[1eyedjack]

---

MP, Instant, 34

This was played on 18 July 2015. Not sure whether that was before or after the implementation of the current-gib-where-possible-at-other-tables change, but the auction was replicated at the majority of tables.

[iandayre]

BBO, we deserve an explanation here. Not so much for the fact that GIB could not find the obvious PASS of 2C, but for the fact that it rebids Diamonds on a 3 card suit, when the description calls for 6 card length!

[Bbradley62]

iandayre, on 2015-July-26, 18:46, said:

BBO, we deserve an explanation here. Not so much for the fact that GIB could not find the obvious PASS of 2C, but for the fact that it rebids Diamonds on a 3 card suit, when the description calls for 6 card length!

Really? We deserve? The best answer would be "Thanks for reporting; it's fixed in the next release".

[iandayre]

We're paying customers BBradley. You don't think we deserve anything from BBO?

Let me add something else. BBO is licensed by ACBL to award masterpoints. Many ACBL members oppose the idea of awarding masterpoints for play against robots. I would like that practice to continue. But I fear that if BBO does not fairly quickly get much busier in improving the product, it will become very hard for ACBL to justify the practice. Thoughts?

[1eyedjack]

My own thoughts are that each version is a noticeable improvement on its predecessor. That much was evident to me when playing Instant tourneys at which the other tables were playing an earlier version. That empirical evidence is likely to disappear in future versions due to the recent policy of matching versions in Instant tourneys. Not saying that the benefits don't outweigh that issue, but it is a definite loss of sorts.

I am also of the view that flakey bizarre actions like this will never be wholly eliminated. The best that we can expect is an ever continuing reduction in their frequency. That has been my observation so far, but it is not going to prevent me from posting the Salvador Dali hands when they arise.

Don't forget also that in all-human tournaments you get players doing ridiculous things. That does not disqualify the event from issuing Master Points. Last time that Uday published his results, he showed that against an average BBO human field the Robots performed at about 57%. Sure, no small contribution to that result is accounted for by their perfect memory and count of the cards that have been played in a hand, but why should that matter?

That 57% result was using a version of GIB that was quite a bit older than the current version 34.

Every month there is a Masters and Royals tournament, at which the human players are on average significantly better than the average run of the mill BBO hoi poloi. And consistently, pretty much every month, a human player partnering a Robot (sometimes several such pairs) ranks well high up there in the pairs event, despite that the opponents may not be playing the same methods as the robot and the robot can never be informed of that.

So I for one think that I am getting value for money. I suppose that much must be obvious by my continued participation in paid tourneys. If and when BBO scales back on its commitment to continue with improvements, at that point I may re-evaluate that conclusion, but in the meantime I recognise that these things take time and I have no reason to doubt the assurances provided by BBO that resources are in play to continue to improve the product.

My main gripe, if I have one, is the lack of explicit feedback from BBO on particular reports. Even if a "fix" is not on the immediate horizon it is comforting to know that they are being read and considered. OK, sometime back we had a blanket assurance from jdonn that they are all considered, but it is not quite as motivating as in times of yore when Georgi would provide input to each post. Perhaps it is just too onerous. But those posts of Georgi did have a secondary benefit, as on occasion they suggested that the problem had not been accurately assessed and or the fix in the process of implementation appeared non-optimal, and we were able so to comment before it became a fait accompli.

[iandayre]

1eyedjack, on 2015-July-27, 11:05, said:

My own thoughts are that each version is a noticeable improvement on its predecessor. That much was evident to me when playing Instant tourneys at which the other tables were playing an earlier version. That empirical evidence is likely to disappear in future versions due to the recent policy of matching versions in Instant tourneys. Not saying that the benefits don't outweigh that issue, but it is a definite loss of sorts.

I am also of the view that flakey bizarre actions like this will never be wholly eliminated. The best that we can expect is an ever continuing reduction in their frequency. That has been my observation so far, but it is not going to prevent me from posting the Salvador Dali hands when they arise.

