A concept for GIB

[sartaj1]

I'm not totally sure that GIB understands the concept of "limited hands dont bid, unless asked to".

 

This was evidenced the other day when GIB on its first go bid 3D (invitational in diamonds). Over that i chose 3NT and now GIB, since it had a super-max 3D hand, chose 4NT.

 

If it had a "I've already limited my hand" flag, it shoudnt go through the "generate hands partner can have" phase and thus be programmed to automatically pass if partner bids game in our known fit/NT.

 

Thats my suggestion !

[fred]

Good suggestion, thanks Sartaj :rolleyes:

 

This is high up on our list of "things that annoy people the most about GIB". Striking things off this list is one of the primary considerations we use in terms of deciding what areas of GIB to work on.

 

This particular problem is going to be more considerably more difficult to solve than, for example, preventing GIB from transferring to 4-card suits (which also rated high in terms of customer annoyance). That being said, we know we are going to have to tackle this problem at some point and it would not surprise me if we make a serious attempt to do so in the next version.

 

Fred Gitelman

Bridge Base Inc.

www.bridgebase.com

[pigpenz]

GIB also used to be very bad it always raised 4 major to 5 major.

It always seemed if you had AKXX raised to six you would go down, I used to dread it when that auction would come up.

[junyi_zhu]

I'm not totally sure that GIB understands the concept of "limited hands dont bid, unless asked to".

 

This was evidenced the other day when GIB on its first go bid 3D (invitational in diamonds). Over that i chose 3NT and now GIB, since it had a super-max 3D hand, chose 4NT.

 

If it had a "I've already limited my hand" flag, it shoudnt go through the "generate hands partner can have" phase and thus be programmed to automatically pass if partner bids game in our known fit/NT.

 

Thats my suggestion !

There are a few big problems of gib bidding.

 

1: once opps set up an invitational or better sequence, gib thinks that overcalls show a hand with 25 points or more.

 

2: once your side has bid a game and opps sac, gib thinks double showing extra and tend to bid more. This can be very very expensive. Cause if you don't double, gib wouldn't double with a defensive oriented hand as well and they may go down quietly

for 3 or more tricks and if you double, gib would likely to pull.

 

3: when you have passed for many times and later double opp's high level contract, gib still thinks they are takeout.

 

4: when gib isn't sure about a double, it rarely stays and tends to pull the double. I think the reverse would serve much better. When gib isn't sure if a double is takeout or not, it's for penalty.

 

5: Gib doesn't often allow competitive bids. For example: 1C 1H 2C 2H 3C: gib thinks it shows extra and is invitational.

 

6: When gib finds your side has 25 points or more, it just doubles opp's high level contract, no matter what. This can also be very expensive, cause often, it's right to bid a suit or to pass and leave the decision to partner.

 

7: Gib rarely make pure penalty doubles at high level, no matter how many aces and trump tricks it holds. I sometimes see gib defending 6NT or even 7NT with two aces and still no double. This is actually one of the biggest problems of gib bidding, cause when gib holds a lot of points, the results tend to be very weak, especially against aggressive players, cause gib tend not to double those overbids, even when they are guaranteed down two and no where to run.

 

8: If possible, a shape counting signal would make the defensive play a lot joyful. Now the defensive carding is a big mass and as I said before, wbridge5 implemented

a couting signal in first 7 tricks which is proved to be a huge edge to its success. It could be a lot of programming to implement attitude signals, but counting should be relatively quite easy and straight forward.

[junyi_zhu]

I'm not totally sure that GIB understands the concept of "limited hands dont bid, unless asked to".

 

This was evidenced the other day when GIB on its first go bid 3D (invitational in diamonds).  Over that i chose 3NT and now GIB, since it had a super-max 3D hand, chose 4NT.

 

If it had a "I've already limited my hand" flag, it shoudnt go through the "generate hands partner can have" phase and thus be programmed to automatically pass if partner bids game in our known fit/NT.

 

Thats my suggestion !

There are a few big problems of gib bidding.

 

1: once opps set up an invitational or better sequence, gib thinks that overcalls show a hand with 25 points or more.

 

2: once your side has bid a game and opps sac, gib thinks double showing extra and tend to bid more. This can be very very expensive. Cause if you don't double, gib wouldn't double with a defensive oriented hand as well and they may go down quietly

for 3 or more tricks and if you double, gib would likely to pull.

 

3: when you have passed for many times and later double opp's high level contract, gib still thinks they are takeout.

 

4: when gib isn't sure about a double, it rarely stays and tends to pull the double. I think the reverse would serve much better. When gib isn't sure if a double is takeout or not, it's for penalty.

 

5: Gib doesn't often allow competitive bids. For example: 1C 1H 2C 2H 3C: gib thinks it shows extra and is invitational.

 

6: When gib finds your side has 25 points or more, it just doubles opp's high level contract, no matter what. This can also be very expensive, cause often, it's right to bid a suit or to pass and leave the decision to partner.

 

7: Gib rarely make pure penalty doubles at high level, no matter how many aces and trump tricks it holds. I sometimes see gib defending 6NT or even 7NT with two aces and still no double. This is actually one of the biggest problems of gib bidding, cause when gib holds a lot of points, the results tend to be very weak, especially against aggressive players, cause gib tend not to double those overbids, even when they are guaranteed down two and no where to run.

 

8: If possible, a shape counting signal would make the defensive play a lot joyful. Now the defensive carding is a big mass and as I said before, wbridge5 implemented

a couting signal in first 7 tricks which is proved to be a huge edge to its success. It could be a lot of programming to implement attitude signals, but counting should be relatively quite easy and straight forward.

Let me post a few more bidding problems.

 

1 in direct seat, gib often fails to overcall 1NT which is in the right range.

2 in balancing seat, the definition of 1NT is wrong. 1C 1H p p 1N: gib thinks it shows a minimum balanced hand and hands with 18-19 would be forced to pass.

3 1C p p 2H: gib take this 2H as unlimited, but would pass partner's 2H with a lot of intermediate hands.

4 the overcalls at two or three levels by gib are generally too weak. This is also a big problem. Basically, gib thinks it can overcall with an opener, which often is wrong, especially after a passed partner with a very weak 5 card suit and balanced hand.

5 I don't know whether you have fixed it or not. In old versions, gib's bidding after

1NT 2M is completely wrong. It takes 2NT as invitational instead of asking for the minor suit.

6 Gib tends to sac when white and when it founds a 8 card fit at four level. That's also a big problem. Basically, gib assumes their game is always cold which just can't be right at the table. My suggestion is the reverse, gib should sac only when it thinks the contract in its own direction is quite good, most likely down one or down two. In money bridge, down -3 and doubled vs. 620 isn't a big deal at all. but -500 vs. +100 or +200 is a big deal.

7 gib tends to ignore the power of distribution. I have seen gib pass 1D 4H x with

a hand like: AJTxx x AKQxxx x, which turns out to be very expensive.