[iandayre]

http://tinyurl.com/jrnmrfg

Surely any bridge player worthy of the name would compete further after the 3S raise is passed around. But, as we all know, GIB has unrealistically high strength requirements for such actions, giving me a problem. 25+??? I am sure our BBO folks know better. I decided I just could not stomach selling out and bid 4C. Now look at the N hand. 95+% of players would pass without a thought. A few might try 4H, even fewer might consider 5C. But not GIB. It likes its hand in view of the massive values promised by 4C, and CUE BIDS!?!?

It is too bad that BBO does not choose to invest the time and resources necessary to eliminate these clear-cut erroneous descriptions. We can all hope for some improvement, some time.

I have a prediction. If BBO does not significantly improve GIB's bidding skills within a reasonable time, the ACBL will withdraw the privilege of awarding Masterpoints. Letters to that effect are already appearing in the Bulletin.

[1eyedjack]

South's explanations are truncated in the reproduced display, but presumably the 4C shows 25+ TP.

I agree that such a requirement is absurd. I also agree that North has a fairly easy pass of a normal 4C bid. It is not abundantly obvious to me that North has a clear cut pass of a 25+ point 4C bid, however absurd that meaning.

Nor is it obvious to me that South should bid 4C even if North is on the same wavelength. You have already shown everything bar the 6th Club, and you have a low ODR in terms of honour mix.

I don't have a strong opinion on whether ACBL masterpoints should be awarded, but I do have an opinion that GIB's bidding quality should be irrelevant to that debate.

[steve2005]

4♣ is a pretty mandatory bid at mp.
4♠ is just out of lala land.
Were a human partner to pull this, probably wouldn't be playing with them much after this.

[lycier]

I think such similar things are normal.
1- Most of wrong descriptions come from abnormal distribution.If Gibs can be gotten great improvement on it in the future,we are sure that Gibs will become a real expert,so it is very difficult for Gib to overcome its defect,especially in its bottleneck period.
2- Most of wrong descriptions come from the limitation of Gib TPs evaluation.
I found a positive correlation between the hand shapes and how well the hcp evaluation adjustment could be consistent with the details of TPs reevaluating.
IMO,I think there can be correlation without causal relations.

3- Maybe there are too many of incorrect comments on Gibs,of course,including me in the past.
Just like Sayc which can't resolve too many bidding issues,Gib system is only as a public playing system instead of as a perfect bidding system. To be fair, Gibs are really better than human in the many cases.

[iandayre]

 1eyedjack, on 2016-January-16, 17:20, said:

South's explanations are truncated in the reproduced display, but presumably the 4C shows 25+ TP.

I agree that such a requirement is absurd. I also agree that North has a fairly easy pass of a normal 4C bid. It is not abundantly obvious to me that North has a clear cut pass of a 25+ point 4C bid, however absurd that meaning.

Nor is it obvious to me that South should bid 4C even if North is on the same wavelength. You have already shown everything bar the 6th Club, and you have a low ODR in terms of honour mix.

I don't have a strong opinion on whether ACBL masterpoints should be awarded, but I do have an opinion that GIB's bidding quality should be irrelevant to that debate.

I think most would agree with my 4C bid, even with a human. GIB has high requirements for competing at the 4 level in this situation, making it more necessary to compete in my opinion. But we needn't discuss that further unless you want to. I am more interested in your final comment. Care to elaborate?

[iandayre]

 lycier, on 2016-January-17, 05:59, said:

I think such similar things are normal.
1- Most of wrong descriptions come from abnormal distribution.If Gibs can be gotten great improvement on it in the future,we are sure that Gibs will become a real expert,so it is very difficult for Gib to overcome its defect,especially in its bottleneck period.
2- Most of wrong descriptions come from the limitation of Gib TPs evaluation.
I found a positive correlation between the hand shapes and how well the hcp evaluation adjustment could be consistent with the details of TPs reevaluating.
IMO,I think there can be correlation without causal relations.

3- Maybe there are too many of incorrect comments on Gibs,of course,including me in the past.
Just like Sayc which can't resolve too many bidding issues,Gib system is only as a public playing system instead of as a perfect bidding system. To be fair, Gibs are really better than human in the many cases.

This is not a system problem Lycier. If you read the system notes, I approve of the great majority of the system AS WRITTEN. Instituting realistic and reasonable strength requirements for competitive calls such as this would not require any changes to the system. GIB's ARE better than humans in some ways, most notably that they watch every spot card on every hand and never miscount. But in terms of the calls they choose, in too many cases, such as this one, they are worse than novices.

