[lamford]

---

Matchpoints. Opening lead J♥

ChCh, South on the above hand, has been practising with the robots on the Solitaire version of Bridgebase, in particular establishing what the rules are for the robots accepting claims. He also learned from barmar that they play the hand single-dummy, and if they make the contract they accept the claim. SB has gone through several partners online, mainly because of his boorish and critical manner, not to mention his R factor of between 0.75 and 1, reflecting the frequency per board of his rants at the director. As a result he was paired with the robot in the North London Online pairs this week, when all other possible partners quickly paired up.

On the above hand, SB, West, led the jack of hearts and ChCh won in South and led low towards the ♠KQTx in dummy. SB put in the nine smoothly of course, and ChCh won with the queen, East following. At this point, ChCh used his new-found knowledge and claimed twelve tricks! SB rejected this of course, but it took him about three or four seconds to realise that there was an inescapable diamond loser, and declarer still had a spade guess. After the claim was rejected, ChCh continued with a low spade to the ace, landing his slam, for 90% of the match points. Charlie berated Molly the Mule for using Stayman on the North hand, which he said would have led to the wrong contract had he not been so alert as South, but she just replied that she always did with a 4-card major.

SB was unhappy and asked ChCh why he had claimed, by private message of course. "Well, I know that the robot rejects the claim if the best single-dummy line does not land the contract. Also this "advanced" dollar-a-day robot knows that the percentage line here is low to the ace on the second round, as West is three times more likely to have J9xx than a singleton 9", he replied. "If you had a singleton, SB, it would have rejected the claim, and this is does in about a tenth of a second. When there was no instant rejection of the claim, I knew that you had to have J9xx, and you should feel complimented that I believed you would find the false-card so smoothly. I also therefore make the contract if you have a singleton 9, as the robot would reject the claim immediately," he concluded.

SB was furious and kept pressing the CALL DIRECTOR button until one arrived. How do you rule?

[shyams]

From what I recall a bot defers to the human in a human-robot partnership. i.e. East does not analyse whether to reject or accept a claim.

Barmar or someone from BBO can confirm if my recollection is indeed correct.

[lamford]

 shyams, on 2020-June-05, 06:52, said:

From what I recall a bot defers to the human in a human-robot partnership. i.e. East does not analyse whether to reject or accept a claim.

Barmar or someone from BBO can confirm if my recollection is indeed correct.

I don't think that robot East even knows who its partner is, so "deference" is not possible. I don't think claims can be accepted unless both defenders (or the declarer for a defensive claim) agree, so if Robot East accepts, SB can still reject. Against two robots, claims are often accepted or rejected, as I have found when teaching a beginner. And I don't know if there is any difference between the advanced and basic robot. It used to be the case that robots could not accept or reject claims, but not any more.

[pescetom]

 shyams, on 2020-June-05, 06:52, said:

From what I recall a bot defers to the human in a human-robot partnership. i.e. East does not analyse whether to reject or accept a claim.

If so that is an anomaly too, as a robot should have the same rights as any other player, including the right to refuse a claim.

If not then it highlights one of the future improvements of electronic play - tempo management - which is both a new possibility (reduce or eliminate the ability to pass UI by varying tempo) but also a new need if robots are capable of tempo which is impossible or unlikely for humans.
It is clearly inappropriate in this procedure that ChCh should be able to detect which opponent has refused.

[ahydra]

Could this be adjudged under L68D(b)(ii)? L68D(b)(ii) explicitly states that L16 doesn't apply, so ChCh can use without penalty the UI that it was not the robot who rejected the claim, and was always entitled to know of and use information about the robots' tendencies in a claim situation (L16A2).

Now it's true that neither SB nor the robot explicitly asked declarer to play it out, as in the wording of 68D(b), but nobody called the TD either, so perhaps this request could be deemed to be made and agreed to implicitly? In particular, if SB sought a ruling later, with his knowledge of the Laws and skill at the game, I would ask him why he didn't call at the time of the claim.

ahydra

[smerriman]

If you claim against a human and robot, the robot does not take part, as shyams mentioned.

[pescetom]

 smerriman, on 2020-June-05, 14:22, said:

If you claim against a human and robot, the robot does not take part, as shyams mentioned.

I trust the human is advised of this fact before he decides.

[lamford]

 smerriman, on 2020-June-05, 14:22, said:

If you claim against a human and robot, the robot does not take part, as shyams mentioned.

This was the North London advanced robot, programmed to reject claims, not the BBO robot.

[Vampyr]

 lamford, on 2020-June-06, 02:51, said:

This was the North London advanced robot, programmed to reject claims, not the BBO robot.

Who is the North London advanced robot? Is it you?

[smerriman]

Also, for that matter, it is not true that GIB accepts the claim if the best single-dummy line lands the contract. GIB accepts if the best single-dummy line lands the contract against all possible opponent hands, even the ones not currently held.

