[billyjef]

One of the ways I learned bridge was by practicing with GIB. As such, I often snicker, knowingly and sympathetically, when people share their frustrating experience with GIB. The one linked below, blew my mind and made me question what goes on heuristically to justify GIB's actions.


http://tinyurl.com/ybapcfeg

[steve2005]

not sure what you expect on this hand?
Gib believes it has 6+ pts.
It will bid 4 card ♠ suit before 7-card ♦ suit on minimum hands.
If it later bids ♦ is <edit>often GF certainly shows much better hand.
Gib is just following it's Walsh type programming.

[Stephen Tu]

I don't know what's so mind blowing. We generally want it to bid spades holding 4 spades and weak hand, no fit. It just is missing a rule to prefer 1nt holding in addition a 7+ minor (plus some of the 4-6s). So it follows what rules it has, bids 1s, then is totally stuck on the next round and picks the closest fit.


I've witnessed many humans make the same mistake.

[The_Badger]

billyjef, on 2017-November-24, 06:49, said:

As such, I often snicker, knowingly and sympathetically, when people share their frustrating experience with GIB. .

No snickering, knowingly or sympathetically here. Walsh-type programming (as steve 2005 says) or not, GIB just doesn't know when to PASS. Period.

There's something fundamentally wrong with GIB's programming when it makes bad decisions on the first round of bidding. Obviously, the more bids that are made the more difficult it is to iron out problems, but here is it just the assumption that length and shortness guarantees a bid: we all know that it should be counting HCPs not total points on the first round without a fit, and there should be some overriding function that stops it from making silly bids like this.

Yes, it blew my mind too.

[Stephen Tu]

I don't think bidding is so horrible. What if partner had diamonds instead of clubs, you could make 5d sometimes. Bidding can also talk e/w out of bidding a vul game sometimes. The problem is mainly missing a rule to prefer burying the spades with a much longer minor.

Also, what the hell do you mean by "overriding function to stop silly bids". It's not like programming languages have a built in standard library call donotmakeasillybid();

Humans have to put in many, many rules to define what is silly and what is not.

[steve2005]

Stephen Tu, on 2017-November-24, 10:22, said:

Also, what the hell do you mean by "overriding function to stop silly bids". It's not like programming languages have a built in standard library call donotmakeasillybid();

Humans have to put in many, many rules to define what is silly and what is not.

Where do I go for the donotmakesillybid convention for my human partners?

[billyjef]

Yeah, while going to 4!h is hopeful by me, I've not experienced, in the past decade, GIB responding with so little value. i.e., it is the first bid that misled me more than the preference to hearts with only a singleton, not that that pleased me either!...LOL. As a human, I certainly would consider where I was going if I chose to respond, and partner bids 2 clubs...knowing that 2!d is GF. Do I really want to show partner we have enough for game if he is a maximum for his opening 1 bid? But to show sympathy, if we do find a fit, GIBs hand could produce as much as 4 tricks. Big IF though.

The_Badger, on 2017-November-24, 09:36, said:

No snickering, knowingly or sympathetically here. Walsh-type programming (as steve 2005 says) or not, GIB just doesn't know when to PASS. Period.

There's something fundamentally wrong with GIB's programming when it makes bad decisions on the first round of bidding. Obviously, the more bids that are made the more difficult it is to iron out problems, but here is it just the assumption that length and shortness guarantees a bid: we all know that it should be counting HCPs not total points on the first round without a fit, and there should be some overriding function that stops it from making silly bids like this.

Yes, it blew my mind too.

[manudude03]

Is 4♥ really that much better with the ♦K? A human player may well decide to bypass the spades to treat the hand as a weak hand with diamonds, but 1♠ certainly isn't unreasonable.

[billyjef]

I don't know Wayne...even if the robots used WJS, at least 3 diamonds shows less than enough strength for game opposite a maximum 1 bid. But again, with long experience with GIB, I personally hadn't experienced a bid like this from GIB and that is what surprised me in retrospect.

manudude03, on 2017-November-24, 12:31, said:

Is 4♥ really that much better with the ♦K? A human player may well decide to bypass the spades to treat the hand as a weak hand with diamonds, but 1♠ certainly isn't unreasonable.