[Bbradley62]

 iandayre, on 2016-January-17, 18:06, said:

GIB's ARE better than humans in some ways, most notably that they watch every spot card on every hand and never miscount.

Except, of course, on the occasional hand like this recently-posted one: http://www.bridgebas...t-the-contract/

[Vampyr]

 1eyedjack, on 2016-January-16, 17:20, said:

I don't have a strong opinion on whether ACBL masterpoints should be awarded, but I do have an opinion that GIB's bidding quality should be irrelevant to that debate.

I suppose it depends on how similar to bridge a game needs to be for the ACBL to award masterpoints.

[1eyedjack]

 Vampyr, on 2016-January-17, 18:57, said:

I suppose it depends on how similar to bridge a game needs to be for the ACBL to award masterpoints.

I would have thought that it was sufficient that it requires the same skill set.

[1eyedjack]

 iandayre, on 2016-January-17, 17:35, said:

I am more interested in your final comment. Care to elaborate?

(For clarity, we are talking about the justification for awarding ACBL masterpoints to robot tourneys).

The question hinges on what are the purposes of master points and then whether the awards in robot tournaments further those aims.

One purpose of master points is to raise money. Yes, awarding master points in robot tourneys raises money for the ACBL, but let's dismiss that, shall we? It is not really a significant feature in the argument.

The main purpose is to provide a player with some measure of skill and/or lifetime achievement at bridge, either for personal gratification or as a measure of comparison with other players or both.

To achieve that success you need a variety of qualities that I summarise in no particular order and with no particular weight attached:

1) The ability to play bridge well
2) A sound understanding of your partner, his abilities and your methods
3) A sound understanding of your opponents, their abilities and their methods.

These 3 qualities are required in both human competitions and in robot tournaments.

GIB's methods and abilities differ from those of other human players, but the capacity for a human player, given the will and effort, to grasp those differences is high. The resources to do so are largely in the public domain and also can be gleaned from experience and practice playing with and against robots.

If they suggest that MPs should not be awarded on the grounds that GIB plays poor bridge, then quite aside from the fact that I might disagree with that premise, what if it does play poor bridge? YOU are competing against other human players in the same position. If you beat them it will be because you have better bridge skills, or a better understanding of GIB's foibles, or most likely a combination of both. Why would that be undeserving of reward/recognition? The same applies if you do well in a human event. At least in a robot event your human opponents have the same opportunities and resources to counteract the second effect.

I do not claim that there are not some arguments against awarding master points in robot tourneys. Only that the fact that GIB plays poorly (if that is the case) or does some inexplicable weird things (as do some humans), is not of itself in my view a relevant factor, and I think that the burden should be on those who think otherwise to develop that point, rather than for me to do the running in defending it.

[helene_t]

I think the implementation of GIB's bidding system suffers from lack of sensible general principles. A lot of specific situations are covered explicitly and in those situations the bids are reasonably well defined, but whenever an undiscussed sequence occurs (this affects many constested auctions but also a few uncontested ones) it defaults to some nonsense which seems to fall into one of four categories:
1: Bids show the values for that level, i.e. a 4m bid must show som 28+ TPs if partner has not shown values. Even if opps have shown the majority of the points.
2: Calls show nothing at all, i.e. almost all passes and most doubles just confirm the "3+ clubs, 11-22 points" or whatever was already shown.
3: Bids just show (say) "biddable spades" without any strength indication, even in constructive auctions where strength ought to be relevant.
4: Bids are unlimited per description but partner frequently passes them. I wonder if the sims assume they are limited or whether it is just that partner figures out that passing is best despite the bid being forcing?

It would probably require a complete overhaul of the bidding engine to get this fixed, since just keeping patching holes by extending the script is a never-ending task. Some AI engine that calculates the optimal assignment of hand types to calls would probably required. The ideas we discussed in the bidding systems designs by computer thread would probably be too radical but maybe something like:

Formulate some generic rules that restrict the design of the bidding system, for example:
- 2NT can be either scrambling or showing a combined minimum of 22-24 HCPs and stopper in opps suit
- We don't look for a new fit once a major suit fit has been found
- If opps have shown the majority of the points, competitive aims take priority
- etc.

Now run an algorithm that optimizes the assignment of hand types to calls, maximizing the expected value of some objective function based on how much extra bidding space is preserved below the safety level relative to the amount of information needed to decide whether to bypass the safety level or not.