That's why eg, if you hold AKQJx vs xxx in trumps and all you have to do is draw them to take the rest of the tricks, it will reject if you claim instantly; only accepting after you draw the first one and both follow, even if the trumps split 3-2.

So even if GIB did play a part in the claim, it would never reject/accept based on what it holds; only what declarer knows it holds. And that's how any robot should work, whether it's GIB or a fictional North London advanced robot.

[barmar]

 smerriman, on 2020-June-06, 14:24, said:

Also, for that matter, it is not true that GIB accepts the claim if the best single-dummy line lands the contract. GIB accepts if the best single-dummy line lands the contract against all possible opponent hands, even the ones not currently held.

That's my understanding as well.

Also, BBO doesn't tell you which opponent rejected the claim. So there's no way to know whether it was the robot, SB, or both.

[pescetom]

 barmar, on 2020-June-06, 14:58, said:

Also, BBO doesn't tell you which opponent rejected the claim. So there's no way to know whether it was the robot, SB, or both.

If only the robot can respond in a fraction of a second and that time is not hidden then there is a way to know, as OP points out.

[barmar]

 pescetom, on 2020-June-06, 15:02, said:

If only the robot can respond in a fraction of a second and that time is not hidden then there is a way to know, as OP points out.

I think it waits for both opponents to respond before reporting the result of the claim, but I'm not absolutely sure.

[lamford]

 barmar, on 2020-June-06, 15:20, said:

I think it waits for both opponents to respond before reporting the result of the claim, but I'm not absolutely sure.

It accepts a concession of the remainder, even if declarer has master trumps, and even if all legal plays win the remainder. It is in breach of

Law 79 A2. A player must not knowingly accept either the score for a trick that his side did not win or the concession of a trick that his opponents could not lose.

This is a MUST regulation and the robot should be given a PP for the first occasion it does this, and, eventually, it should face expulsion from BBO for repeat offences.

The basic robot appears to accept all claims in the casual area, while the advanced robot is more discerning. More relevantly, if chess and backgammon are anything to go by, there will be a plethora of different robots as more and more programmers turn to the challenge of bridge, having "solved" games like chess, go, and to a lesser extent backgammon. The foibles of the different robots will therefore be important and their wrong explanations and wrong acceptances or rejections of claims will become more relevant.

[smerriman]

 lamford, on 2020-June-06, 18:01, said:

The basic robot appears to accept all claims in the casual area, while the advanced robot is more discerning.

Again, that's not true in the slightest; it works as I described above, and has nothing to do with what type of bot you're using. Have played (and claimed) with robots for a long period of time. It is designed as such to make it completely impossible to achieve deception of the form mentioned in your story.

The only way it is exploitable is eliminating memory lapses - eg if you're forgotten whether one of your two remaining cards is high, you can claim a trick and see if it accepts. This is obviously not ideal, but far better than the type of situations you're describing, which will never happen, and about the best you could ever do with robots, who will never be able to interpret claim statements.

There are many laws that are broken on BBO; robots are primarily there to speed things up; if a human wants to concede and get on with the next hand, they are free to do so, even if it's against the laws of bridge.

[lamford]

 smerriman, on 2020-June-06, 21:03, said:

This is obviously not ideal, but far better than the type of situations you're describing, which will never happen, and about the best you could ever do with robots, who will never be able to interpret claim statements.

There are many laws that are broken on BBO; robots are primarily there to speed things up; if a human wants to concede and get on with the next hand, they are free to do so, even if it's against the laws of bridge.

There is an anomaly in that the platform uses a robot to award a score when the hand is not completed, for sure in pairs games. There it uses double dummy play even though it could, just as easily, play the hand out without looking at all four hands in a fraction of a second. A human tournament director then has to adjust for the gross incompetence of the BBO programmer. Far from speeding things up, extra work is created for the TD.

A competent programmer would also be able to have the robot comply with Law 79A2, so that when the human claims 0 tricks in a three-card ending when he validly meant to claim 3 tricks, it does not breach the Law and should do, as any ethical opponent would do, reject the claim. This again causes work for the TD who has to adjust for the breach of the Law by the robot. I recall Gary Kasparov beating an early chess version of Deep Blue and saying that the only two things the computer had to learn were chess and manners. He soon had to eat his words.

You write, "robots are primarily there to speed things up". I don't see what basis you have for that assertion.

You write: "about the best you could ever do with robots". I think you should have written: "about the worst you could ever do with robots".

And I have just played some hands against robots in Solitaire. It accepted a claim when I had KQT9x opposite A8xx but should have waited until I led the king. Actually, what it should have done was "awarded one down", because there was a careless line which failed and "any doubtful point as to a claim shall be resolved against the claimer" (Law 70A). So, rejecting the claim and waking up the declarer is also incorrect, and the robot should act as TD and award one down.

