How can GIB possily go down in this slam?

[MightyMoe]

GIB finessed the ♦Q then cashed the ♠A, playing the ♦ A and J, losing to the T when it didn't drop for down 1.

 

[hv=pc=n&s=saqt52h7daqj9cak6&w=s984hj83dkt865c83&n=sj7hakq9d74cqjt95&e=sk63ht6542d32c742&d=n&v=0&b=1&a=1cp1sp1np2d(NMF)p2hp6nppp&p=h4h7hjhahkh6ckh8d4d2dqdkh3hqh5s2c5c2cac8sas4s7]399|300|the bidding is a bit crude, but you can't be subtle when GIB is your partner.[/hv]

 

This was played at total points. There are 12 guaranteed tricks if GIB uses just a bit of basic technique in the ♠ suit. Even the 'rookie play' of finessing the ♠J produces 13 tricks.

 

What is up with this? Doesn't GIB count its developable tricks?

 

Ross

[Yu18772]

Wow, GIB doesn't usually make such mess. I cant even see how can ♦ line be right?

3♠+3♥+1♦+5♣=12 guaranteed tricks.

http://www.bridgebase.com/forums/public/style_emoticons/default/rolleyes.gif

[cloa513]

GIB finessed the ♦Q then cashed the ♠A, playing the ♦ A and J, losing to the T when it didn't drop for down 1.

 

[hv=pc=n&s=saqt52h7daqj9cak6&w=s984hj83dkt865c83&n=sj7hakq9d74cqjt95&e=sk63ht6542d32c742&d=n&v=0&b=1&a=1cp1sp1np2d(NMF)p2hp6nppp&p=h4h7hjhahkh6ckh8d4d2dqdkh3hqh5s2c5c2cac8sas4s7]399|300|the bidding is a bit crude, but you can't be subtle when GIB is your partner.[/hv]

 

This was played at total points. There are 12 guaranteed tricks if GIB uses just a bit of basic technique in the ♠ suit. Even the 'rookie play' of finessing the ♠J produces 13 tricks.

 

What is up with this? Doesn't GIB count its developable tricks?

 

Ross

GIB doesn't use cardplay- only simulations and if the simulations just happen to be the most unlikely ones then it makes a mess.

[Yu18772]

You mean GIB cant count tricks at all?

http://www.bridgebase.com/forums/public/style_emoticons/default/rolleyes.gif

[uday]

Were you playing in a tournament, or with a rental (presumably basic) bot? I'm curious too. The basic bot isn't too strong.

 

Uday

[Bbradley62]

Were you playing in a tournament, or with a rental (presumably basic) bot? I'm curious too. The basic bot isn't too strong.

 

Uday

Probably this one: http://webutil.bridgebase.com/v2/tview.php?t=3534-1325560741

[MightyMoe]

Uday,

 

It was in a Race/Best Hand tournament and the link you show could be the one. I've played several of these tournaments.

 

You can imagine the shock when the movie said 6N-1.

 

Ross

[barmar]

When GIB does its simulations, it performs double dummy analysis. So if the simulation has a ♦ layout where it's possible to take 3 or 4 tricks, it assumes it will achieve it. And there turn out to be many such layouts. Any time East has the K or T, it "knows" it will get 3 tricks (finesse the 9 if East has T, finesse Q if it has K). And when the K is on side and the T drops, it gets 4.

 

In the 250 hands it simulated, it counted the following results:

 

96 D4: 1020 SJ: 1020

62 D4: 1020 SJ: 990

10 D4: 990 SJ: 1020

82 D4: 990 SJ: 990

 

So on 178 of the hands, it doesn't make a difference which suit it attacks. But on the remaining 72, the ♦ is a 6:1 favorite to get an overtrick. So it leads the ♦.

 

Of course, once it does this it's not actually playing double dummy, so now it has to guess whether to finesse the Q or 9. And once this finesse loses, it decides that playing for the T dropping is better than now trying the ♠ finesse. This works whenever the ♦T is in the short hand, while the ♠ finesse is just 50-50. I think there may be some squeeze possibilities as well, if the same hand has the ♦T and ♠Q.

