MP Scoring Query

[eagles123]

I don't know if this is universal, but sometimes in our results there will be a random 0.1 for getting a bottom and 0.1 less than full marks for getting a top: example

 

 

Pairs Contract Scores MPs

NS EW Bid By Ld NS EW NS EW

6 6 3NT+1 E 430 5.5 16.5

5 4 2S W 110 21.9 0.1

4 2 3NT+1 E 430 5.5 16.5

3 13 5HX-3 N 500 0.1 21.9

2 11 2NT+1 E 150 17.5 4.5

1 9 2NT+1 E 150 17.5 4.5

12 5 3NT E 400 11 11

11 3 3NT+1 E 430 5.5 16.5

10 1 2NT+2 E 180 13.2 8.8

9 12 3NT+1 E 430 5.5 16.5

8 10 2NT+1 E 150 17.5 4.5

 

 

How come on the bolded scores its not 22 - 0?

 

and whats the difference to something like this when it is 22-0:

 

12 12 2H N 110 18 4

11 10 1NT N 90 11 11

10 8 1NT N 90 11 11

9 6 1NT N 90 11 11

8 4 1NT N 90 11 11

7 2 1NT+1 E 120 0 22

6 13 2H+1 S 140 22 0

5 11 1NT N 90 11 11

4 9 1NT N 90 11 11

3 7 1NT-1 N 50 3 19

2 5 1NT+1 N 120 20 2

1 3 1NT-1 N 50 3 19

 

obviously not a huge issue but just interested :)

 

thanks,

 

Eagles

[Zelandakh]

This happens when there is a sit-out table and the board is not played as many times as it should be. Incidentally that is probably also the reason that 13.2 is not 13 - the slight amendment applies across the board, not only for tops and bottoms. It is simply more noticeable there.

[eagles123]

Thanks!

[ArtK78]

Z may be correct in describing the reason for the peculiar result, but in my opinion the scoring program is wrong to do this. There is no reason why normal matchpointing should not apply just because there is an extra EW pair that did not play the board. In my opinion, the correct scoring of the board should be:

 

 

Pairs Contract Scores MPs

NS EW Bid By Ld NS EW NS EW

6 6 3NT+1 E 430 5 15

5 4 2S W 110 20 0

4 2 3NT+1 E 430 5 15

3 13 5HX-3 N 500 0 20

2 11 2NT+1 E 150 16 4

1 9 2NT+1 E 150 16 4

12 5 3NT E 400 10 10

11 3 3NT+1 E 430 5 15

10 1 2NT+2 E 180 12 8

9 12 3NT+1 E 430 5 15

8 10 2NT+1 E 150 16 4

 

Note that the board was played 11 times, so top on a board should be 20 (for those in North America, the board is scored on a 10 top and then the scores are doubled to eliminate half points).

If some boards are played 12 times, then this board has to be "factored up" to a 22 top so that each board counts the same. This is accomplished by multiplying each score by 22/20. In that case, the scores should be:

 

6 6 3NT+1 E 430 5.5 16.5

5 4 2S W 110 22 0

4 2 3NT+1 E 430 5.5 16.5

3 13 5HX-3 N 500 0 22

2 11 2NT+1 E 150 17.6 4.4

1 9 2NT+1 E 150 17.6 4.4

12 5 3NT E 400 11 11

11 3 3NT+1 E 430 5.5 16.5

10 1 2NT+2 E 180 13,2 8.8

9 12 3NT+1 E 430 5.5 16.5

8 10 2NT+1 E 150 17.6 4.4

 

This is classic matchpointing. I have never understood why computer scoring uses some other method which results is peculiar scores like a "top" of 21.9 and a "bottom" of 0.1. If you get the best score on a board, without any ties, why should you get any score other than 100% of the available matchpoints?

[StevenG]

There is a formula for multiplying up, when boards have not all been played the same amount of times.

 

http://en.wikipedia.org/wiki/Neuberg_formula

 

Results are flattened slightly leading to the 0.1 scores.

[PeterAlan]

As Zel has said, it's what you get when there is a sit-out, but it also depends on you having the Neuberg formula in operation - this is a method of adjusting scores to a common matchpoint top when different boards are played by different numbers of pairs (the sit-out instance is the commonest special case of this).

 

So-called "normal matchpointing" was a computational convenience in the days of hand scoring, but with computerised scoring there is little reason not to use the Neuberg approach and, as the article explains, this is increasingly what is adopted. There's usually a parameter in most scoring programs to switch between the two approaches.

 

Edit: I see this has crossed with StevenG's reply - since you're UK based, I've given you a link to the EBU's (Max Bavin's) article on the topic.

[aguahombre]

In the old days a difference of less than .25 was considered a tie for that position. It would seem that today a difference in final score of .01 (or some number which could be attributed to rounding) should be considered a tie.

