[iandayre]

http://tinyurl.com/zq6ju5b

At all tables the bidding began, 1H - 1S - P - P - 2H. Do you continue? Vul. vs Not, ACBL Robot Duplicate MP's.

As it happens your answer will not affect the final contract, but it will certainly affect GIB's declarer play, as I'll show after a few replies. Answer the bidding question before watching the play for a fair problem.

[iandayre]

Well, bidding aside, the point of the hand is the vagaries of GIB's declarer play. About half the field subsided after the 1S overcall. GIB won the SA and bizarrely cashed the HA, throwing the K under it! This seemingly no-win play created a trump loser for GIB. It then took the D finesse but was unable to return to its hand to repeat it. We took our trump trick and 4 black suit winners for -2. Those who bid again over 2H with the S hand, whether with a double, 2C bid or whatever, also led the SK. GIB won, took the D finesse, cashed 4 rounds of H beginning with the K, and repeated the D finesse for 2 overtricks. A normal and successful line of play. Why GIB chose the actual one when the only competition was a 1S overcall is anyone's guess.

[johnu]

No matter what the auction was, crashing the ♥K is a nullo play that can never gain. Different actions by N-S shouldn't make any difference to the play in 4♥. In the past, I've seen the choice of a minor suit opening make a huge difference in the play (when it shouldn't have affected the outcome), or even the choice of a (immaterial) spot card lead make a huge difference in the line of play

[Stephen Tu]

I think this is one of those where bidding again tends to lock DK into south hand making it a nearly sure thing and making GIB less likely to consider hooking diamond to be a potentially bad MP play (extra undertricks if somebody can ruff a diamond, like South has stiff, or North has Kx of diamonds and can pitch one on a spade, and later ruff a D). Combined with the Gibson single dummy engine not having kicked in yet (trick 3 normally?) and the base double dummy Monte Carlo engine not really being able to have any sort of plan of what it's actually going to do on subsequent tricks, not knowing what it's playing for yet, just trying the best card now. So I guess it scores drawing trumps as better, but HA is better than low to K because that preserves option of unblocking and taking the diamond finesse anyway if it's right (which double dummy it will get right at trick 2). And unblocking K is fine since against 4-1 trumps it's going to like hook the diamond then hook the heart since it's playing double dummy after all. It's one of those fundamental problems with DD sim play, it can recommend very weird things currently when the assumption is omnipotent guessing in the future, but then the future actually comes and it's not a DD situation anymore.

On my home GIB, basically if I force par mode and make GIB use it's highest level of thinking (kicking in Gibson immediately) it gets it right, but otherwise does HA unblock thing. It also gets it right if playing IMP scoring, there you need the diamond onside so it does it at trick 2.

So it seems to me a combo of:
- perhaps should consider more of a "make the damn contract bonus" at MP in the evaluation, considering that some may have stopped short of game and going down loses to those people anyway.
- devote more CPU time and kick in the single dummy reasoning earlier if/when economics make it feasible to do so.

[manudude03]

I think the bug is cashing the ace of hearts in the first place. Once you do, you really need to unblock since you need a way back to hand after taking the diamond hook. Also having said that, if you really did do that, then surely you are taking a diamond hook immediately rather than continue to draw trumps.

[1eyedjack]

As I recall, GIB waits until about trick 3 before its "advanced" card play algorithm, such as it has one, kicks in. With 12 or 13 cards per hand the resources consumed in catering for the exponentially greater permutations are supposedly just too onerous.

That is a shame, if so, because so much of the destiny of the hand is typically wrapped up in the first two tricks.