[eagles123]

---

what the heck is this pass by GIB West

[1eyedjack]

Is this a case of passing being the default action when it cannot think of a specific suitable alternative?

[iandayre]

Do you suppose that anyone at BBO is even the slightest bit embarrassed by this?

No, I don't know exactly how much time and expense BBO puts toward improving GIB. I do know, that for a program that has been around for many years, this is unacceptable.

[Stephen Tu]

7-6 hands don't come up very often, it's not surprising it's not optimized to be able to evaluate them properly and have the right rules in place to come up with a reasonable bid with them.

I imagine the double of 3♠ is probably ill-defined also.

I don't see why it would be particularly embarrassing. Remember that computers don't have "common sense", they can't construct rules on the fly to deal with the unusual, they just follow what rules it has. And when you go about defining rule sets for a computer you probably aren't emphasizing how to deal reasonably with 7-6 hands.

[MrAce]

eagles123, on 2015-May-04, 03:19, said:

---

what the heck is this pass by GIB West

It has only 15 hcp

[manudude03]

---

Clearly GIB simulated the above layout

[johnu]

I believe this is the 1, 2, 3, 4 rule. All those levels of bid have been made so West cannot bid.

[iandayre]

Stephen Tu, on 2015-May-04, 15:31, said:

7-6 hands don't come up very often, it's not surprising it's not optimized to be able to evaluate them properly and have the right rules in place to come up with a reasonable bid with them.

I imagine the double of 3♠ is probably ill-defined also.

I don't see why it would be particularly embarrassing. Remember that computers don't have "common sense", they can't construct rules on the fly to deal with the unusual, they just follow what rules it has. And when you go about defining rule sets for a computer you probably aren't emphasizing how to deal reasonably with 7-6 hands.

It would be one thing if GIB had made the incredibly wimpy call of 5C. You could then say that it doesn't appropriately handle extreme distribution. But to PASS with this much high card strength? And once again, had GIB been introduced recently, OK, give it time. But it's been around more than 20 years and owned by BBO many years-don't recall the exact date. To have so many significant blind spots - see also the recent hand with 8 solid Diamonds that never bid - simply doesn't cut it for me.

[Stephen Tu]

I don't think you fully appreciate the incredibly large number of possible bidding sequences, especially competitive ones. It's hard to patch them all. Humans can adapt to new sequences easily, it makes intuitively makes sense for us to bid some number of clubs. Even when we have never encountered a specific sequence, we can create an analogy to something similar and work out what to do. But for a computer, if it has never held 7-6 after the sequence 1h-p-2d-(3s)-dbl-(4s)-? maybe there are not explicitly defined rules for what 5c/6c mean, or they are defined such that this hand doesn't fit them for some reason. Humans can usually work out something sensible to do when encountering a freak hand and lots of interference (but not always, just look at the human forums here asking what to do after preempts).

Computers can't just decide what things ought to mean on the fly, they have to be taught explicitly on almost each individual sequence, they can't think "well, this kind of feels like this other sequence I know, I suppose bids should mean ...". If it's a new sequence not likely to be encountered often, I'd expect there to be holes even after 18 years, . It's not like they have a whole team of dedicated programmers working full time to build a super bridge playing program like IBM did with chess. It was like a one man operation, a side hobby for half a dozen years, then put on the shelf for a few years, then shoved off to BBO where again it's like a couple guys at most working on it on a part time basis. You expect too much, given the amount of resources being dedicated to the project, and I don't think appreciate how differently computers think and how incredibly dumb they can be without being given explicit rules on each possible sequence. Do you have significant computer programming experience?

[iandayre]

Stephen Tu, on 2015-May-05, 13:18, said:

I don't think you fully appreciate the incredibly large number of possible bidding sequences, especially competitive ones. It's hard to patch them all. Humans can adapt to new sequences easily, it makes intuitively makes sense for us to bid some number of clubs. Even when we have never encountered a specific sequence, we can create an analogy to something similar and work out what to do. But for a computer, if it has never held 7-6 after the sequence 1h-p-2d-(3s)-dbl-(4s)-? maybe there are not explicitly defined rules for what 5c/6c mean, or they are defined such that this hand doesn't fit them for some reason. Humans can usually work out something sensible to do when encountering a freak hand and lots of interference (but not always, just look at the human forums here asking what to do after preempts).

Computers can't just decide what things ought to mean on the fly, they have to be taught explicitly on almost each individual sequence, they can't think "well, this kind of feels like this other sequence I know, I suppose bids should mean ...". If it's a new sequence not likely to be encountered often, I'd expect there to be holes even after 18 years, . It's not like they have a whole team of dedicated programmers working full time to build a super bridge playing program like IBM did with chess. It was like a one man operation, a side hobby for half a dozen years, then put on the shelf for a few years, then shoved off to BBO where again it's like a couple guys at most working on it on a part time basis. You expect too much, given the amount of resources being dedicated to the project, and I don't think appreciate how differently computers think and how incredibly dumb they can be without being given explicit rules on each possible sequence. Do you have significant computer programming experience?

