[steve2005]

Say you are missing QJ in suit and have AKT9 and combined 8-9 cards in the suit. Also, assume no entry problems etc.

Normally if opponents play Q or J after you play Ace it is a 2-1 favourite to finesse next time suit is played.

Theory
If GIB defender has QJ doubleton it will always play the Q.

So, if Gib plays the Q restricted choice doesn't apply and you have normal odds for that suit.
However, if Gib plays the J it is now a 100% play to finesse as GIB won't have QJ.

Is this theory correct?

[mythdoc]

In my experience, yes, it is. Whenever GIB has equivalent honors it always seems to play the higher one first.

[shyams]

I have observed this GIB behavior as well.

[smerriman]

According to GIB system notes it randomises its card when it has equivalent cards in a doubleton, though there's other stuff in the system notes that is 100% provably wrong. I don't have evidence either way, though I don't recall seeing such a pattern. Could do some digging to find out.

[zhasbeen]

I haven't done a study but have a large sample of games to base my guesstimate on, and it seems random to me. I'm sure that GIB has held QJ a few times in this situation for boards I've played. However, it is almost always a top and bottom proposition to take the finesse since more than half of most fields don't play RC. I usually go for the top since it would bug the crap out of me if I wimped out and it worked.

[steve2005]

smerriman, on 2020-March-09, 13:04, said:

though there's other stuff in the system notes that is 100% provably wrong. .

Exactly the notes are not totally reliable.

[pescetom]

No idea about this question, but the notes have proven unreliable about actual behaviour many times.

[smerriman]

I think you might be right. A while ago I did some probability analysis on 2000 GIB challenge hands - once you restrict it to one of West/East holding QJ and someone playing a high honor earlier in the trick, that reduced it to a tiny sample of 8, but 8/8 of those occasions GIB dropped the queen.

That must have been deleted from the code along with the other notes about signalling etc. That is awful.

Perhaps spreading the news about this might be the catalyst needed to get BBO to actually do something about the code (at least opening it up to others), rather than say it's too hard to fix anything but bidding holes.