One suggestion on Gib defense

[xxhong]

Can gib implement some simple logic unit like this:

you hold

xx

QJxx

-

AKJTxxx

 

dummy shows:

Axx

xx

AKJxxxx

x

 

You successfully pushed opps to 5S.

You lead CA and switch HQ. Partner wins HA and returns H, declarer holds good trumps and DQ, so he claims after drawing trumps.

Here the major defensive mistake is due to the lack of definition of HQ from partner, which should deny HK. So the only line to defeat the contract is to cash HA and offer partner a ruff. This logic unit should only apply when partner leads a honor, which shows top of nothing when leading or switching. This hand happened in money bridge game of human gib pair vs. human gib pair. There are simply many such kind of defensive failures every day, which cost the good guy a lot of money and make the game random and unpleasant to play.

[barmar]

GIB knows that the ♥Q is the top of a sequence.

 

Other than knowing about standard honor leads and count signals, GIB has no logic units in its defense. It doesn't have a way of inferring automatically that you played this way because you have a diamond void.

 

Everything in its defense is based on simulations. It deals out a few hundred hands that it thinks are consistent with the bidding and play so far, performs double dummy analysis of them, and finds the plays that have the best expected value. If it doesn't deal out a bunch of hands with diamond voids in your hand, it won't find the ruff.

 

There are no plans to redesign GIB's card play algorithm. I'm not competent to do it.

 

I'm not even sure I would find that defense myself at the table. If I can't do it myself, I certainly can't program a computer to do it.

[cloa513]

GIB knows that the ♥Q is the top of a sequence.

 

Other than knowing about standard honor leads and count signals, GIB has no logic units in its defense. It doesn't have a way of inferring automatically that you played this way because you have a diamond void.

 

Everything in its defense is based on simulations. It deals out a few hundred hands that it thinks are consistent with the bidding and play so far, performs double dummy analysis of them, and finds the plays that have the best expected value. If it doesn't deal out a bunch of hands with diamond voids in your hand, it won't find the ruff.

 

There are no plans to redesign GIB's card play algorithm. I'm not competent to do it.

 

I'm not even sure I would find that defense myself at the table. If I can't do it myself, I certainly can't program a computer to do it.

If GIB knows that QH is top of sequence then there can be no point in returning that suit.

Then you eliminate the hearts option and clubs so what is left. If its spades then partner will get that the trick anyway so the only reasonable lead is diamonds.

 

 

What a bunch of dumb simulations to miss something so obvious.

[barmar]

If GIB knows that QH is top of sequence then there can be no point in returning that suit.

It's a safe exit. Declarer is always getting his K, it doesn't give away a trick. I'm pretty sure I've made that play hundreds of times. If I don't have a better play, I'll just continue the suit.

 

You haven't shown partner's hand or the auction, so it's hard to tell what GIB might infer from them. But we know that GIB is not always good at selecting hands in its simulations.