[calm01]

GIB held

Q63
105
Q8532
1054

The bidding began

pass 2C pass 2D (by GIB)
pass 3C pass 3S (by GIB)
...


The description of the 3S bid made by my GIB partner was:

4+ S; 5+ total points; forcing

Obviously GIB' understanding of the 3S bid differs markedly from the description of the bid.

I repeat my request for the descriptions of GIB bids to corrrespond with GIB's undestanding. Until then the GIB bidding will be flaky at best.

It matters less if the descriptions make some or little bridge sense - GIB must act in a manner consistent with its bid descriptions or a reputation of poor bidding will remain and GIB bidding will continue to underperform its declarer play.

what do you feel about the issue of GIB bid descriptions not being respected?

[arigreen]

Hi calm01,

Thanks for your suggestion. I understand that it is frustrating when GIB's hand differs from the description, but I'm sure you understand that it is not always easy to choose a bid with a 100% accurate description. For example, in your auction 2♣-P-2♦-P;3♣-P-? here are the descriptions of several possible bids:

3♦ Cheaper minor -- 4- total points
3♥ 4+ H; 5+ total points; forcing to 3N
3♠ 4+ S; 5+ total points; forcing
3N 5- H; 5- S; 7-12 HCP
4♣ 3+ C; 3-5 total points

With the hand you gave, 4♣ would match the description. However, if you change the hand slightly to Qxx.Tx.Qxxxx.Jxx (6 total points) then the hand doesn't match any of the above descriptions.

Obviously in a perfect world the descriptions would encompass all or most possible handtypes, but we aren't in that perfect world. Therefore, there will frequently be times when GIB is forced to simulate and choose a "close" bid.

One could make the argument that if only one bid exists which is an accurate description of GIB's hand, then GIB should be required to make that bid without simulating. This would work well on the hand you gave, as GIB would be forced to bid 4♣. However, such an approach may produce poor results when deciding whether or not to accept a game try, as GIB relies on simulations for dynamic hand evaluation.

One thing to note is that the database of rules that GIB uses for deciding what bid to make is similar to but not identical to the database of descriptions of what bids mean.

[Bbradley62]

The descriptions of the alternatives doesn't seem to be internally consistent. If 3M shows 4+M and 5+points, shouldn't 3N deny the ability to bid 3M? Why would we bid 3N with a 5-card major? Why isn't 3N an attempted sign-off: 5-8ish HCP with no 4-card major?

[Stephen Tu]

arigreen, on Jul 21 2010, 07:30 PM, said:

3♦  Cheaper minor -- 4- total points
3♥  4+ H; 5+ total points; forcing to 3N
3♠  4+ S; 5+ total points; forcing
3N  5- H; 5- S; 7-12 HCP
4♣ 3+ C; 3-5 total points

This set seems horrible to me, GIB is forced into bad bids by bad DB choices
not covering a large enough subset of the hands it could have.

I'd suggest something like:
3♥ 5+ H; 5+ total points; F3n
3♠ 5+ S; 5+ total points; F3n
3N 5-7 points, 2-3 S, 2-3 H, 2-5 D, 4- CL
4♣ 3+ CL; 5+ total points
3♦ 4- total points any shape; OR 4-7, 4H or 4S or 5D

[arigreen]

Hi Stephen,

Thanks for your system suggestion -- it looks great to me. I agree that the continuations in this auction are rather horrible. My point was that GIB can't simply choose the bid that best matches the descriptions, because the descriptions themselves are poor.

The best thing we can do is to add rules for as many specific auctions as possible. I'll start by adding Stephen's suggestions over 2♣-P-2♦-P-3♣.

[calm01]

Bbradley62,

I agree about the lack of self-consistency inherent in the GIB's options.

The approach you suggest makes bridge sense to me - and is one I actually employ with one particular partner - it has worked reasonably well for over a decade.

Ari,

Thanks for your prompt analysis and reply.

I accept that GIB is selecting from a poor range of options in this situation.

So the problem is not of GIB programming, but the options it has been provided with have little inherent bridge logic.

Based on playing with over 20 partners and six completely different bidding systems, there are probably only three broad but practical approaches to responding to strong bid opening after having given a first negative/relay depending on support/strength combination:

First let us assume that 2D is a first negative based on a point count/control agreement then a second response to a suit rebid by the strong hand probably needs to be based on something like one of three broad styles:

- when less than say any one King or the Queen of partners suit, support with 2 good/any 3 cards or if less support give second negative,

- with say at least any one King or the Queen of partners suit,

a) support or bid own reasonable 4/5 card suit (4/5 and definition of
reasonable according to style),

or support or show stops/features for 3NT when partners suit is a
minor,

or c) support or give second point count/control answer.

The choice of a,b,c is a matter of partnership style and partnership perception of frequency/utility.

If the first 2D response is just a relay then the strength requirements for an own suit bid are obviously higher but the bridge logic is basically the same.

I appreciate that bridge logic may be without the programmers scope but if GIB is to improve, both the bridge logic needs to be sound (self-consistent and take account of frequencies) and the programming needs follow the options. Here the problem appears to be in the bridge logic not in the programming.

For convenience I repeat your descriptions of the options available to GIB after 2C 2D 3C.

"it is not always easy to choose a bid with a 100% accurate description. For example, in your auction 2♣-P-2♦-P;3♣-P-? here are the descriptions of several possible bids:

3♦ Cheaper minor -- 4- total points
3♥ 4+ H; 5+ total points; forcing to 3N
3♠ 4+ S; 5+ total points; forcing
3N 5- H; 5- S; 7-12 HCP
4♣ 3+ C; 3-5 total points"

Assuming you agree there is little bridge logic in these options, what is the best way forward for all four of 2C 2D then 2H/2S/3C/3D by the strong hand.

Will a review of the bridge logic behind all the GIB bid descriptions pay equally high dividends in terms of effort and improvement is GIB bidding?

[arigreen]

I fully agree with you that improving the bridge logic behind GIB's decisions is paramount in improving the quality of the robots. The problem isn't that the bridge logic itself is too difficult, but rather that there are thousands of auctions that each need to be implemented separately. Even the definition of "biddable suit" differs based on the context of the auction. We try to focus our time on improving the more frequent auctions.