Computer dealt hands - with a hiccup

[bluejak]

A friend told me of an experience in a club using computer dealt hands. During round seven, he played a board. When he came to the next board he summoned the TD [part way through the hand] and told him it was the same hand with compass directions changed. After some initial disbelief, the TD discovered this was accurate, down to the last pip.

 

One of the reasons it had not been noticed before was that the hands were rotated anti-clockwise, so a different hand was held by dealer, and the auctions were invariably different.

 

Anyway, assume the hands were 9 and 10. What scores are you going to give to the players on board 9 on the first six rounds? At this table? On board 10 on the first six rounds? At this table [ok, this last is the easy one]?

[TylerE]

An interesting problem, because this happening is actually NOT mathematically impossible, especially if you ignore pips.

[blackshoe]

I would leave board nine alone, on the theory that in most cases players will have seen that one first. For board ten, well. hm. Without claiming it is at all legal, and without consulting my law book, it seems to me that in view of the fact that apparently no one who has already played board ten seems to have noticed a problem, I would let previous results stand, tell the players at the table to which I was called to redeal the board, and treat it as a fouled board.

 

Innovative, I think, possibly practical. It might even be legal, though I suspect not. :ph34r:

[axman]

A friend told me of an experience in a club using computer dealt hands. During round seven, he played a board. When he came to the next board he summoned the TD [part way through the hand] and told him it was the same hand with compass directions changed. After some initial disbelief, the TD discovered this was accurate, down to the last pip.

 

One of the reasons it had not been noticed before was that the hands were rotated anti-clockwise, so a different hand was held by dealer, and the auctions were invariably different.

 

Anyway, assume the hands were 9 and 10. What scores are you going to give to the players on board 9 on the first six rounds? At this table? On board 10 on the first six rounds? At this table [ok, this last is the easy one]?

There is insufficient information to make any ruling, for instance it is premature to dismiss the possibility that the hands are legitimate [the lack of instructions for duplication]. Personally I experienced the oddity of deja vu during about the third round of an American swiss where the hands are shuffled every round. I had a board that I estimated was perhaps 2 or 3 pips different from being an exact duplicate from the previous day's pairs. The auction was indeed the same but on Sunday it could take one more trick.

 

At this point, it would seem that the astute friend could have been premature in drawing attention to his conclusion. If the hand had been indeed fouled it couldn't hurt to first complete the play; while if the interruption unnecessarily causes cancellation of the board, you perhaps see my point...

[pran]

With my more than 30 years experience in programming for computer dealt hands I would first of all request a thorough analysis on what has happened.

 

This includes interviews with the person ho did the actual card dealing, inspection of the files created by the card dealing computer program and special tests of this program. (Such tests are not trivial and require some expert knowledge!)

 

If all four hands are indeed pairwise identical except for positions between two different deals there is some reason to suspect an error, but the possibility that this was purely incidental is still significant. If however, two different deals are identical to the smallest detail (i.e. also with respect of positions) I shall say that this proves an error. (Be aware that pips are as significant as honours when comparing deals for the purpose of revealing such errors!)

 

And for the event in progress suspicion alone is not sufficient to cancel results on any board.

 

regards Sven

[bluejak]

I am never sure the point in telling the opening poster that the situation is not as he says it was, and posts suggesting this seem incredibly unhelpful in helping people rule.

 

When he came to the next board he summoned the TD [part way through the hand] and told him it was the same hand with compass directions changed. After some initial disbelief, the TD discovered this was accurate, down to the last pip.

That is the situation which occurred and suggestions it did not occur do not help.

 

So, would people please like to answer the question. If you want to answer some other question over something that did not happen in this case, why not start a new thread?

[barmar]

Did the computer actually deal the same hand twice, or was the problem in the duplication? The Director should be able to look at the hand record to tell. If the problem was in the duplication, and the boards were hand-duplicated, I think a warning or PP for the pairs at the table that duplicated the boards would be appropriate. The Director should then duplicate the board correctly, so the remaining tables can play it correctly. It should be treated as a fouled board for the scoring.

[bluejak]

The computer dealt the same hand twice.

[pran]

I understand that all four hands were identical from deal 9 to deal 10 except that they were shifted one position counter-clockwise. Strictly speaking this makes the two deals not identical. I also understand that the similarity between the two deals was only detected by one player during auction and play, and that it was a one-time only occurrence.

 

If this is correct I would let the obtained results stand, but cause the computer program in question to be suspended pending further investigations.

 

And I believe this is exactly what I wrote on these circumstances in my first post (which was of a more general nature).

