[oscarb]

I played the robot IMP challenge and at both tables the auction went 1♣ - 1♥ - 1♠ - 3N. At one table the bot led a ♠ and at the other table they led a ♦. I have seen this before. Any idea why?

[mycroft]

I'll leave the technical to SMerriman who has studied this intently.

But the basic is that in order to determine what to lead, GIB generates a bunch of hands consistent with the auction, then determines what the best lead is from each of them. Then decides what the best lead is, most of the time, based on those simulations.

Because different robots will generate different sample hands, especially if the choices are close, there can be situations where one set of hands suggests a club is best and another suggests a spade.

Given sufficient hands to analyze, these choices should all line up. But the number of hands generated is limited, because players expect the opening lead to show up less than a minute after the auction ends :-)

[smerriman]

mycroft, on 2025-January-14, 15:13, said:

Because different robots will generate different sample hands, especially if the choices are close, there can be situations where one set of hands suggests a club is best and another suggests a spade.

In a challenge / robot tournament etc, BBO ensure the random seed is synchronised so this never occurs.

But GIB was just updated yesterday, so most likely one person completed the challenge with the old version of GIB and one with the new. This type of thing always used to happen with Instant Tournaments after each update, until enough time had passed that the old tournament had taken place with the newer robots.

[barmar]

We deployed a new version of GIB a couple of days ago. So the version used when you play the hand may not be the same as when the hand was originally played for comparison, so it may make different decisions in some cases.