[djanmfinn]

Why do not the robots bid exactly the same way in the same environment.Discripancies are sometimes quite astonishing. In exactly the same bidding environment 3 to 4 different sequences can sometimes be noted. Take this hand for instance:

NORTH GIB
♠V9xx
♥D7x
♦10x
♣KQ10X

SOUTH GIB DEALER ALL VUL.
♠AQ109x
♥AK109x
♦x
♣xx

At my table and 4 others the bidding went undisturbed: 1♠ 2♠; 2NT 4♠
At 2 other tables at least, also undisturbed, : 1♠ 2♠ all pass
At other tables opps made it difficult for Gib who didn't deem his hand Worth a go so he passed as opps would play 3♦. The obvious question is why not the same action at tables where opps remained silent. These différences in Gib's reactions are not very bot-like and the reasons why this may happen should be made clear: is Gib sensitive to opponents level, for example, or does its play dépend on computer performance.

[Antrax]

This has been discussed multiple times. There's a random element to GIB's bidding and play - GIB chooses some random hands to see what's common when trying to decide what to do next. For basic GIB, it might choose a small number of random hands, and then you might get unlucky when all exhibit some property that's not really probable.

[1eyedjack]

I don't understand. The conclusion expressed in this thread is at odds with the advice that I received here:
http://www.bridgebas...post__p__735380

[Antrax]

Hmm, that would make a lot of sense and shouldn't be difficult to implement.
So, the other option is that some people use the old client so GIB runs on their computer?

[Stephen Tu]

In tourney supposedly identical actions should be taken by the bots given identical bid sequence. If random main bridge club hand, there are 3 versions of GIB in play that may do different things, the basic, the advanced, and the old ones running on the old PC client, which also has variable strength settings (time it takes to make decisions) and an older buggier bidding database than the web versions.