[rduran1216]

Maybe I'm missing information on how the robots defend, but they are terrible at making obvious no-trump defenses.

[Bbradley62]

Did you have a specific example to show, so that the developers might see a specific problem, or are you just venting?

[rduran1216]

Bbradley62, on 2011-March-17, 20:16, said:

Did you have a specific example to show, so that the developers might see a specific problem, or are you just venting?

I've played hundreds of hands against NT lately with all robots. GIB seems prone to make terrible switches against 3NT, I'm actually curious if other players have experienced the same thing. Maybe I'll take a log of hands to post.

[Bbradley62]

Actually, this one was pretty funny. What's up with West pitching the ♠A on Trick 3?

---

[manudude03]

Here's one I had tonight:

Link to hand

Not sure why East unblocked the ♦Q either.

[rduran1216]

manudude03, on 2011-March-17, 20:25, said:

Here's one I had tonight:

Link to hand

Not sure why East unblocked the ♦Q either.

this is the kind of crap I'm talking about, and when ur the partner of someone who does that, u sit and wonder wtf is going on.

[Bbradley62]

Same board, another table. Why does West drop his ♣Q under South's ♣K?

---

[pettert]

Bbradley62, on 2011-March-17, 20:32, said:

Same board, another table. Why does West drop his ♣Q under South's ♣K?

---

Give partner Jxx in clubs, and this is the winning defence.

[cloa513]

pettert, on 2011-March-18, 07:14, said:

Give partner Jxx in clubs, and this is the winning defence.

Jxx would only keep declarer to game as you can see dummy and pretty sure that declarer has the spades values- tricks for defence 1 club, 2 diamonds and 1 spade.

[barmar]

Since GIB uses double-dummy analysis, it defends as if declarer is also playing double-dummy. So it assumes that if a finesse is working, declarer will always take it, so it's not giving anything away by playing the card that's going to be finessed.

[cloa513]

barmar, on 2011-March-19, 00:25, said:

Since GIB uses double-dummy analysis, it defends as if declarer is also playing double-dummy. So it assumes that if a finesse is working, declarer will always take it, so it's not giving anything away by playing the card that's going to be finessed.

Why doesn't GIB just concede or claim after the first lead then!

[ahydra]

It wouldn't take much modification to get GIB to play the lowest card instead of dropping finessable honours... and the hand with the pitch of the spade ace is just insane, no idea quite how a bug like that gets through QA.

ahydra

[barmar]

QA? I think GIB was written by one guy (Matt Ginsberg) working alone. He eventually sold or licensed it to BridgeBase, and they've tweaked it, but I suspect it's still 90% what they got from him.

[helene_t]

cloa513, on 2011-March-19, 03:01, said:

Why doesn't GIB just concede or claim after the first lead then!

GIB doesn't know the opponent's (or his partner's) cards. What DD means in this context is that it makes its decisions based on DD analysis of a number of random hypothetical deals, which are consistent with the bidding and play so far.

But yeah, I recognize the problem of awful switches against notrumps. Maybe upgrading to "advanced" robots would help, if the issue is a too small sample size for the simulations.

But obviously those simulations have to be based on certain assumptions about which plays are consistent with which holdings. Those assumptions are bound to be wrong on some hands.