[steve2005]

billyjef, on 2017-November-24, 12:47, said:

I don't know Wayne...even if the robots used WJS, at least 3 diamonds shows less than enough strength for game opposite a maximum 1 bid.

The way to show a weak hand with diamonds would be to bid 1N and then cheapest possible number of ♦.
If Gib did this everyone would complain about the possible missed spade fit.

[billyjef]

I am aware of that condition, but lacking a weak jump shift, and I would not use a WJS with the hand even if I had it available, a 1NT response should still say, at the very least, I am bidding because if you have a maximum for your opening bid partner, we have game. I don't believe GIBs hand is good enough opposite a maximum 1 bid (outside a magic fit), and that for me is a criteria for making a response with minimum values; and in my world, what should be programmed into GIB's heuristic projective options.

steve2005, on 2017-November-24, 14:21, said:

The way to show a weak hand with diamonds would be to bid 1N and then cheapest possible number of ♦.
If Gib did this everyone would complain about the possible missed spade fit.

[smerriman]

billyjef, on 2017-November-24, 16:26, said:

I don't believe GIBs hand is good enough opposite a maximum 1 bid (outside a magic fit), and that for me is a criteria for making a response with minimum values;

What would you call a magic fit? I did a quick sim, and any 4 spades along with 17-21 HCP is making 4♠ about 60% of the time. So it's not out of the question hoping for 4 card support.

[billyjef]

Apparently I need to adjust my restrictions for responding to 2/1 opening 1 bids.

smerriman, on 2017-November-24, 17:02, said:

What would you call a magic fit? I did a quick sim, and any 4 spades along with 17-21 HCP is making 4♠ about 60% of the time. So it's not out of the question hoping for 4 card support.

[johnu]

smerriman, on 2017-November-24, 17:02, said:

What would you call a magic fit? I did a quick sim, and any 4 spades along with 17-21 HCP is making 4♠ about 60% of the time. So it's not out of the question hoping for 4 card support.

What are the odds of partner having 4 card spade support and 17-21 HCP???

[johnu]

Stephen Tu, on 2017-November-24, 09:19, said:

I don't know what's so mind blowing. We generally want it to bid spades holding 4 spades and weak hand, no fit. It just is missing a rule to prefer 1nt holding in addition a 7+ minor (plus some of the 4-6s). So it follows what rules it has, bids 1s, then is totally stuck on the next round and picks the closest fit.


I've witnessed many humans make the same mistake.

Responding 1NT with this hand may not be successful. Is GIB going to pass responder's 2♦ rebid with a void?

[billyjef]

And holding 5♥

johnu, on 2017-November-24, 17:28, said:

What are the odds of partner having 4 card spade support and 17-21 HCP???

[smerriman]

I was solely commenting on your point about being worth a bid if game was possible opposite a maximum. The HCP are therefore a given.

[billyjef]

The chance for game is certainly an important factor to weigh in to the decision to bid, but does that fact solely determine if one should bid, outside of context? I wonder how much context GIB is able to process and inform it's actions. Or does it just look at the possibility of a 60% game and that satisfies the equation to bid? All very curious.

I've always thought of myself as an aggressive and adventurous bidder, but, according this discussion, clearly I am more conservative than I imagined I was.

smerriman, on 2017-November-24, 20:50, said:

I was solely commenting on your point about being worth a bid if game was possible opposite a maximum. The HCP are therefore a given.

[Stephen Tu]

Chance for game opposite partner's strong jump shift hand shouldn't be the only criteria for a response. Those hands only make up a small fraction of partner's hands. The real question is whether, on average, bidding leaves you in a better contract than passing. This includes sometimes talking opps out of their best contract.

[billyjef]

I think it can be tricky because when partner does have a game forcing hand and makes a game forcing jump shift, after responder promised minimum values, there is no way to stop the train that I am aware of. Certainly, responder will sometimes have agreements to show defensive intentions, but in the realm of GIB, I didn't think 1/1 was one of them.

Stephen Tu, on 2017-November-25, 10:57, said:

Chance for game opposite partner's strong jump shift hand shouldn't be the only criteria for a response. Those hands only make up a small fraction of partner's hands. The real question is whether, on average, bidding leaves you in a better contract than passing. This includes sometimes talking opps out of their best contract.