[pescetom]

So you get 12

Can anyone explain what happens at trick 11?

[smerriman]

Anyone? Yes.

But the only people that can don't

Can try to invent an explanation, like it rules out West having the ♥K from the lead, thus double dummy says both a diamond and a club are +1 until half a trick later when suddenly it isn't, but Advanced GIB is programmed to avoid that type of logic, and we've already seen other cases where GIB chooses a card that is strictly worse than another card in every possible arrangement of cards.

[sfi]

My first thought is "bid better" . But then I wondered what would happen if South had bid 7NT instead. Does anyone have the tools to test that?

[johnu]

At trick 11, double dummy, 13 tricks are guaranteed by leading a diamond. If West has ♦Q, finesse. If East has ♦Q then it is doubleton since East is known to have ♥K.

So declarer leads a diamond, West follows low, and now??? Panic!!! Should have cashed good club, oh no!?! Odds are 4 to 3 that East has ♦Q, so play for the drop.

[pescetom]

johnu, on 2022-February-06, 19:10, said:

At trick 11, double dummy, 13 tricks are guaranteed by leading a diamond. If West has ♦Q, finesse. If East has ♦Q then it is doubleton since East is known to have ♥K.

So declarer leads a diamond, West follows low, and now??? Panic!!! Should have cashed good club, oh no!?! Odds are 4 to 3 that East has ♦Q, so play for the drop.

Ok, but surely ♣T will also show 13 tricks double dummy. Is there is no single dummy override, even with just 3 tricks remaining? It would only be a few lines of code if based entirely on top tricks (which would be sufficient here), a few pages if you want it to unblock and pull suit (I seem to remember reading that there is already something similar in the claim evaluation mechanism).

[smerriman]

The GIBson algorithm kicks in after the first 2 tricks (or maybe it's 3; definitely less than 10) for the advanced robots so as mentioned earlier that's not really a valid explanation.

[pilowsky]

On a side note, robots sometimes replace humans mid-hand.
Does the replacement GIB get to peek at the played tricks or do they suffer the same disadvantage as human subs?

[pescetom]

sfi, on 2022-February-06, 18:47, said:

My first thought is "bid better"

I was tempted to do so but managed to resist Partner is a robot and it's a MP tournament on BBO, after all: 6NT+1 would have been 92% and even 6NT= was 52%, to give you an idea.

With a human I would bid 3♣ over 2NT, expecting that to be game forcing and to elict 3♠ showing 3-card support: after that I could bid to 7NT in various ways with my regular partner, or use RKCB and vanilla Specific Kings to 7♠ with anyone else. But I've not tried bidding constructively with robot after a 2NT jump and I wasn't in a hurry to experiment - my impression based on previous traumatic experiences is that robot is not interested in responder's major and is likely to get over-excited during slam investigation too.

[pescetom]

pilowsky, on 2022-February-07, 03:43, said:

On a side note, robots sometimes replace humans mid-hand.
Does the replacement GIB get to peek at the played tricks or do they suffer the same disadvantage as human subs?

It's a good question, but I imagine the answer is that they get to peek, as this saves writing extra code.