[eamongall]

Last night while running my Ireland Imps Pairs nightly event a competitor drew my attention to his robot dropping a trick despite having AK of spades and 2 high trumps remaining

https://www.bridgeba...9975-1601667088

Having used verious levels of GIB since Matt Ginsberg sent me the original text version in 1996 I was amazed at this play.
The player was not complaining despite paying to use what is supposedly the advanced version of the robot.
I will also send to tournaments at BBO just for the crack

I have heard and read many complaints about GIBs bidding but thats okay as it runs on samples of hands and a bad sample can flaw its bidding
However generally its cardplay is up to a standard of claiming with 4 high cards left ... I expect my text version will claim the above hand before the end of play or before taking the practice finesse
Some poor chap shelled out 25 cents for this play and losing about 13 imps.

Same player Woody1980 also indicated hand 28 as interesting play by the robot

https://www.bridgeba...9850-1601667088

Here the robot turns a possible 11 tricks into 8 tricks ......

Later I will try these hands on my OFFLINE versions of GIB and see how they go

Thanks
Eamon Galligan

[Huibertus]

My guess is with the introduction of AI the black-outs that come with intelligens are being introduced too?

[pilowsky]

In addition to my terrible bidding and card play, this happens so commonly to me that I started a little thread for these exemplary hands - feel free to join in.

[smerriman]

This is just standard GIB. It makes assumptions based on the bidding, and never deviates from them. It knows that West is 'guaranteed' to have the spade queen for the double, and thus it can't tell the difference between playing the king and the jack - it sees them as equals.

Likewise, in the second example, it doesn't understand Multi, and assumes West has a weak two in diamonds and East a strong hand with hearts. That of course is ludicrous but its whole play engine is based around it, so it's really not unexpected tnat it would go haywire.

[goffster]

smerriman, on 2020-October-03, 13:08, said:

This is just standard GIB. It makes assumptions based on the bidding, and never deviates from them. It knows that West is 'guaranteed' to have the spade queen for the double, and thus it can't tell the difference between playing the king and the jack - it sees them as equals.

Likewise, in the second example, it doesn't understand Multi, and assumes West has a weak two in diamonds and East a strong hand with hearts. That of course is ludicrous but its whole play engine is based around it, so it's really not unexpected tnat it would go haywire.

The proper algorithm would say (at IMPS)
"Can I guarantee the contract with no assumptions" and play that way.
falling back to
"Can I guarantee the contract with assumptions" if it cannot.