[gordontd]

In the old days a difference of less than .25 was considered a tie for that position. It would seem that today a difference in final score of .01 (or some number which could be attributed to rounding) should be considered a tie.

Don't we want to split ties, not create more of them?

[mycroft]

The reason the 22/20 factoring method (which is what was used pre-computers to do formulas) is flawed is that who knows if the "phantom" 12th pair to play it has the system to find 2210 on that hand that everybody else in the room got 1460 on - or conversely, if the defence is good enough/declarer is bad enough to go -800 on the hand that everyone else is either -500 or -620 on? There's a non-zero chance of this.

 

Take it down farther. Assume all boards are played 12 times, so top is 22; but one board had to be replaced halfway through because of fouling, and it was played 6 times (as was the unfouled board). Now top on the board is 10; now we have to multiply the scores by 22/10. Seriously, is there not a chance that the top (and there are two of them, one for each fouling state) isn't going to get beat if it were played another 6 times? So, shouldn't your "top" reflect that chance?

[barmar]

This can also happen if the board is fouled part-way through the session. The rule for scoring a fouled board is to put the tables that played each version of the board into separate groups, and matchpoint them each separately. Then each group is factored up to the normal top.

 

With computer scoring, is there really any need for traditional raw matchpoints? I don't think they've ever been used in online bridge (at least not on OKbridge or BBO, the two sites I've used), which just reports percentages.

[aguahombre]

Don't we want to split ties, not create more of them?

Maybe. But, I don't think a .01 rounding fluke should decide a significant event.

[ArtK78]

The reason the 22/20 factoring method (which is what was used pre-computers to do formulas) is flawed is that who knows if the "phantom" 12th pair to play it has the system to find 2210 on that hand that everybody else in the room got 1460 on - or conversely, if the defence is good enough/declarer is bad enough to go -800 on the hand that everyone else is either -500 or -620 on? There's a non-zero chance of this.

 

Take it down farther. Assume all boards are played 12 times, so top is 22; but one board had to be replaced halfway through because of fouling, and it was played 6 times (as was the unfouled board). Now top on the board is 10; now we have to multiply the scores by 22/10. Seriously, is there not a chance that the top (and there are two of them, one for each fouling state) isn't going to get beat if it were played another 6 times? So, shouldn't your "top" reflect that chance?

With all due respect, factoring up the score on the board by multiplication to bring the top on the board up to the top on the rest of the boards is not flawed. It is mathematically accurate and absolutely appropriate.

 

The board was played with a field consisting of one less comparisons than the field the other boards were played in. There is no "phantom" pair, and any speculation about what the score would be if there were another comparison is just that - speculation. There is no need to account for it, and any method that does so is just making a guess with no substantiation.

 

The fact that anyone came up with some sort of computational algorithm that results in a low score of .01 and a top score of .01 less than the typical top is mere hocus pocus. There is no rationality to it. I accept that fact that the powers that be have chosen to use this scoring system when there are boards with different numbers of comparisons. I don't have to agree with it, and I don't have to like it.

 

I note that in match point pair games on BBO, each board is scored using the number of comparisons on the board and the scores are converted to percentages of the top score on that board. That eliminates the need for any other computational algorithms to compute matchpoint scores on different tops - each board is scored with a top of 100% to two decimal places. That works, and it is accurate to two decimal places.

 

As for fouled boards, Barmar correctly reports the procedure for dealing with fouled boards using manual scoring. And I agree with that method for the same reasons that I agree with the pure factoring up method. There is some other method used with computer scored fouled boards, and I have no idea what the rationale is. I know that, "in the old days," there was a different method for dealing with fouled boards at sectional and higher rated tournaments, even when scoring was done manually.

 

By the way, when factoring was done "in the old days" it was always factoring up - never factoring down. This was due to the fact that making the scores larger increased the chance of breaking ties due to fractional matchpoints resulting from the factoring. As agua noted, the old rule was that score differentials of less than .25 matchpoints were considered to be tied results. The .25 rule sometimes created really peculiar results. For example, suppose three pairs scored 145.60 MPs, 145.40 MPs and 145.20 MPs, and suppose these were the 2nd, 3rd and 4th best scores. Using the .25 rule, the top pair tied for 2/3, the middle pair tied for 2/3/4, and the bottom pair tied for 3/4.

 

[gordontd]

This is classic matchpointing. I have never understood why computer scoring uses some other method which results is peculiar scores like a "top" of 21.9 and a "bottom" of 0.1. If you get the best score on a board, without any ties, why should you get any score other than 100% of the available matchpoints?

Actually classic matchpointing as performed before computers were used for scoring would remove a whole matchpoint (half if you are American) from the top and add one to the bottom, so that the total matchpoints available on the board remained the same but you didn't have problems with tied results, as in this example:

 

+150____9

+120____6

+120____6

Ave_____5

-50_____3

-100____1

 

Neuberg scoring is just a more subtle implementation of this.