[iandayre]

 helene_t, on 2016-January-18, 05:37, said:

I think the implementation of GIB's bidding system suffers from lack of sensible general principles. A lot of specific situations are covered explicitly and in those situations the bids are reasonably well defined, but whenever an undiscussed sequence occurs (this affects many constested auctions but also a few uncontested ones) it defaults to some nonsense which seems to fall into one of four categories:
1: Bids show the values for that level, i.e. a 4m bid must show som 28+ TPs if partner has not shown values. Even if opps have shown the majority of the points.
2: Calls show nothing at all, i.e. almost all passes and most doubles just confirm the "3+ clubs, 11-22 points" or whatever was already shown.
3: Bids just show (say) "biddable spades" without any strength indication, even in constructive auctions where strength ought to be relevant.
4: Bids are unlimited per description but partner frequently passes them. I wonder if the sims assume they are limited or whether it is just that partner figures out that passing is best despite the bid being forcing?

It would probably require a complete overhaul of the bidding engine to get this fixed, since just keeping patching holes by extending the script is a never-ending task. Some AI engine that calculates the optimal assignment of hand types to calls would probably required. The ideas we discussed in the bidding systems designs by computer thread would probably be too radical but maybe something like:

Formulate some generic rules that restrict the design of the bidding system, for example:
- 2NT can be either scrambling or showing a combined minimum of 22-24 HCPs and stopper in opps suit
- We don't look for a new fit once a major suit fit has been found
- If opps have shown the majority of the points, competitive aims take priority
- etc.

Now run an algorithm that optimizes the assignment of hand types to calls, maximizing the expected value of some objective function based on how much extra bidding space is preserved below the safety level relative to the amount of information needed to decide whether to bypass the safety level or not.

I'm no programmer, here's how I look at it. GIB is currently programmed, on the most part, to comply with the bidding descriptions as written. Unfortunately, those who wrote those descriptions had some very serious misconceptions about what constitutes sound bridge principles. If the descriptions were re-written and improved, couldn't they then be programmed in the same fashion the current ones were?

On the question of ACBL Masterpoints, they have always meant a little something to me and I enjoy that I'm gaining some again without having to leave my home or try to find compatible partners. A lot of people here apparently feel the same way, as tons of them are being won here. I'm not anywhere near the top tier of online winners. Let's say this is another, possibly more persuasive, reason to convince BBO that their current policy regarding GIB improvements may not be sustainable.

[helene_t]

 iandayre, on 2016-January-18, 17:30, said:

GIB is currently programmed, on the most part, to comply with the bidding descriptions as written. Unfortunately, those who wrote those descriptions had some very serious misconceptions about what constitutes sound bridge principals.

I am sure GIB has always been written and maintained by expert bridge players. I believe the rules are generally ok, there are just too many convoluted auctions that are not covered by any rules so GIB defaults to some very crude generic principle like "bids show the values to make the contract bid", "bids show length in the suit bid" or "bids show nothing at all".

[iandayre]

 helene_t, on 2016-January-19, 05:03, said:

I am sure GIB has always been written and maintained by expert bridge players. I believe the rules are generally ok, there are just too many convoluted auctions that are not covered by any rules so GIB defaults to some very crude generic principle like "bids show the values to make the contract bid", "bids show length in the suit bid" or "bids show nothing at all".

No offense Helene, but I emphatically disagree. One simple example - the actual hand here. What expert would put in place a 25+ requirement for further action? And there are many similar situations in other auctions. Note I am referring to the original programmers. If they were in fact experts, they did a very careless and haphazard job. I respect the BBO people who work with GIB, but they are employees who must allocate their time and resources as instructed. It is the responsibility of Mr. Gitelman to see that GIB improves. It's his business. If he doesn't choose to prioritize improving GIB, well again, I made my prediction in the OP.

[fred]

Iandayre - I have tried to explain this to you several times already, yet apparently either you don't want to hear me or you think I am lying. I am not sure why I bother, but I will try again.

You said: If he doesn't choose to prioritize improving GIB...

I have said (repeatedly) and will say again: For many years we have prioritized improving GIB and I expect we will continue to do so for as long BBO continues to exist.

You said: I'm no programmer, here's how I look at it. GIB is currently programmed, on the most part, to comply with the bidding descriptions as written. Unfortunately, those who wrote those descriptions had some very serious misconceptions about what constitutes sound bridge principles. If the descriptions were re-written and improved, couldn't they then be programmed in the same fashion the current ones were?

