GIB failure to show false preference

[calm01]

GIB is consistently passing to leave opener in a 4-3 fit when a known 5-2 fit is available.

 

For example opener opens 1S and rebids 2H over a 1NT response by GIB. GIB holding a hand with 6-9 point passes holding 2S and 3H.

 

This is so easy to fix and is a critical skill GIB is currently missing that one wonders why it has not been fixed before.

[mtvesuvius]

GiB is a good robot - It does not execute actions which are false.

 

In any case, leaving you in 2♥ is often right, since you may also be 5-5. Correcting to 2♠ on 8-9 HCP hands may work better since 2♥ could still be fairly strong. This is more of a judgement thing, personally I'd rather be in 2♥.

[barmar]

On the 2 level, 5-2 fits usually play better than 4-3, since it's easy to lose control of the hand if they declarer in the side suits.

 

I always thought it was pretty standard to take preference back to the known 5-card suit. But maybe that's only when opener's rebid is in a minor, which is often only 3 cards (and could be 2 if you're 4=5=2=2).

[uday]

What gib does is: ( constructive auctions )

 

After 1M and a 2m rebid, responder will blindly preference back to 2M with the expected hand type

 

After 1S and a 2H rebid, responder "thinks" about it ( ie, runs some what-ifs ) and chooses from alternatives. I'll ask Fred G if we should remove this what-if and make this rule-based for the common cases.

[xxhong]

I think it is a very common situation. For 5-4 majors, no much extra value, 2S is generally better than 2H.

For 5-5, no much extra, 2H is better. For hands with extra values and game is possible, 2S is clearly better than 2H because it keeps the bidding alive. Therefore, 2S is better than pass, especially when opener holds some extra (here pass can be very bad when you may belong to 3NT, 4H or 4S)

What gib does is: ( constructive auctions )

 

After 1M and a 2m rebid, responder will blindly preference back to 2M with the expected hand type

 

After 1S and a 2H rebid, responder "thinks" about it ( ie, runs some what-ifs ) and chooses from alternatives. I'll ask Fred G if we should remove this what-if and make this rule-based for the common cases.

[fred]

If we had to force GIB to either always Pass or always bid 2S, we would go with 2S, but I hope we can do better than that.

 

The other day I held something like this:

 

[hv=pc=n&s=s32h432djt8ckq432]133|100[/hv]

 

We were vulnerable at IMPs. My partner opened 1S and rebid 2D over my (semi-forcing) 1NT. It is highly unusual for us to rebid 2m with a 3-card suit.

 

My judgment strongly suggested that it was right to Pass (good diamonds, weak spades, no desire to hear partner bid again). This seemed kind of strange at the time and I suspect that Passing 2D would not even occur to a lot of players. Turns out I was right, but obviously that proves nothing. I would have had a harder problem if partner had rebid 2H instead of 2D.

 

I am not suggesting that we rate to be able to impart human-expert-quality judgment in GIB when it comes to handling this sequence, but the rules I proposed for Uday to code will certainly be a big improvement over what exists now.

 

Fred Gitelman

Bridge Base Inc.

www.bridgebase.com