[thepossum]

---

It was a bit embarrassing. The simplest possible bidding and play yet it scored 100% for yours truly
For everyone else East seemed able to find 3 winners

On a very small sample I appeared to trick GiB by playing the J clubs on trick two (I was declaring as North )

[smerriman]

If you want a GIB-related answer, rather than a real-bridge answer, you'll need to post more details about what happened at the other tables.

[thepossum]

OK

Brief summary. Almost every other table bid 4S, either the same way as me or via 2NT - EDIT correction, several never made it past 2 but that's irrelevant
What appears to be the difference was the play of the J or x of Clubs on trick 2 leading in my case to a suicidal switch to Hearts by East, yet on other tables maintaining a successful diamond attack on trick 3

Enough to "trick" a sim obviously, somehow

And I have sampled about 60% of the tables that also bid 4 and my hypothesis on the play of Clubs on trick 2 appears to hold up

[smerriman]

thepossum, on 2022-August-19, 17:53, said:

What appears to be the difference was the play of the J or x of Clubs on trick 2 leading in my case to a suicidal switch to Hearts by East, yet on other tables maintaining a successful diamond attack on trick 3

Well, there's your answer If you play a different card to every other table, GIB will come up with a different set of simulations at your table compared to others.

When I put GIB in this position repeatedly, it decided to play a heart almost every time, but a diamond about once every 30 occasions.

This is independent of what card you played - it'll pick a diamond once every 30 times if you play the jack too.

In this case, the other tables stumbled upon the unlucky "one in 30" shot.

It was just as likely that you'd stumble upon the unlucky "one in 30" shot and get a 0% score.

[thepossum]

smerriman, on 2022-August-19, 18:05, said:

Well, there's your answer If you play a different card to every other table, GIB will come up with a different set of simulations at your table compared to others.

When I put GIB in this position repeatedly, it decided to play a heart almost every time, but a diamond about once every 30 occasions.

This is independent of what card you played - it'll pick a diamond once every 30 times if you play the jack too.

In this case, the other tables stumbled upon the unlucky "one in 30" shot.

It was just as likely that you'd stumble upon the unlucky "one in 30" shot and get a 0% score.

Maybe if the defence is that flaky it would be better put to more like a coin toss decision on which line of defence. Maybe more accurate representation of variance. I am surprised at many things about how Gib "sims" work or not and understand problems with how seeds are used or generated it still seems a bit unfair

How would a reasonably competent human have defended?

Would the play of the Jack versus x really be enough to change a Double Dummy calculation on that defence. Its curious to some of us

I will remember to do it next time the situation arises and see what happens

EDIT. I tried it in another Bridge program and the initial defence was a heart, not a diamond. However after forcing a diamond defence it continued with it

[smerriman]

thepossum, on 2022-August-19, 18:48, said:

Would the play of the Jack versus x really be enough to change a Double Dummy calculation on that defence. Its curious to some of us

In case I wasn't clear, the card you played had zero, zip, nada influence on the robot's logic.

Imagine a different robot at each table. You ask it to generate a random number from 1 to 30, and you win if you get a 30.

You have a 1 in 30 chance of winning, right?

BBO has adjusted the robots so that they all get the same random number if the human does everything identically.

You did something different, so your robot generates a different random number. You still have the same 1 in 30 shot at winning.. and fluked the winning number, but it wasn't because of what you did.

In real bridge, humans signal, so East would know to continue with a diamond.