[helene_t]

One possibility is to treat this is analogous to bad shuffling. Say some people had been playing rubber bridge (i.e. stacking tricks instead of each playing keeping his cards before returning them to the deck), and subsequently the cards were dealt clockwise with a simple "cut" instead of shuffling, then the same hands will appear, and if one of the rubber players then turn up at the club night he will recognize the hand. There must be clear rules for what to do in that case (I would think that the board is not fouled).

 

Another possibility is to treat it as analogous to the director mixing up duplicated boards, in which case the boards are fouled I suppose, although maybe only board 10 should be fouled as per Blackshoe's argument.

 

I agree wirth everything Pran said but finding out how it happened takes time and the TD must make a decision in the meantime. One could argue that if it turns out to be a software bug it is analogous to bad shuffling while if it turns out to be an operator error it is analogous to the TD mixing up the boards, but IMO it wouldn't be reasonable to make that distinction.

 

FWIW I think if it's really true that the software produced the same hand twice, it is either due to an operator error (a typo in some configuration file or w/e) or some very old software that uses 32-bit entropy sources. With a modern dealing program, identical boards are not supposed to happen.

[jeremy69]

When computer deals were first introduced at the club I play at there was an interesting variant in that the progran produced the same board for board 3, 13 and 23. There was pandemonium when it cam eto light and the program was put into the garage until properly checked out and I wouldn't have thought it that likely for something similar to happen now. However in the given situation I think one should not interfere with the hand nor indeed if it produced two identical non rotated boards. One day when the computer deals us all 13 of a suit the hand will never get played becuase someone will go "ha ha" and give the hand away. It did happen in a European Womens Championship but the reason was a prosaic one of it not having been dealt at all. It didn't stop the TD asking the women concerned "Has it been played at the other table?"

I agree with Pran that if this happens the program needs suspending and checking out at the first available opportunity. I don't know all that much about the programs that deal hands but if

With a modern dealing program, identical boards are not supposed to happen.

If identical deals are thrown out by the computer then isn't it's owner in breach of Law 6E?

[gnasher]

When a director makes a decision as to facts, he can never have absolute certainty. The chance that a situation like this occurred randomly rather than as a result of software or human error is vanishingly small. If the director assumes a dealing error, he's far more likely to be correct than in any other determination of fact that he makes in his career.

 

So:

- For people who played board 9 before board 10, board 10 breaches some part of Law 6, and the results for board 10 should be cancelled.

- For people who played board 10 before board 9, board 9 breaches some part of Law 6, and the results for board 9 should be cancelled.

- He should now redeal one of the two boards, whichever makes most sense. If everyone played board 9 before board 10, he redeals board 10, obviously.

[mjj29]

With a modern dealing program, identical boards are not supposed to happen.

If identical deals are thrown out by the computer then isn't it's owner in breach of Law 6E?

Depends how the dealing program works, I very much doubt that it actively throws out duplicate deals. Sven probably meant to say "unless the random chance of identical boards actually occurs then ....", however, it is possible that the random number generator is such that it can't produce the appropriate repeated sequences in the same run. This would of course be bad and, as you say, in breach of 6E, but I feel only a technical breach that doesn't actually matter in practice. I know my program does it right, but it would not surprise me to find all sorts of biases and non-random behaviour in many popular programs.

[gordontd]

I wouldn't have thought it that likely for something similar to happen now.

Actually it does. If the dealing program that comes with the Duplimate machine crashes while in use, it quite commonly re-deals the same board as the next board's number. So, for example, Board 19 & Board 20 will be identical (apart from the dealer and vulnerability). It's something that needs to be checked for whenever there's a crash of the program.

 

If the movement were a Howell, and especially if the two adjacent boards were not part of the same board-set, it would be quite possible for it not to be noticed for a long time - conceivably the whole event.

 

I have no idea how the second board might have been rotated before being played though.

[blackshoe]

In what way did my post not answer the question?

 

Sven has a point in that whether and how the software was at fault bears investigation, but none of that can have anything to do with the table ruling, as there isn't time to do the investigating before ruling. Certainly very few directors will have Sven's expertise in this area, and from his post even he will take considerable time to complete his investigation.

[suokko]

Odds for generating same bridge hand twice ever in once life time is so small that it should never happen. (number of different deals is 53,644,737,765,488,792,839,237,440,000)

 

But normal random number generation in computers is deterministic so if you have same seed for generation process you will get same boards. This same problem is huge in cryptographic key generation where problem is solved by providing random bits from truly random sources like user input timing, timing of hard driver seek operations etc.

