| |
BGonline.org Forums
Snowie vs XG in backgames
Posted By: phil simborg In Response To: Snowie vs XG in backgames (higonefive)
Date: Friday, 24 December 2010, at 2:52 a.m.
There is another methodology to test the strength of the bot in back games compared to humans.
Set up a complicated back game position.
Invite WC players to play the position out against the bot. I would have each player selected play only one side, say 20 times, but they will not be able to see their errors during the game or between games (so they cannot learn from the bot).
Have another group of WC players play the other side against the bot, again, not seeing their errors during or between games.
See who wins more points from each side, the players or the bot.
e.g. Let the No. 1, 3, 5, and 7 Giants play Red 10 games against the bot. Let the No. & 8 Giants play White 10 games against the bot. Compare the bot's record against the players' record. The reason I wouldn't want any one player to play more than 10 games is that he might find a flaw in the bot's play on a given type of play or situation and try to exploit it, and that would only prove that humans have a better ability to adjust to opponent's flaws, not necessarily that they play better when they are not aware of a given flaw.
Naturally, this type of competition could be done in a more practical way, using a few WC players in the study who are willing to play under the above conditions and honestly report their results.
By the way, I'll bet on the bot.
(I did a form of the above study myself a few years ago when I was in a 2-3-4-5 back game will all 15 checkers back and I couldn't win from either side anywhere near as many games as the bot said I should win. So I started studying how the bot played both sides, began playing more that way myself, and I began winning much more from either side. This, of course, only proved that the bot played better than I did, which doesn't prove the bot is better than a human given my questionable origin.)
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.