[Andy_L]

From #3692 Robot Duplicate - Matchpoints 2011-08-11. What can N GIB possibly be thinking? 6NT on 31HCP has an 95% line of play: At trick #2 play a ♦2 to the ♦A and then ♦8 back to the N hand... with the intent of continuing ♦'s to knock out the ♦K. Then Tricks are: ♠(3) - ♥(1) - ♦(4) - ♣(4) = 12 Tricks. At Trick #3 if N learns the ♦'s are 5-1, N can then still make if the ♠'s are 3-2: ♠(6) - ♥(1) - ♦(1) - ♣(4) = 12 Tricks. GIB N played this hand at 6NT eight times and 7NT once. GIB N made 6NT once and went down 3 at 6NT seven times and down 4 at 7NT once. Brings up another GIB question I have: When there is no E-W bidding and different N-S bidding (yet N-S arriving at the same contract) why does GIB play the hand so differently? In this case are the minute inferences drawn from E-W passes during different N-S auctions enough to abandon a 95% line of play? Playing ♦'s at trick #2 yields 84% success rate when ♦'s are 4-2 or 3-3... for the 16% when ♦'s are 5-1 or 6-0 there is still a 68% success rate play switching to ♠'s... still 12 tricks as long as ♠'s are 3-2. 84% + (16% * 68%) = 95% If I've done the apriori percentages right... how can the GIB not take a 95% line of play in a GREAT contract of 6NT when on 31HCP?

[mtvesuvius]

---

Seems to me this is a backfiring of the standard GIB greed -- It was trying to make 7 when the diamond hook works. I have no idea why it didn't play on diamonds immediately, as that offers the same opportunity but will still make when it loses. Of course a duck by West would give North a decision whether to play for spades to come in or settle for 12 tricks, but it still is much better than the line taken.

GIB pays no attention to whether it is in a field contract or not. If it had a 98% line of play in a miraculous 21 HCP slam, it might not take it if it deems another line more likely to take a greater number of tricks on average. This seems like a major weakness for GIB, but it is nearly impossible to come up with a reasonable fix.