[E Laurvick]

I think the title says it all. I get to play an extraordinary number of hands, almost always either one too high when my robot partner makes a ridiculous bid over what I thought was the final contract, or one too low (such as passing my slam try with an Ace more than already shown). I'm not going to bother listing hands here, just a general comment.

Good for practicing play of the hand.

Lousy for practicing defense. Have you noticed your robot partner almost NEVER leads any suit you have bid?

[1eyedjack]

If you can't be bothered to list examples, why should anyone be bothered to take note of your opinion?

Perhaps if you were at least to advise us of the user ID under which you log in when you play against GIB then we could examine the examples without your having to list them (you have not played wirth or against GIB at all in the last 2 months under E Laurvick).

Have these "extraordinary number of hands" been played in robot tournaments? I ask the question only because you would have been competing against humans sitting in your seat playing with and against identical robots presented with identical problems, in bidding and in defensive play as well as declarer play. How you fared in that field would have been largely dictated by how you fared in comparison with other humans in all three of those aspects of the game.

GIB's overall competence has been tested against the human BBO population over thousands of hands. Basic GIB plays a 55% game. Advanced GIB plays a 57% game. If it were as incompetent at bidding and defense as the OP suggests, I doubt that these results would be attainable. Which is not to say that it does not make mistakes, of course, some of which are gross, as the threads in this forum show. Diminishing with the release of later versions over the effluxion of time, I should hope. The problem to watch out for is when fixing one problem starts to create others in its stead.

[helene_t]

The thing is that GIB makes different mistakes, not more mistakes. It will sometimes pass low-level takeout doubles with only modest defensive values. It sometimes works, sometimes not, but in any case it is something no sane human would do. It will make futile-willy-style slam bids and fail to understand basic psychology, like for example that it is silly to lead a singleton trump king because declarer would not find the drop.

So GIB appears stupid when seen through human eyes. But imagine how stupid humans are as seen by robots - humans misclick, miscount, fail to calculate modestly complex probabilities, forget whether system is on after p's 1nt overcall, sometimes even forget which suit is trump.