No I am certainly not a programmer. I don't suggest it is easy to produce a good bridge playing program. But you have identified the problem - just a "couple guys at most working on it on a part time basis". You have seen my posts and that is exactly what I have called for, for BBO to devote more time and resources to improving GIB. There are other bridge-playing programs out there but GIB is the only one licensed to award ACBL Masterpoints, so that is where I do my playing. Given that relationship between BBO and ACBL, I think that BBO has an obligation to put a higher priority on improving GIB. Don't you?

[Stephen Tu]

It would be nice if GIB were better, sure. But the question is how much are BBO users collectively willing to shell out for a better bot? Would they pay 3x fees per tournament? 8x? People gripe about it but they are still playing the robo tourneys because it's still basically a level playing field since the bot screw up against everyone. I gripe about it but still prefer to practice vs. bots than against BBO randoms, because the bots don't mind if my partner and I stop to fully discuss a bidding sequence, you don't have to worry about opps constantly leaving for whatever reason, plus the bots despite their frequent screwups are still arguably better than the average BBO random, who also do a lot of screwups, although screwing up in human ways rather than machine ways.

BBO is only really incentivized to improve the bots if people quit playing the robo tourneys because GIB is so bad. Or if they think they'd significantly increase revenue if they improved the bot, enough to pay extra programmers to attack the problem. Is another site going to pop up offering a better robot that will take BBO's business?

So basically you are whining in a "I want a pony" type way. The problem is probably a lot harder than you think it is, probably a lot more expensive than you think it would be to fix at a rate that would make you happy. Probably not feasible given BBO's revenue stream. They are doing what they can, IMO your expectations are unrealistic.

[fred]

Stephen Tu has a good grasp of some of the main reasons why this is so difficult. He also explains things well (thanks).

But his guess as to the number of people-hours that BBO dedicates to improving GIB understates the actual number by well over 100%. I promise that improving GIB has been one of our highest development priorities for several years and I expect that this will remain the case for the foreseeable future.

We really do care about this. Sorry that some of the hands that GIB plays cause some people speculate that this is not the case.

Fred Gitelman
Bridge Base Inc.
www.bridgebase.com

[1eyedjack]

Stephen Tu, on 2015-May-05, 13:18, said:

I don't think [you] appreciate how differently computers think and how incredibly dumb they can be without being given explicit rules on each possible sequence.

And that's the rub, in the opinion of one with almost no programming experience.

Continuing with the above caveat, it seems to me that on the hand at the head of this thread, one of precisely two possibilities governs West's action over 4S:
1) There is a rule that has been programmed to tell West what to do, xor
2) There is no such rule (yet).

If there is a rule, and that rule specifies that West should pass, then the rule is crap, or a bug if you like, and needs to be changed.

If there is no rule, then I would have expected GIB to be required to simulate a sample of hands on the basis of which it would suggest a final contract. It defies belief that such a procedure would result in a sample of any magnitude that would on balance suggest that passing 4S is optimal.

There is some evidence from past threads to suggest that in case 2 (there being no rule) it simply passes.

[lycier]

Stephen Tu, on 2015-May-05, 13:18, said:

I don't think you fully appreciate the incredibly large number of possible bidding sequences, especially competitive ones. It's hard to patch them all. Humans can adapt to new sequences easily, it makes intuitively makes sense for us to bid some number of clubs. Even when we have never encountered a specific sequence, we can create an analogy to something similar and work out what to do. But for a computer, if it has never held 7-6 after the sequence 1h-p-2d-(3s)-dbl-(4s)-? maybe there are not explicitly defined rules for what 5c/6c mean, or they are defined such that this hand doesn't fit them for some reason. Humans can usually work out something sensible to do when encountering a freak hand and lots of interference (but not always, just look at the human forums here asking what to do after preempts).

Computers can't just decide what things ought to mean on the fly, they have to be taught explicitly on almost each individual sequence, they can't think "well, this kind of feels like this other sequence I know, I suppose bids should mean ...". If it's a new sequence not likely to be encountered often, I'd expect there to be holes even after 18 years, . It's not like they have a whole team of dedicated programmers working full time to build a super bridge playing program like IBM did with chess. It was like a one man operation, a side hobby for half a dozen years, then put on the shelf for a few years, then shoved off to BBO where again it's like a couple guys at most working on it on a part time basis. You expect too much, given the amount of resources being dedicated to the project, and I don't think appreciate how differently computers think and how incredibly dumb they can be without being given explicit rules on each possible sequence. Do you have significant computer programming experience?

Stephen Tu, on 2015-May-05, 15:11, said:

It would be nice if GIB were better, sure. But the question is how much are BBO users collectively willing to shell out for a better bot? Would they pay 3x fees per tournament? 8x? People gripe about it but they are still playing the robo tourneys because it's still basically a level playing field since the bot screw up against everyone. I gripe about it but still prefer to practice vs. bots than against BBO randoms, because the bots don't mind if my partner and I stop to fully discuss a bidding sequence, you don't have to worry about opps constantly leaving for whatever reason, plus the bots despite their frequent screwups are still arguably better than the average BBO random, who also do a lot of screwups, although screwing up in human ways rather than machine ways.