[ArtK78]

Actually classic matchpointing as performed before computers were used for scoring would remove a whole matchpoint (half if you are American) from the top and add one to the bottom, so that the total matchpoints available on the board remained the same but you didn't have problems with tied results, as in this example:

 

+150____9

+120____6

+120____6

Ave_____5

-50_____3

-100____1

 

Neuberg scoring is just a more subtle implementation of this.

All I can say to this post is "Huh??" I have been playing since 1972, and I have been an ACBL Certified Director since about 1976. I have never heard of this. And, frankly, it makes no sense.

[Trinidad]

All I can say to this post is "Huh??" I have been playing since 1972, and I have been an ACBL Certified Director since about 1976. I have never heard of this. And, frankly, it makes no sense.

I think Gordon meant to add the clause "if the number of times the board has been played is one less than the number of times that the other boards have been played" somewhere in his post.

 

Rik

[gordontd]

All I can say to this post is "Huh??" I have been playing since 1972, and I have been an ACBL Certified Director since about 1976. I have never heard of this. And, frankly, it makes no sense.

So how would you matchpoint a board with those six scores?

[gordontd]

I think Gordon meant to add the clause "if the number of times the board has been played is one less than the number of times that the other boards have been played" somewhere in his post.

 

Rik

Well that's what the Ave result is.

[campboy]

The fact that anyone came up with some sort of computational algorithm that results in a low score of .01 and a top score of .01 less than the typical top is mere hocus pocus. There is no rationality to it. I accept that fact that the powers that be have chosen to use this scoring system when there are boards with different numbers of comparisons. I don't have to agree with it, and I don't have to like it.

It's not hocus pocus, it's statistics. It's exactly the same reason that you should use Bessel's correction for the sample variance. If you don't like it, fine, but the mathematics is correct.

[ArtK78]

So how would you matchpoint a board with those six scores?

Sorry. For some reason I did not notice the Ave score that was in the middle. Yes, the classic way of adjusting a board with an Ave score is that the Ave score ties every other result, and the matchpoints shown are correct.

[fromageGB]

I'm happy with generated fractions because I look at it in a different way. I don't think one top is worth as much as another. If you played in a 4 player movement, and score 6 UK mps, by beating the other 3 pairs, then that's a top. But if you played a board against 5000 pairs and beat all 4999 of them, isn't this a more convincing victory?

 

I don't know what the Neuberg formula would make of this scenario, but it would probably agree with me. (Or I with it!)

[ArtK78]

I'm happy with generated fractions because I look at it in a different way. I don't think one top is worth as much as another. If you played in a 4 player movement, and score 6 UK mps, by beating the other 3 pairs, then that's a top. But if you played a board against 5000 pairs and beat all 4999 of them, isn't this a more convincing victory?

 

I don't know what the Neuberg formula would make of this scenario, but it would probably agree with me. (Or I with it!)

Your example is a bit extreme. In almost all instances, the difference in the number of comparisons will be very small. In a club game, it is unlikely to be more than one (except in the case of fouled boards). Online, there can be a larger difference in the number of comparisons, and the Neuberg formula is not used.

 

Maybe you would think differently if you lost a major event (or failed to advance to the next stage of a major event) by .01 matchpoint because of the Neuberg formula.

 

By the way, in worldwide bridge competitions which do not have fixed matchpoint results, gaining a world-wide top is typically the result of getting some huge number. There is very little practical difference between getting a top on a 5001 comparison board and getting a top minus .01 matchpoints when top on a board is 5000 (or 10000 European style). When the top on a board is much lower - say, 12 or less (24 or less European style), the chance that .01 matchpoints will break a tie is much higher, as the chance of a tie is much higher.

 

I have witnessed tie breaks between pairs for the final qualifying position in a 4-session pair event with the top 3 qualifying for the National finals (District North American Pairs - two sessions qualifying, two sessions final). It would be difficult to explain how the Neuberg formula broke a tie by arbitrarily awarding an extra .01 matchpoints to a pair that otherwise would have gotten a zero, or .01 less matchpoints to a pair that would have otherwise gotten a top. Of course, the tie-breaking procedures in use are also difficult to justify. Quite frankly, I don't know what they are. I believe that the first tie break is on a board a match style comparison if the pairs played in the same direction, but I am not sure about that.

[barmar]

In normal club games or f2f tournaments, I think it's unlikely that Neuberg would result in just .01 difference from normal matchpoints. Factoring between 11 and 12 tops (or 22 and 24 in British style) will result in multiples of something like .08.

 

Where you get smaller fractions that break ties are when you have multi-day events with carryover. The carryover typically contains fractional matchpoints, and if you also had to use Neuberg for some boards, the final results can be close by tiny fractions.