Author |
Topic: Technical issues about self-modifying bots (Read 3745 times) |
|
99of9
Forum Guru
Gnobby's creator (player #314)
Gender:
Posts: 1413
|
|
Re: Technical issues about self-modifying bots
« Reply #15 on: Jan 24th, 2006, 10:30pm » |
Quote Modify
|
Ryan the objective of these books so far is a far less worthy aim than you are describing. Gnobby's aim is direct plagiarism from winning play of identical positions. Bomb's aim seems to be to drag it toward previous winning lines, and away from previous losing lines (especially those it played itself). I have a question for you. When you were a 1600 rated player, would you have been willing to spend 0.5 seconds of your thinking time to find out whether a top rated player had ever played that exact position, and if so, be told what they had played? Surely! Of course you are right that sometimes that exact position came up with a cat switched for a dog. You will not be alerted to that fact. But the identicals are still useful in themselves (if only for the first 3-5 moves of a game).
|
|
IP Logged |
|
|
|
Ryan_Cable
Forum Guru
Arimaa player #951
Gender:
Posts: 138
|
|
Re: Technical issues about self-modifying bots
« Reply #16 on: Jan 24th, 2006, 11:19pm » |
Quote Modify
|
on Jan 24th, 2006, 10:30pm, 99of9 wrote:I have a question for you. When you were a 1600 rated player, would you have been willing to spend 0.5 seconds of your thinking time to find out whether a top rated player had ever played that exact position, and if so, be told what they had played? Surely! |
| Absolutely, I would love to be able to do that right now!
|
|
IP Logged |
|
|
|
fotland
Forum Guru
Arimaa player #211
Gender:
Posts: 216
|
|
Re: Technical issues about self-modifying bots
« Reply #17 on: Jan 24th, 2006, 11:36pm » |
Quote Modify
|
Bomb has had the book at least since 2005, and probably since 2004. I used the same code in my chess program and go program. Bomb just records its own games, since the game database wasn't public at that time. The go and chess programs also include many strong games. The main purpose is to prevent it from losing the same way twice with the same move sequence. This is important for tournaments where it might play the same program multiple times, or play a program that trained against it. It also makes it more interesting for human opponents, who tend to repeat winning openings.
|
|
IP Logged |
|
|
|
Ryan_Cable
Forum Guru
Arimaa player #951
Gender:
Posts: 138
|
|
Re: Technical issues about self-modifying bots
« Reply #18 on: Jan 25th, 2006, 1:13am » |
Quote Modify
|
on Jan 24th, 2006, 11:36pm, fotland wrote:The main purpose is to prevent it from losing the same way twice with the same move sequence. This is important for tournaments where it might play the same program multiple times, or play a program that trained against it. It also makes it more interesting for human opponents, who tend to repeat winning openings. |
| Yes, I think this is a very good idea. Especially, if it allows you to remove or at least reduce the opening randomization. And reducing the randomness feeds back helping the book find more matches. However, I was able to bash bot_Bomb2005Fast, bot_Bomb2005Blitz, bot_Bomb2005CC, etc. using the same Bait and Tackle opening ~50 times, with the first 9 or 10 moves being identical on my part. Also, Gnobby was able to copy a couple games identically against one of the Bomb variants, so I think the book must be turned off or broken for these bots.
|
|
IP Logged |
|
|
|
|