Author |
Topic: Winning with random moves (Read 2530 times) |
|
leon_messerschmidt
Forum Senior Member
Arimaa player #6344
Gender:
Posts: 26
|
|
Winning with random moves
« on: Feb 27th, 2014, 2:19am » |
Quote Modify
|
I'm dabbling with the idea of writing a bot. To test the mechanics of my move-generator I wrote a bot that selects random moves and matched it against the random player in AEI. Curiously when I pitch the two together my bot wins roughly 70% of the time. When two instances of the AEI random player face off, they win about 50/50. Likewise for two instances of my random bot. The only reason I can think of is that my bot culls duplicate moves and that the AEI random player selects from all available moves. But it doesn't sound right does it?
|
|
IP Logged |
|
|
|
chessandgo
Forum Guru
Arimaa player #1889
Gender:
Posts: 1244
|
|
Re: Winning with random moves
« Reply #1 on: Feb 27th, 2014, 4:24am » |
Quote Modify
|
If "culls duplicate moves" means that you select a post-move position uniformly at random instead of a uniformly random move, you give more weight to captures and fast rabbit advances (which mostly can only be achieved with one move, while useless moves replicate more easily, especially the two-effective plus 2 reverse steps)
|
|
IP Logged |
|
|
|
clyring
Forum Guru
Arimaa player #6218
Gender:
Posts: 362
|
|
Re: Winning with random moves
« Reply #2 on: Feb 27th, 2014, 7:01am » |
Quote Modify
|
A random-mover is indeed stronger than a random-stepper, but not to the point of having a 70% winrate. If I had to guess, the other difference is that only one of the setup protocols is random.
|
|
IP Logged |
I administer the Endless Endgame Event (EEE). Players welcome!
|
|
|
Ail
Forum Guru
Rabbits can't push Rabbits!
Gender:
Posts: 52
|
|
Re: Winning with random moves
« Reply #3 on: Feb 27th, 2014, 7:35am » |
Quote Modify
|
I'm rather curious how many moves on avarage a match of 2 random-movers takes. I also started coding some arimaa-stuff. Right now I'm still at my move-checker. It would have been easy if you could only move your own pieces but checking the validity of pushing/pulling is not a simple task. ^^ Is there probably any open-source bot where one could copy&paste the basics so one doesn't have to reinvent everything from scratch?
|
|
IP Logged |
|
|
|
petermck
Forum Junior Member
Arimaa player #9440
Gender:
Posts: 6
|
|
Re: Winning with random moves
« Reply #4 on: Feb 27th, 2014, 10:41am » |
Quote Modify
|
Looks like there's a bunch of source code here that you could probably use: http://arimaa.com/arimaa/download/
|
|
IP Logged |
|
|
|
leon_messerschmidt
Forum Senior Member
Arimaa player #6344
Gender:
Posts: 26
|
|
Re: Winning with random moves
« Reply #5 on: Feb 27th, 2014, 1:58pm » |
Quote Modify
|
@chessandgo I generate all possible positions (4 steps deep) and use a transposition table to remove duplicates. And then randomly select a position from that set. Thank you for your insights. That makes a heap of sense.
|
|
IP Logged |
|
|
|
clyring
Forum Guru
Arimaa player #6218
Gender:
Posts: 362
|
|
Re: Winning with random moves
« Reply #6 on: Feb 27th, 2014, 10:11pm » |
Quote Modify
|
Apparently there was already some research into this in 2003. See also this thread.
|
|
IP Logged |
I administer the Endless Endgame Event (EEE). Players welcome!
|
|
|
Janzert
Forum Guru
Arimaa player #247
Gender:
Posts: 1016
|
|
Re: Winning with random moves
« Reply #7 on: Feb 28th, 2014, 10:07am » |
Quote Modify
|
on Feb 27th, 2014, 7:01am, clyring wrote:If I had to guess, the other difference is that only one of the setup protocols is random. |
| The "simple_engine.py" included with AEI is a constant setup, random stepper. So like clyring I would guess that your bot is most likely either using random setups or possibly a constant setup that favors random movement more (maybe some or all rabbits forward?). Janzert
|
|
IP Logged |
|
|
|
|