Help
I don't know how playing a small diamond to the 8 can be right unless you knew the JT was onside.... Mind you, you'd think it would at least get the trumps right afterwards!
Page 1 of 1
Sometimes I wonder if GIB can actually see all 4 hands
#1
- - - A AKQJT9876543
-
- Group: Advanced Members
- Posts: 2,614
- Joined: 2007-October-02
- Gender:Male
Posted 2024-August-07, 17:35
I don't know how playing a small diamond to the 8 can be right unless you knew the JT was onside.... Mind you, you'd think it would at least get the trumps right afterwards!
Wayne Somerville
#2
- Group: Advanced Members
- Posts: 4,055
- Joined: 2014-March-15
- Gender:Male
Posted 2024-August-08, 01:19
I assume basic robot and MPs?
Probably the most useful point here is that GIB's system strictly requires a lead of K with AKx+, and A from AK doubleton. So after the first two tricks, it's assuming clubs split 7-2 in all simulations, which affects the odds quite a bit. At the same time, it still makes the play dangerous, as if it loses a club back is likely getting overruffed..
Getting the old version of GIB to run a huge sample, it's saying that on 23.5% of deals it doesn't matter which diamond it plays, 50% of the time it has to play the ace, and 26.5% of the time it has to play the 8.
That sounds like a big majority, but remember GIB runs very few simulations.. a sample of 100 hands with those odds would conclude ace is best 99.7% of the time. But a sample of 30 hands would conclude 8 is correct 6.7% of the time, which is often enough. I wouldn't be surprised if basic GIB runs even less than that.
Play your clubs in the other order, and it'll play the ace 99.995% of the time even on a 30 hand sample.
Probably the most useful point here is that GIB's system strictly requires a lead of K with AKx+, and A from AK doubleton. So after the first two tricks, it's assuming clubs split 7-2 in all simulations, which affects the odds quite a bit. At the same time, it still makes the play dangerous, as if it loses a club back is likely getting overruffed..
Getting the old version of GIB to run a huge sample, it's saying that on 23.5% of deals it doesn't matter which diamond it plays, 50% of the time it has to play the ace, and 26.5% of the time it has to play the 8.
That sounds like a big majority, but remember GIB runs very few simulations.. a sample of 100 hands with those odds would conclude ace is best 99.7% of the time. But a sample of 30 hands would conclude 8 is correct 6.7% of the time, which is often enough. I wouldn't be surprised if basic GIB runs even less than that.
Play your clubs in the other order, and it'll play the ace 99.995% of the time even on a 30 hand sample.
#3
- - - A AKQJT9876543
-
- Group: Advanced Members
- Posts: 2,614
- Joined: 2007-October-02
- Gender:Male
Posted 2024-August-08, 14:17
smerriman, on 2024-August-08, 01:19, said:
I assume basic robot and MPs?
Probably the most useful point here is that GIB's system strictly requires a lead of K with AKx+, and A from AK doubleton. So after the first two tricks, it's assuming clubs split 7-2 in all simulations, which affects the odds quite a bit. At the same time, it still makes the play dangerous, as if it loses a club back is likely getting overruffed..
Getting the old version of GIB to run a huge sample, it's saying that on 23.5% of deals it doesn't matter which diamond it plays, 50% of the time it has to play the ace, and 26.5% of the time it has to play the 8.
That sounds like a big majority, but remember GIB runs very few simulations.. a sample of 100 hands with those odds would conclude ace is best 99.7% of the time. But a sample of 30 hands would conclude 8 is correct 6.7% of the time, which is often enough. I wouldn't be surprised if basic GIB runs even less than that.
Play your clubs in the other order, and it'll play the ace 99.995% of the time even on a 30 hand sample.
Probably the most useful point here is that GIB's system strictly requires a lead of K with AKx+, and A from AK doubleton. So after the first two tricks, it's assuming clubs split 7-2 in all simulations, which affects the odds quite a bit. At the same time, it still makes the play dangerous, as if it loses a club back is likely getting overruffed..
Getting the old version of GIB to run a huge sample, it's saying that on 23.5% of deals it doesn't matter which diamond it plays, 50% of the time it has to play the ace, and 26.5% of the time it has to play the 8.
That sounds like a big majority, but remember GIB runs very few simulations.. a sample of 100 hands with those odds would conclude ace is best 99.7% of the time. But a sample of 30 hands would conclude 8 is correct 6.7% of the time, which is often enough. I wouldn't be surprised if basic GIB runs even less than that.
Play your clubs in the other order, and it'll play the ace 99.995% of the time even on a 30 hand sample.
It was in the Diana and friends tourney, which I think uses the advanced bots.
Wayne Somerville
#4
- Group: Advanced Members
- Posts: 4,055
- Joined: 2014-March-15
- Gender:Male
Posted 2024-August-08, 15:48
Ah, yep. Well, barmar once said the advanced robots simulate "a few dozen" hands, so it's probably well within the range of possibility - even a 50 card simulation would still have it playing the 8 about 3% of the time. Or it could simply be another case of it playing a random card for no reason at all, which it does from time to time.
Page 1 of 1