BBO is only really incentivized to improve the bots if people quit playing the robo tourneys because GIB is so bad. Or if they think they'd significantly increase revenue if they improved the bot, enough to pay extra programmers to attack the problem. Is another site going to pop up offering a better robot that will take BBO's business?

So basically you are whining in a "I want a pony" type way. The problem is probably a lot harder than you think it is, probably a lot more expensive than you think it would be to fix at a rate that would make you happy. Probably not feasible given BBO's revenue stream. They are doing what they can, IMO your expectations are unrealistic.

Hi Stephen Tu

This is best best reply which I have seen.

What are the unfortunate things on this forum?
There are two things ：

First:Our original intention is to help bbo, and bbo ruled that we basically complain about Gib.
What is the meanings of "complain or whining"?
I think "complain or whining"ish reports are no usefull and meaningless,and might even do some damages to bbo.It let us very shocked.
We really don't know how to report.Beg you pardon since I have posted too many in the past.

Second:The tagline of this forum :
Report Gib bugs,suggest systems that Gib should play,discuss Gib strategies,report ahem,good plays by Gib.
Today I just know that is not always true,only occasionally needing.
Here many thanks to many helpful and valueable replies,they are Georgi,Bbradley62,Helene_t and Gwnn .
Of course,the best best reply is Stephen Tu.

To be honest,you should tell me earlier, not just today let us know .Hoping that you would better try to persuade them- modify the subtitle appropriately, we won't be free to make our reports, and avoid us to get confused about it because we really don't know programming.

1- reporting bug as possible - I am very willing to do so
2- For goodness' sake,for fred' sake,for uday' sake,stop whining now,it's not much further to go, any report which complain of some possible issues ,might be doing damages to bbo.
I remembered when I came here, just saw the two speakers - one is Bbradley62,another is 1eyedjack.
When the temperature makes more and more higher,its cooling is required,just like a hot day forget the winter chills.People are so.This is a very normal natural phenomenon.

Thank you very very much for your kind help.

Best Wishes

lycier

.

[Stephen Tu]

lycier, on 2015-May-06, 03:17, said:

First:Our original intention is to help bbo, and bbo ruled that we basically complain about Gib.
What is the meanings of "complain or whining"?
I think "complain or whining"ish reports are no usefull and meaningless,and might even do some damages to bbo.It let us very shocked.
We really don't know how to report.Beg you pardon since I have posted too many in the past.

Only MrAce was really thinking that you are "posting too many bugs", and I don't think anyone else here really agrees with MrAce on this. There is nothing wrong with posting bugs IMO, in any amount. Continue posting bugs!

I think the attitude should be:
- post bidding bugs that you think are fixable. There are tons of somewhat ridiculous bids that GIB makes, usually on competitive sequences. If rules can be added on more sequences as needed, hopefully GIB will slowly get better; posting bugs helps the programmers find areas where rules need to be tweaked and added. OTOH, as explained before there is not much value in posting errors of GIB's declarer play or defense. For declarer play, GIB advanced is actually quite good after the first few tricks (if it is not bamboozled by misunderstanding the auction, which might justify reporting the hand as a bidding bug). But basic GIBs are lobotomized to play really fast and will make some really bad errors due to inherent issues with double-dummy based reasoning vs. single-dummy, there is not point in complain about basic GIB since it is made dumb on purpose to lessen resource consumption. Defensive play is really fundamentally borked by inability to understand signalling, plus putting too much faith in declarer's bidding, plus the issues in assuming that declarer is double-dummy omniscient. Maybe eventually BBO will have a team capable of addressing the defense issues, but for now they don't seem to, and also reporting defense errors doesn't really help because the issues are known, and also aren't the type that can be fixed by simply adding rules that say "if this do that".


- try to be helpful to BBO and discussion of the bug by posting the actual hand with included explanations if possible (best, learn to use the 'hv' tag and cut/pasting the URL containing the LIN code), or at least the tinyurl link to myhands (second best, the main annoyances with this method are you force people to open another tab/window to look at the hand, plus the link eventually becomes invalid after some period of time).

- but don't get bent out of shape if your bug doesn't get addressed for awhile. Or make rather pointless whining about the rate at which bugs are getting fixed, bashing the BBO programmers who are doing the best they can. This isn't helpful IMO, and it shows a lack of understanding of the difficulties involved.

[iandayre]

Excellent points regarding GIB's card play. I too am willing to be patient about improving GIB's defense, and I agree that GIB's declarer play, while not at expert level, is above average. Those of us who play the ACBL and the Instant Robot Tournaments don't get the benefit of GIB's strongest point since we usually opt to play all the hands ourselves.

This should be remembered though, when BBO points to the statistics that GIB typically scores a few percentage points above average in random MP games. It is out-declaring the average player, so therefore bidding and defending somewhat below average.