I say: That is exactly what we are doing (rewriting and improving descriptions), but there are a massive number of possible auctions and unfortunately bridge bidding does not lend itself well to making global descriptions that make sense in all cases and don't contradict other global descriptions one might want to make.

To me your statement is analogous to saying: I am not a doctor but I feel like I am in a position to demand from my doctor a quick and complete cure with no side effects to any of a quadrillion health problems that might befall me.

You say: I respect the BBO people who work with GIB, but they are employees who must allocate their time and resources as instructed.

I say: Our GIB programmers are instructed to spend the vast majority of their time and resources working on GIB. In addition, Uday and I (who are not primarily GIB programmers) spend a significant % of our working hours discussing and implementing possible improvements for GIB. Barmar (another non-GIB programmer on our staff) sometimes gets involved in GIB-related discussions and programming as well.

It turns out that I spent most of December (including about 12 hours per day when most people were on holiday) writing a new program that I believe has the potential to greatly increase the rate at which we are able to improve GIB. If my optimism proves to be well-founded there will be considerably more programming effort required and we may well hire more bridge experts to work on GIB. When the programming is too difficult for our team we have been known to pay up to $400 per hour for consultants who have specific areas of expertise that we lack.

If this sounds like "low priority" to you then I doubt there is anything we could do that would satisfy you.

You said: No offense Helene, but I emphatically disagree.

I say: Of course it is your right to disagree, but you don't know what you are talking about (no offense to you either). Helene, who I believe knows more than a little about computer programming in general and how GIB works in particular, is 100% correct.

I am sorry you apparently find it super-frustrating that your GIB partners sometimes make terrible bids. Probably, if your GIB partners were capable of it, they would find some of your bids to be terrible as well. Most of my bridge partners are World Champions and even they make terrible bids on occasion. If it were possible for either a computer or a human to play bridge without making terrible bids then I suspect that bridge would not be a very interesting or popular game.

I promise you that we find the rate at which GIB makes terrible bids to be completely unacceptable and that for many years we have been trying very hard to reduce the frequency of these terrible bids. Fine if you choose not to believe me, but please stop the public speculations about what you think my priorities might or might not be.

Fred Gitelman
Bridge Base Inc.
www.bridgebase.com

[mgoetze]

Now I'm sad, thinking about what Fred could do if he spent one of those 12-hour days on improving the usability of the web client (especially the audio) instead of sinking it all into GIB.

[fred]

 mgoetze, on 2016-January-20, 18:52, said:

Now I'm sad, thinking about what Fred could do if he spent one of those 12-hour days on improving the usability of the web client (especially the audio) instead of sinking it all into GIB.

Don't be sad. Probably you know this already, but just to be clear...

- I don't normally work 12-hour days or work on weekends or holidays anymore (I did do that for many years but nowadays my work schedule is pretty normal)
- I don't normally spend a lot of time doing programming that is directly related to GIB
- Almost all the programming work I do involves the BBO web-client, the mobile client, and the HTML client that we use for our solitaire games

It turns out I had what I thought was a possible big idea related to GIB in December and, since I didn't have much other work to do at the time, I decided to go full steam ahead with it.

If you can describe what exactly you would like to see with respect to audio in the web-client and why it is important to you, I might be able to do something about it. There is a massive new release of the web-client that will hopefully be available soon and I won't be able to do anything audio-related in that release, but it is possible that we can do some improvements in that area in the version after that.

Fred Gitelman
Bridge Base Inc.
www.bridgebase.com

[mgoetze]

 fred, on 2016-January-21, 08:35, said:

If you can describe what exactly you would like to see with respect to audio in the web-client and why it is important to you, I might be able to do something about it.

So as to not derail this thread further, I have started a new thread.

http://www.bridgebas...b-client-audio/

Thank you very much for your consideration.

[iandayre]

 fred, on 2016-January-20, 17:04, said:

Iandayre - I have tried to explain this to you several times already, yet apparently either you don't want to hear me or you think I am lying. I am not sure why I bother, but I will try again.

You said: If he doesn't choose to prioritize improving GIB...

I have said (repeatedly) and will say again: For many years we have prioritized improving GIB and I expect we will continue to do so for as long BBO continues to exist.

You said: I'm no programmer, here's how I look at it. GIB is currently programmed, on the most part, to comply with the bidding descriptions as written. Unfortunately, those who wrote those descriptions had some very serious misconceptions about what constitutes sound bridge principles. If the descriptions were re-written and improved, couldn't they then be programmed in the same fashion the current ones were?