Don't forget also that in all-human tournaments you get players doing ridiculous things. That does not disqualify the event from issuing Master Points. Last time that Uday published his results, he showed that against an average BBO human field the Robots performed at about 57%. Sure, no small contribution to that result is accounted for by their perfect memory and count of the cards that have been played in a hand, but why should that matter?

That 57% result was using a version of GIB that was quite a bit older than the current version 34.

Every month there is a Masters and Royals tournament, at which the human players are on average significantly better than the average run of the mill BBO hoi poloi. And consistently, pretty much every month, a human player partnering a Robot (sometimes several such pairs) ranks well high up there in the pairs event, despite that the opponents may not be playing the same methods as the robot and the robot can never be informed of that.

So I for one think that I am getting value for money. I suppose that much must be obvious by my continued participation in paid tourneys. If and when BBO scales back on its commitment to continue with improvements, at that point I may re-evaluate that conclusion, but in the meantime I recognise that these things take time and I have no reason to doubt the assurances provided by BBO that resources are in play to continue to improve the product.

My main gripe, if I have one, is the lack of explicit feedback from BBO on particular reports. Even if a "fix" is not on the immediate horizon it is comforting to know that they are being read and considered. OK, sometime back we had a blanket assurance from jdonn that they are all considered, but it is not quite as motivating as in times of yore when Georgi would provide input to each post. Perhaps it is just too onerous. But those posts of Georgi did have a secondary benefit, as on occasion they suggested that the problem had not been accurately assessed and or the fix in the process of implementation appeared non-optimal, and we were able so to comment before it became a fait accompli.

I agree with you Jack. I think we can agree that declarer play is GIB"s strongest area, which is why it does pretty well in those tournaments you mentioned. But when we play in Robot Tournaments, the human usually declares all the hands, which magnifies GIB's shortcomings on defense and, in particular, bidding. In fact I would not mind if they left defense completely alone, and focused on the weaker areas of bidding.

And with regard to bidding errors, I am more tolerant of those where GIB varies somewhat from the descriptions for its bids. I am less tolerant of the cases where the descriptions themselves are contradictory or just don't make bridge sense, meaning that both GIB and the human partners have little chance of reaching the best contract.

[cloa513]

Bbradley62, on 2015-July-26, 19:00, said:

Really? We deserve? The best answer would be "Thanks for reporting; it's fixed in the next release".

GIB takes the forcing part of non-forcing seriously even if it puts into the silliest bid choice. It also has many forcing descriptions written in an invisible non-readable font.
That's the answer that we have had over and over and its still broken. How fundamental is its to have coding for forcing and non-forcing bids- it should have been in GIB Version 1- obviously Ginsberg couldn't programme at all. They have to do a serious investigation not just put a crappy patch in the next version.

The best answer probably would be GIB is too broken- instead we will take up the offer for Bridge Baron 13 bots to be licensed for BBO.

[lycier]

I think some comments are too radical,this issue should be a bug , or robot bidding server is too old and too worse.
So the best answer is " Thanks for reporting,it's fixed in the next version.If you can help robot generously with some money for today,the next version may be released tommorrow."

[1eyedjack]

Hey, good job robot.

Placed first, second AND third in the July 2015 masters and royals tourney.

[steve2005]

1♦-2♣ for the average club player is a very ill-defined sequence especially as a passed hand. So I think we should give Gib a break here. With my regular partner we have a lot of artificial sequences so can handle most anything. Obviously this is not a possibility for Gib. However, opposite an unpassed hand or if your not going to pass 2♣ (best bid) should be bidding 2N with short ♦ and major suit stoppers NF. I suspect this shows extras for Gib then u have no choice but to pass 2♣.

[1eyedjack]

The average club bidder can leave a lot of sequences undefined on an explicit partnership level where the computer programmer is not afforded that luxury. For one thing, he may have a solitary opinion as to its meaning and, while not agreed with partner, might justifiably or otherwise form the reasonable expectation that his partner has a similar opinion.

But if the average club bidder with a partner with whom he intended to play again encountered this sequence, I should hope that they would engineer not to repeat it.

I don't see it as "giving GIB a break". This is not a blame assignment project. It is an avoidance of repetition project.