[42krunner]

8 board 25-penny Robo-dup
MP's

I had to leave immediately and I didn't even pay attention to partner's response to Blackwood. I managed to bid up to 7NT missing a keycard and threw in a XX on a wing and a prayer for a lead. Looks like it worked.

link


Looks like he did the same against 7NTx at another table

link

[iandayre]

Amazing this is still happening after all the threads that have reported it. GIB cannot be taken seriously as a bridge playing program as long as it continues to make this blunder, over and over again.

[42krunner]

Can you post example threads?

[1eyedjack]

Here is one thread. It has more than one example hand

http://www.bridgebas...post__p__556518

[Stephen Tu]

Well it's kind of inelegant to hack in an "always lead ace vs. 7nt" type of thing, it's really totally against the underlying workings of the play engine which isn't based on following humanistic rules. The basic idea is it is dealing hands consistent with the auction, then choosing the double dummy best card on average. Not leading the ace tends to maximize the undertricks on the samples it deals. So how to fix it? Really it has to better tweak its sample, when the opps bid to 7nt missing an ace it's hard for it to generate truly representative hands that fit the auction since the opps probably don't have their bids. (Here south 7nt supposedly show "25+HCP" which is arguably a bug in the database, and holding 7 hcp I don't know how it tries to generate possible hands for South since it would be impossible). Somehow one has to tweak the dealing algorithm to generate wilder distributions and wider HCP ranges to the opps, so that it encounters enough more hands where the opps do have 13 tricks outside where it does have to lead the ace. Being more flexible on the sample when the opps might be psyching or distorting (opening 1nt with stiffs/voids, outside of HCP range as is common in robot tourneys), should help it in more defensive situations, way more than just programming in a "lead ace vs. 7nt" rule.

Really not leading aces vs. 7nt is kind of low on the things I'd like tweaked about GIB, if they ever get enough resources to really work on the play engine then presumably this can come out organically without being programmed in as a special case.

[42krunner]

Thanks Stephen

I wonder what it would have led in IMPs. Especially against the redoubled contracts.

[1eyedjack]

The way that a human would think on these hands is that maximising undertricks is a low priority (in comparison with defeating the contract). We would take the view that one down is going to be a good score even in a MP event, because we do not expect many others to be bidding this contract.

So while I may agree that hard coding GIB simply to cash an Ace against 7N is inelegant, hard coding it to defeat a contract as number one priority and then within that scenario to maximise undertricks may be a reasonable alternative strategy. Operating on such a strategy, it may naturally find the Ace lead without being told to.

I know that this is more IMP strategy than MP, but perhaps that is not so bad even in an MP event where opponents have a high probability of departing from system. I would hope that in an IMP event it would cash the Ace anyway, for just the reason stated above. Unfortunately most of the threads on the subject do not state conditions of contest (this thread being an exception).

In any case, inelegant it may be, but I suspect not difficult to program, and for that reason I would suggest hard-coding it if that is a 5 minute fix, and then put in place a more elegant solution in due course.

[broze]

A brilliant example that I played. Also 7NTxx. GIB partner makes on a marvelous triple squeeze.

http://www.bridgebas...-ultimate-fail/

[steve2005]

Stephen Tu, on 2014-August-19, 14:22, said:

Not leading the ace tends to maximize the undertricks on the samples it deals.

This happens so often I suspect there is a bug in the simulation program. Perhaps something like if the simulation shows no difference in what GIB leads then it makes a random lead.

So if GIB thinks 7N is hopeless is 13:1 against GIB leading the Ace