[nige1]

I wouldn't have thought it that likely for something similar to happen now.

Actually it does. If the dealing program that comes with the Duplimate machine crashes while in use, it quite commonly re-deals the same board as the next board's number. So, for example, Board 19 & Board 20 will be identical (apart from the dealer and vulnerability). It's something that needs to be checked for whenever there's a crash of the program. If the movement were a Howell, and especially if the two adjacent boards were not part of the same board-set, it would be quite possible for it not to be noticed for a long time - conceivably the whole event. I have no idea how the second board might have been rotated before being played though.

Gordon's explanation convinces me. Although the rotation does seem peculiar -- I doubt whether there was a rotation -- Sorry Bluejak :( -- I think it was just a different dealer and vulnerability, as Gordon implies. In principle, however, there may be nothing untoward. Obviously, it is just as likely that the machine's next deal is the same as the previous deal as that it is any other specific deal. I think the director should let other results stand and warn the vigilant player, in future, not to voice his suspicions until play is complete.

[blackshoe]

When we are given the facts of a case here, it is folly to waste time arguing about whether those facts are accurate, unless you were there.

 

On what legal grounds do you issue such a warning, Nigel? Did he do something wrong?

[pran]

I wouldn't have thought it that likely for something similar to happen now.

Actually it does. If the dealing program that comes with the Duplimate machine crashes while in use, it quite commonly re-deals the same board as the next board's number. So, for example, Board 19 & Board 20 will be identical (apart from the dealer and vulnerability). It's something that needs to be checked for whenever there's a crash of the program.

This is a shocking allegation.

 

I have no business with Duplimate systems (other than that I at times am asked to deliver random dealt deal files to be processed by Duplimate owners), but if the quoted allegation is correct the program should be suspended immediately.

 

And what does it take to test a card dealing program? A set of 8000 deals run through my testing procedure will give an initial answer within seconds. More such sets created independently over some time and tested will usually be sufficient to provide a definite answer, but preferably the basic random generator algorithm should also be available for a separate test as this is the kernel of any such system.

[pran]

Odds for generating same bridge hand twice ever in once life time is so small that it should never happen. (number of different deals is 53,644,737,765,488,792,839,237,440,000)

 

But normal random number generation in computers is deterministic so if you have same seed for generation process you will get same boards.

The important point is that with an identical seed the entire run will be identical to the previous, there will not be just one identical board.

[gordontd]

This is a shocking allegation.

I'm not sure why you are shocked by the idea that a program crash will produce uncertain results that need to be checked.

[suokko]

This is a shocking allegation.

I'm not sure why you are shocked by the idea that a program crash will produce uncertain results that need to be checked.

Because they shouldn't! It is bug that program crashes in first place. There is 2nd bug that program doesn't handle crash correctly.

[McBruce]

Surely when you ask the dealing program to produce a set of deals it will either respond "DONE" or send you some error message. If the former, you take the file to the machine and start the dealing of the actual cards; if the latter, you regenerate a new set of hands, don't you?

 

"Computer-dealt hands at a club" and "the computer dealt the same hand twice" can mean several similar but different things. I'm assuming from David's post that the hand records for the set included the two consecutive deals with all four hands the same but rotated. I think in that case these are two different deals (albeit remarkably similar) but agree that the coincidence strongly suggests that the software needs to be looked at.

 

But it could also mean that the computer dealt the same hand twice and there was an error by the person who took the cards from the dealing machine and put them in the slots. Once you add the human factor there is no limit to what can happen. I was called to the table in a Swiss Teams (player-dealt hands) recently to find a player debating with himself whether AJJ97 was a biddable spade suit. :(

[jeremy69]

If you haven't got time to investigate (and finding out who played board 10 before 9 may not be so easy) then

1. You will beat up the software (and/or it's owner) after the event

2. You will remonstrate with the player if he announced it in such a way as to poison the board for others (but not if he told you quietly)

3. You will direct that the boards continue to be played because you have no evidence that this is other than a correct deal (although I agree probability is on the side of suspicion).

[mjj29]

And what does it take to test a card dealing program? A set of 8000 deals run through my testing procedure will give an initial answer within seconds. More such sets created independently over some time and tested will usually be sufficient to provide a definite answer, but preferably the basic random generator algorithm should also be available for a separate test as this is the kernel of any such system.

Seriously, how hard is it to write a random board generating algorithm. Every programming language has a library for getting good (secure) random numbers, plug that into a Knuth shuffle and you're done... Seeding it isn't a problem either, the platform should have a way of randomly initializing the seed each time.

 

Out of interest what's your testing procedure?