A complete failure in every sense

[pilowsky]

As you know I am a pretty one-eyed supporter of GIB. But this just happened. The contract only makes 5♠. I could see that while I was bidding, but in a fit of hubris and stupidity I ignored the warnings of my wise robot partner and bid 6♠ anyway. The hand is matchpoints from a daylong. I know that I have the most HCP, so I also know that North cannot possibly have what I need. When I create a LIN file and give it to GIB to play, they bid and play in 5♠!

Now I get the results. Here is the PBN.

 

1. Me 6♠-1 4.84% my movie
2. Someone else 4♠+2 91.94% by one person - their movie
3. Top board misplayed by GIB (it happens) 100% 6♠- their movie

 

What's the bug?

 

The contract is only makeable if you don't bid it. It is an auto da fe. GIB truly is like the Spanish Inquisition!

 

Each player gets a different lead:

 

1. I get the ♥3
2. 4♠ gets the ♦10 duck!
3. And 1♠ / 2NT / 6♠ gets the ♣A, ♣

[hv=pc=n&s=sakqt4hqj862d7ck2&w=sh9743dqt9632cat8&n=sj852hakdkj5cq964&e=s9763ht5da84cj753&d=s&v=e&b=3&a=1sp2np6s(why%20not%3F)ppp]399|300[/hv]

 

I suppose it is not really a "bug". If you read "The Dream Machine" by Mitchell Waldrop, which discusses the origins of personal computing and the critical role played by JCR Licklider, you will learn that the original bug was an insect that shortcircuited the hardware. What happens in software is something else.

space for smerriman

Here, in order to play against GIB you would need to know through some sort of mystical disclosure system what sort of lead system it follows depending on the bidding sequence made. Clearly, by giving the computer no information at all, Top Board succeeded in gaining the edge.

another space for smerriman

This, of course, means that the people in the optimum contract of 3NT+2 missed out on their rightful place in the sun. I guess even robots lie sometimes.

[akwoo]

GiB's leading system is actually well known.

 

It randomly generates some number (I think 1000) hands which fit the parameters described by its opponents' bids.

 

It then picks the lead that takes the highest average number of tricks among the hands it has generated (at MPs) - at IMPs it weights the tricks by their IMP value and then computes the weighted average.

[smerriman]

This is very normal for GIB. If you look at the top scorers of daylong tournaments, every single time they're getting good results by psyching/bidding badly and having GIB defend atrociously as a result.