
Gold | 4.99 | 4 | 8 |
Militia | 4.97 | 1 | 4 |
Library | 3.48 | 5 | — |
Village | 2.23 | 6 | ∞ |
Copper | 0.03 | 4 | — |
Smithy | -1.80 | 7 | — |
Cellar | -3.27 | 6 | 7 |
Laboratory | -3.32 | 5 | — |
Curse | -4.58 | 8 | — |
Chapel | -5.35 | 0 | 5 |
Estate | -5.92 | 5 | — |
Moneylender | -8.88 | 2 | 4 |
Once a player algorithm purchases up to the minimum desired value, the priority of the card is halved, so once the player has 2 witches he will start buying duchies. Also, if the player has more gold than is required to purchase something, the value of that thing is reduced (if you are 1 over, the value is halved, and if you are 2 over it is cut by 2/3 etc), so with 6 gold this player will buy golds and not silvers, and with 8 he will always buy provinces
I am a little surprised that it did not pick chapel with undeniable witches on the board, but perhaps it has some faulty trashing reasoning.
*I found out recently that a flaw in the code was making it actually search for weaker players after a while instead of stringer players, so don't take this dominion strategy to be that good :)
My next step is to use this genetic algorithm to train a neural network to set good buy priorities for the game given a set of ten kingdom cards. I have never written a neural network before, so hopefully it will go well. I have also never trained a neural network, so perhaps that will go well, but it is hard to say.
Andrea is learning things all the time. She loves to help me fix things. Alison has told me that most of the time when she goes out of a walk she has to stop to 'fix' my bicycle with pine-cones and sticks.
She has also decided that maple helicopters are tasty food products. Some searching online seems to indicate that they are at least not poisonous, so I'm not too worried. It is not like they are here all year, or like she is eating a lot of them.
We recently moved some of our surviving grape vines to new locations in order to spread them apart. Hopefully they survive the transplanting process.
I am thinking of having a game day this Saturday. If anyone reading this is interested, you are invited. (Tell me if you are coming, though).
I have been thinking about 'Race for the Galaxy' a lot recently since I found a good AI for it online. Perhaps we would play that.
Andrea is learning things all the time. She loves to help me fix things. Alison has told me that most of the time when she goes out of a walk she has to stop to 'fix' my bicycle with pine-cones and sticks.
She has also decided that maple helicopters are tasty food products. Some searching online seems to indicate that they are at least not poisonous, so I'm not too worried. It is not like they are here all year, or like she is eating a lot of them.
We recently moved some of our surviving grape vines to new locations in order to spread them apart. Hopefully they survive the transplanting process.
I am thinking of having a game day this Saturday. If anyone reading this is interested, you are invited. (Tell me if you are coming, though).
I have been thinking about 'Race for the Galaxy' a lot recently since I found a good AI for it online. Perhaps we would play that.
This comment has been removed by the author.
ReplyDelete