[virgosrock]

you hold third hand all vul
JTx
aqTx
752
863

p-1s-p(you)-4s all pass

heart(LOL) in hand you lead H ace. heart in hand because 90% of GIBBO contracts make on lead. and gibby leads from aq(xxx) quite often in such situations.
Dummy comes down - weird but true
qx
9876
qxx
AKJx

your pard (GIBBO) follows with h2

now? and why?

vrock

[virgosrock]

virgosrock, on 2018-April-21, 07:11, said:

you hold third hand all vul
JTx
aqTx
752
863

p-1s-p(you)-4s all pass

heart(LOL) in hand you lead H ace. heart in hand because 90% of GIBBO contracts make on lead. and gibby leads from aq(xxx) quite often in such situations.
Dummy comes down - weird but true
qx
9876
qxx
AKJx

your pard (GIBBO) follows with h2

now? and why?

vrock

what? no opinions

vrock

[smerriman]

GIB doesn't play defensive carding. Take a wild guess.

[steve2005]

smerriman, on 2018-April-24, 15:08, said:

GIB doesn't play defensive carding. Take a wild guess.

Not completely true. But the few signals Gib is supposed to give on opening leads according to documentation you can't be sure Gib is giving them.
So effectively there are no signals.

[smerriman]

steve2005, on 2018-April-24, 17:31, said:

Not completely true. But the few signals Gib is supposed to give on opening leads according to documentation you can't be sure Gib is giving them.
So effectively there are no signals.

Yeah, I know, I did the analysis

But the conclusion was that the fact that it signals wrong a large amount of the time, plus the fact that it's impossible to read anyway, makes it equivalent to assuming no signals at all.

In this specific case, your partner might have:

- a singleton
- any doubleton, since it plays low from doubletons (even small doubletons) 50% of the time
- Kxx or better - it normally won't play the 2 from this, but occasionally forgets for no reason
- basically any other hand.

I wish I could get a response to that thread as to why GIB is so buggy in this situation

[virgosrock]

smerriman, on 2018-April-24, 17:47, said:

Yeah, I know, I did the analysis

But the conclusion was that the fact that it signals wrong a large amount of the time, plus the fact that it's impossible to read anyway, makes it equivalent to assuming no signals at all.

In this specific case, your partner might have:

- a singleton
- any doubleton, since it plays low from doubletons (even small doubletons) 50% of the time
- Kxx or better - it normally won't play the 2 from this, but occasionally forgets for no reason
- basically any other hand.

I wish I could get a response to that thread as to why GIB is so buggy in this situation

to complete this thread GIBBO had KJ2 and played the 2.
what is mind-blowing is the creator(s) were so addled they did not even think of this.

vrock

[Stephen Tu]

It's not particularly mind-blowing. Signalling is hard to teach to humans. It's way harder to teach to computers.

Especially this situation, it's not always clear depending on rest of hand that it's safe to play the J. Give declarer Q stiff or Q doubleton of hearts. It ought to be possible to construct hands where playing low prevents dummy's good spots from being set up for discards (declarer throws one loser on 9 while partner wins T, then throws another loser on the 8), while keeping the J completely prevents this.

Not to mention there exist hands where it absolutely needs/prefers say a diamond shift now, and there is no urgency to continue hearts, not that GIB thinks of signalling in this way.

Signalling is not so easy that you can hard code in simplistic things like signal J from KJx when partner leads ace and Q not in dummy without it looking ridiculous on many other hands, expert signalling is way more complicated and varies drastically from hand to hand even holding same cards in a suit. With GIB, at the barely signals at all stage, the emphasis is mainly on not blowing tricks outright rather than giving attitude signals which are hard to code anyway. Supposedly it will signal the K when it has worthless spots to signal, say from K82.

[smerriman]

Stephen Tu, on 2018-April-25, 06:53, said:

Supposedly it will signal the K when it has worthless spots to signal, say from K82.

That's the problem. I proved that it sometimes doesn't make signals like that, on occasions where it cannot possibly cost. That is very easy to program (if double dummy shows 0 difference between the options, opt for the signal), and is how the developers said it was meant to be programmed, yet it still doesn't