I have become tired of putting down all the mundane mistakes GIB makes, because it seems like enough isn't being done. In an express tourney, why doesn't GIB just return the lead? People DO make mistakes.
http://tinyurl.com/6yv5rde
Page 1 of 1
Singleton
#1
Posted 2011-June-24, 14:29
"It's not enough to win the tricks that belong to you. Try also for some that belong to the opponents."
"Learn from the mistakes of others. You won't live long enough to make them all yourself."
"One advantage of bad bidding is that you get practice at playing atrocious contracts."
-Alfred Sheinwold
"Learn from the mistakes of others. You won't live long enough to make them all yourself."
"One advantage of bad bidding is that you get practice at playing atrocious contracts."
-Alfred Sheinwold
#2
Posted 2011-June-26, 20:54
Probably because it expected South to have more spades, for its jump to slam with no support from partner, so it 's not likely to have so many clubs. So the lead looks more like a doubleton than singleton.
#4
Posted 2011-June-26, 22:08
When GIB sees a short suit in dummy, it often leads trumps to prevent declarer from ruffing.
GIB isn't programmed with the logic "If there's only one layout where you can set the contract, assume that layout." It deals out a bunch of hands that are consistent with the auction, and performs a double-dummy analysis of each of them to find the plays that results in the most tricks. Then it chooses the play that shows up the most. This algorithm is also why it doesn't take safety plays when declaring.
This obviously isn't the way to play optimally, but this is how GIB is written. It would presumably take a major redesign to fix it.
GIB isn't programmed with the logic "If there's only one layout where you can set the contract, assume that layout." It deals out a bunch of hands that are consistent with the auction, and performs a double-dummy analysis of each of them to find the plays that results in the most tricks. Then it chooses the play that shows up the most. This algorithm is also why it doesn't take safety plays when declaring.
This obviously isn't the way to play optimally, but this is how GIB is written. It would presumably take a major redesign to fix it.
#5
Posted 2011-June-28, 08:41
Maybe not, it would slow the bots down, but how about after x simulations, if there are 2 plays or calls which have identical results, run another x simulations and repeat again if necessary. If after 3x simulations there is still no difference, then allow it to pick one at random.
Wayne Somerville
Page 1 of 1

Help
