GIB mistakenly discards Ace in 3-card ending
#1
Posted 2013-June-14, 12:03
http://tinyurl.com/mnrha8l
In the 3-card ending, the bot playing East discards the A♣ instead of the 10♣ on the play to the 11th trick. Any possible logic for doing this, or is it a bug? Declarer has only one unknown card at the time of the discard, and the discard of the A♣ can only ever hurt (as it did here). If declarer had a heart instead of another club, then the discard makes no difference, but that is inconsistent with the bidding (declarer denied a four card major). I suppose it could be that the initial underbid by declarer lead the bot to assume that West had the K♣ in which case the discard makes no difference. But the logic here should be to go with the play that dominates (as good or better in all circumstances).
I must say the tournament bots in general really do play (and in particular defend) a lot better than the basic bots!
Peter
#2
Posted 2013-June-14, 12:15
Basically, you confused it.
#3
Posted 2013-June-14, 16:42
#4
Posted 2013-June-14, 21:55
#5
Posted 2013-June-14, 21:56
manudude03, on 2013-June-14, 16:42, said:
fixed, didn't read play thoroughly
#6
Posted 2013-June-15, 12:24
Antrax, on 2013-June-14, 21:55, said:
So an easy fix is to pitch the lower card when it wants to pitch "whatever."
This particular issue doesn't seem to come up much with the advanced bots, but it comes up a lot with the basic bots.
#7
Posted 2013-June-15, 21:09
More to the point, is this really the problem with GIB, that psyching 1NT with a 2NT opener makes it give up an overtrick? Assuming you have limited time to work on it, wouldn't you rather fix something else?
#8
Posted 2013-June-16, 16:25
Antrax, on 2013-June-15, 21:09, said:
More to the point, is this really the problem with GIB, that psyching 1NT with a 2NT opener makes it give up an overtrick? Assuming you have limited time to work on it, wouldn't you rather fix something else?
It is an easy fix - if the simulations say that GIB is indifferent between two cards of the same suit, then discard the lower one.
I grant this particular example with the !NT psyche (not by me I would add!) is in and of itself not that important, but with the basic bots this sort of error happens much more, even without a psychic bid. I assume the smaller number of simulations lead to more cases where it doesn't seem to matter what discard to make, even when in fact it does.
#9
Posted 2013-June-17, 15:47
That said, if GIB correctly decides that there is no difference between two discards, then there is clearly no reason to prefer the lower of two cards.
#10
Posted 2013-June-18, 03:58
FM75, on 2013-June-17, 15:47, said:
[nitpick]Only 6 possible hands, you're double-counting.[/nitpick]