[cloa513]

♦

When GIB does its simulations, it performs double dummy analysis. So if the simulation has a ♦ layout where it's possible to take 3 or 4 tricks, it assumes it will achieve it. And there turn out to be many such layouts. Any time East has the K or T, it "knows" it will get 3 tricks (finesse the 9 if East has T, finesse Q if it has K). And when the K is on side and the T drops, it gets 4.

 

In the 250 hands it simulated, it counted the following results:

 

96 D4: 1020 SJ: 1020

62 D4: 1020 SJ: 990

10 D4: 990 SJ: 1020

82 D4: 990 SJ: 990

 

So on 178 of the hands, it doesn't make a difference which suit it attacks. But on the remaining 72, the ♦ is a 6:1 favorite to get an overtrick. So it leads the ♦.

 

Of course, once it does this it's not actually playing double dummy, so now it has to guess whether to finesse the Q or 9. And once this finesse loses, it decides that playing for the T dropping is better than now trying the ♠ finesse. This works whenever the ♦T is in the short hand, while the ♠ finesse is just 50-50. I think there may be some squeeze possibilities as well, if the same hand has the ♦T and ♠Q.

Which doesn't explain why it leads the ace of spades, Double Dummy that makes no sense- it should keep that for a finesse if the 10 ♦ doesn't drop. It must have ignored its simulations to do that.

[barmar]

Actually, at this point in the play, GIB had switched from simulations to GIBson, which tries to enumerate the different distributions of the suits and important cards, and figure out how many tricks it can make with each layout. Here's its analysis of how many tricks it can take after cashing ♠A:

 

05C 03D 25H 06S - HT -> 12

05C 35D 26H 03S - DT -> 12

05C 56D 26H 06S + DT D8 D6 D5 D3 -> 11

05C 06D 36H 35S + HT - SK -> 12

05C 03D 26H 05S - SK -> 12

05C 06D 26H 46S -> 11

05C 06D 23H 46S + SK -> 12

11C 55D 44H 33S -> 11

01C 16D 56H 06S + DT -> 12

05C 45D 26H 06S - DT -> 12

22C 55D 44H 22S + HT H2 - H5 H3 -> 12

22C 44D 55H 22S + C8 C2 - C7 C4 C3 -> 12

33C 55D 33H 22S + C8 C3 C2 - C7 C4 -> 12

 

This is its list of possible holdings in the West hand. The first line means "If he has 0-5 ♣, 0-3 ♦, 2-5 ♥, 0-6 ♠, and doesn't have ♥T, we'll take 12 tricks." There are only a few layouts where it goes down (the third line shows the layout that actually exists).

 

I'm not sure, but I think cashing the ♠A at this point may be necessary as a Vienna Coup, to squeeze out the K while running the clubs (see line 7, where it thinks it will make even if West has ♠K).

[MightyMoe]

The discussion of squeeze positions is utterly immaterial.

 

There is no way GIB should have even looked at the ♦ finesse to begin with. There is plenty of transportation so the consideration is:

Play on ♦s: slightly more than 50%

Play on ♠s: 100%. 12 tricks, guaranteed.

 

Then there is the actual line, where GIB cashed a second ♥ to pitch the ♣ K which was utterly gratuitous. This kind of 'flair' is absolutely unnecessary. It would be appreciated if that kind nonsense be shelved for the sake of some more technically sound play.

 

I know it sounds boring but if the play had gone: win the ♥ lead, ♣ to the A, ♠ 2 to the J, any double dummy analysis is not required.

 

Had GIB West doubled the NMF call and gotten a ♦ lead, there would be merit to this discussion of squeezes, otherwise, it is moot. It should have not even been in this situation.

[barmar]

When GIB made the initial choice between playing on ♦ or ♠ it wasn't just trying to 12 guaranteed tricks, it was trying for the maximum number of tricks. And with double dummy play, there are more hands where it can pick up the ♦ suit for the maximum number of tricks.