[smerriman]

 lamford, on 2020-June-07, 06:13, said:

There is an anomaly in that the platform uses a robot to award a score when the hand is not completed, for sure in pairs games. There it uses double dummy play even though it could, just as easily, play the hand out without looking at all four hands in a fraction of a second. A human tournament director then has to adjust for the gross incompetence of the BBO programmer. Far from speeding things up, extra work is created for the TD.

Given the TD needs to adjudicate anyway, how is providing additional information to base the decision on creating extra work? (Not saying that argumentatively; I don't know how TDing works, so maybe there is something that makes it harder).

 lamford, on 2020-June-07, 06:13, said:

A competent programmer would also be able to have the robot comply with Law 79A2, so that when the human claims 0 tricks in a three-card ending when he validly meant to claim 3 tricks, it does not breach the Law and should do, as any ethical opponent would do, reject the claim.
..
You write, "robots are primarily there to speed things up". I don't see what basis you have for that assertion.

Quote from barmar: "BBO was designed for more casual play, rigid adherence to the Laws was not a priority."

I do not believe in the slightest the ability to concede was an error by an incompetent programmer. Robot subs were introduced to prevent the hassles around waiting for + dealing with human subs, and I suspect if you ask anyone that plays with robots intentionally, their main reason will be wanting a quick game.

 lamford, on 2020-June-07, 06:13, said:

You write: "about the best you could ever do with robots". I think you should have written: "about the worst you could ever do with robots".

No, I meant exactly what I wrote; I believe BBO's claiming algorithm is about as perfect as it can get. It's basically the only aspect of GIB that I find faultless.

 lamford, on 2020-June-07, 06:13, said:

And I have just played some hands against robots in Solitaire. It accepted a claim when I had KQT9x opposite A8xx but should have waited until I led the king. Actually, what it should have done was "awarded one down", because there was a careless line which failed and "any doubtful point as to a claim shall be resolved against the claimer" (Law 70A). So, rejecting the claim and waking up the declarer is also incorrect, and the robot should act as TD and award one down.

You seem to be all over the place in this thread - first implying you can tell that a robot with a human partner didn't reject a claim (you can't), then stating a robot would reject a claim based on its holding (it doesn't), and now we seem to have changed topic entirely and are saying robots should deny claims where there is a 100% winning line. It really sounds like you've played very little with robots - as for me, I'm perfectly happy with their claiming system and definitely would not want this to be changed.

[lamford]

 smerriman, on 2020-June-07, 14:25, said:

we seem to have changed topic entirely and are saying robots should deny claims where there is a 100% winning line. It really sounds like you've played very little with robots - as for me, I'm perfectly happy with their claiming system and definitely would not want this to be changed.

Indeed I have only played with robots since Lockdown started, and also have indeed changed my view of what they do and don't do by trying out various scenarios against them. I don't have access to their code, and appreciate that BBO does not have to comply with the Laws of Bridge entirely. However, the principles of claims should be followed as closely as possible

With our KQT9x opposite A8xx scenario, if we play two robots, they both accept the claim, because there is a 100% winning line. If we play two humans, they "should" not reject the claim, but should call the TD who "should" rule one off, only when Jxxx is over the KQT9x of course! If you think that this difference is acceptable, we have to disagree. Also if you think it acceptable for a robot to accept the concession of a trick it cannot win, then we have to disagree. A trivial change to establish whether there is a legal sequence allowing the concession is what the TD would have to do, and this is duck soup to a robot.

You quote from barmar: "BBO was designed for more casual play, rigid adherence to the Laws was not a priority." Note the two relevant words there, "was" and "was".

[lamford]

 smerriman, on 2020-June-07, 14:25, said:

Given the TD needs to adjudicate anyway, how is providing additional information to base the decision on creating extra work? (Not saying that argumentatively; I don't know how TDing works, so maybe there is something that makes it harder).

If GIB, when awarding a score for an incomplete hand, played the last few cards single dummy rather than double dummy, a trivial change, then there would be no need for the TD to have to get through to that point in the play to consider what is known about the hand and decide on the correct adjusted score. This would be the right method in my opinion, for any hand that has started but is slow (or someone has lost the connection). Four advanced robots are now good enough to complete the hand and this is much better than the double dummy outcome which can be a nonsense.

This excellent post of 11 years ago, from the maestro, is now out of date, because advances in robots mean that we don't have to do as we did then:
https://www.bridgeba...justing-scores/

[lamford]

 smerriman, on 2020-June-07, 14:25, said:

I believe BBO's claiming algorithm is about as perfect as it can get. It's basically the only aspect of GIB that I find faultless.

In the second hand of The Good, The Bad and The Ugly, which I put in Interesting Bridge Hands today, GIB incorrectly rejects the claim of 1NT= at trick one. It is a problem I composed twenty years ago, and declarer has a (I believe only one) 100% line.