I say: That is exactly what we are doing (rewriting and improving descriptions), but there are a massive number of possible auctions and unfortunately bridge bidding does not lend itself well to making global descriptions that make sense in all cases and don't contradict other global descriptions one might want to make.

To me your statement is analogous to saying: I am not a doctor but I feel like I am in a position to demand from my doctor a quick and complete cure with no side effects to any of a quadrillion health problems that might befall me.

You say: I respect the BBO people who work with GIB, but they are employees who must allocate their time and resources as instructed.

I say: Our GIB programmers are instructed to spend the vast majority of their time and resources working on GIB. In addition, Uday and I (who are not primarily GIB programmers) spend a significant % of our working hours discussing and implementing possible improvements for GIB. Barmar (another non-GIB programmer on our staff) sometimes gets involved in GIB-related discussions and programming as well.

It turns out that I spent most of December (including about 12 hours per day when most people were on holiday) writing a new program that I believe has the potential to greatly increase the rate at which we are able to improve GIB. If my optimism proves to be well-founded there will be considerably more programming effort required and we may well hire more bridge experts to work on GIB. When the programming is too difficult for our team we have been known to pay up to $400 per hour for consultants who have specific areas of expertise that we lack.

If this sounds like "low priority" to you then I doubt there is anything we could do that would satisfy you.

You said: No offense Helene, but I emphatically disagree.

I say: Of course it is your right to disagree, but you don't know what you are talking about (no offense to you either). Helene, who I believe knows more than a little about computer programming in general and how GIB works in particular, is 100% correct.

I am sorry you apparently find it super-frustrating that your GIB partners sometimes make terrible bids. Probably, if your GIB partners were capable of it, they would find some of your bids to be terrible as well. Most of my bridge partners are World Champions and even they make terrible bids on occasion. If it were possible for either a computer or a human to play bridge without making terrible bids then I suspect that bridge would not be a very interesting or popular game.

I promise you that we find the rate at which GIB makes terrible bids to be completely unacceptable and that for many years we have been trying very hard to reduce the frequency of these terrible bids. Fine if you choose not to believe me, but please stop the public speculations about what you think my priorities might or might not be.

Fred Gitelman
Bridge Base Inc.
www.bridgebase.com

Thank you for your response. I don't recall ever having received a personal response from you, but this is certainly one!

In no way do I think you are a liar, on the contrary I have the highest respect for you, for your success as a player but more so for the fact that you have built BBO into the top online bridge product in the world. I am glad to hear that you consider GIB to be a high priority, and I wish you success in implementing improvements. I have never, and do not now, expected them to come overnight.

I will say that you have misinterpreted my disagreement with Helene. It had nothing to do with programming, it had to do with that fact, as I said originally, that the descriptions in GIB, in many cases, are not in keeping with sound bridge principles. This strongly indicates to me that the original programmers were not experts. If they were, in my opinion they did a poor job. But of course that is now the past, it can't be helped and you'll be moving forward from here.

I will say that I believe most of the GIB regulars are somewhat taken aback in that there have no GIB upgrades for well over 6 months, and that BBO personnel have not been active on this forum for some time. That is the main source of my impatience. I would also add that, as a paying customer, I would have hoped that, not just in the 2+ years I have been playing with GIB, but in the many years that BBO has offered the product before that, that we would be further along now.

You have given me reason for optimism! Once again, thank you.

[Stephen Tu]

 iandayre, on 2016-January-21, 18:06, said:

I will say that you have misinterpreted my disagreement with Helene. It had nothing to do with programming, it had to do with that fact, as I said originally, that the descriptions in GIB, in many cases, are not in keeping with sound bridge principles. This strongly indicates to me that the original programmers were not experts. If they were, in my opinion they did a poor job.

It's just that your opinion is coming from someone with admittedly no background in computer science, who has no programming experience. You don't have a solid grasp of just how many possible auctions there are in bridge, and how hard it is to construct enough rules for a stupid machine to try to cover every one that comes up in practice. It's like criticizing a dog trainer that when they are done you don't have a talking dog with human intelligence like Brian on the animated Family Guy show. Unrealistic expectations for even expert bridge players who are also expert programmers.

I think professional computer programmers, like me, are more understanding of how difficult it is to train a computer for a game with as many possible sequences as bridge.

Go look at Truscott's bidding dictionary. See how many pages he has to cover, and that's only for first 4 bids of uncontested sequences. Imagine writing rules to cover all of those, and all the contested ones, and the ones later in the auction. And ordering the rules so the computer picks the best bid consistently. It's not so easy!!