[lycier]

---

At the 7th round,Gib E needn't finesse since Gib E can get 10 tricks,obviously this is a miscalculation hand.

[hotShot]

This Basic GIB does not plan the play. So it doesn't know that it can take the winners and run home.

Instead GIB uses informations from the auction for its sample set and given that S showed 5 ♦ the change that S has both ♦Q and ♦J is about 4/6.

Again there are many tricks for the declarer and no way to keep defenders from taking ♣A (and ♣Q). The typical set where many cards make no difference at all, reducing the selectiveness of the sample set.
If you press the DD-GIB button it tells you that playing ♣ will just make and any other card will allow an overtrick.

You can't see the situation when it's Wests turn to decide which card to play, but we expect about 4/6 of the samples show that the finesse will work.
So playing the finesse is the logical choice from the sample set.
GIB does not even know what a finesse is and is unfamiliar with the concept of safety.
So although there are samples where E will go down playing the finesse, the finesse is the move that works 4/6 of the time providing an extra trick.

[Bbradley62]

It might be nice to know the form of scoring on this hand.

Same hand, but completely different topic: the 3♦ bid is weird. Was West hoping for a re-opening double from partner, so he could leave it in? If 3♦ is "forcing to 3NT", West shouldn't expect East to be able to bid 3N, unless 3♦ shows diamond control (which the description doesn't say). But West initially passed with 10+ total points, so does that indicate some problem like this hand? Or maybe ♠KQx, ♥xx, ♦xxx, ♣AQxxx? Or would that hand (or others) bid 4♥ over 3♣ instead of 3♦?

[Stephen Tu]

hotShot, on 2014-May-20, 07:12, said:

This Basic GIB does not plan the play. So it doesn't know that it can take the winners and run home.

Instead GIB uses informations from the auction for its sample set and given that S showed 5 ♦ the change that S has both ♦Q and ♦J is about 4/6.

I'd suspect it's 6/6, because of the "18 total points" requirement. This is a long standing limitation, if you bid like you have more points than you actually have, GIB will play you for those points and do goofy things. Its play would be improved if someone could introduce more "fuzziness" in the samples it generates, for the opponent's bidding, in both play & defense, since opps may have a looser idea of point and distributional requirements than GIB does. It's tricky though, since if you loosen the restrictions too much then it's not taking the bidding into account in play enough. Maybe it has to calculate twice, best play with strict sample, and best play with loose sample, and go with play for looser sample if expected score drops more than some threshold number.

But this kind of improvement may never happen given BBO's current GIB development resources.

[johnu]

Stephen Tu, on 2014-May-20, 11:31, said:

I'd suspect it's 6/6, because of the "18 total points" requirement.

I think that's probably correct. Lycier has once again psyched showing a strong hand and confused (basic?) GIB.