result calculation

[GMarco24]

Sry for my bad english, but i hope you will understand...

 

You should try this:

in the middle of the hand you are recieving message : "Goodbye, tralala... to tournament."

Claim all yours. CPU will accept it.

Now you can see your result of 6NT-9, and result will make 6NT= sooooooooo goooooooooooood, when there is cold 7NT.

There are players who are beginners, and they should play, but some players suffer big injustice because of them.

If we 'discard' 2 top results, and 2 bottom, we will get better aperture for calculation.

 

Thanks to all people on BBO for effort to make us satisfy

[barmar]

Players who do this should be banned. Also, the director can adjust the score of boards like this. It doesn't happen enough to require changing how scoring is done.

[GMarco24]

Maybe I was not quite clear...

I wasn't talking just about this funny claim. I'm talking about all results on BBO. did you noticed that you are never bad if you made six? with 38 pts??? always two pairs plays 1 club or something like that. On other side, there are some pairs which made four where I was proud on myself for making two, and I'm not such a bad declarer. I just suggest that we should count score with 2 or 3 or ,I don't know how many results, overruled for general aperture

[Gerardo]

BBO uses CrossIMPs (just like MPs are calculated, comparing all results again any other result), NOT Butler (optionally take the extreme results, take the average, compare against the average).

 

There is no sensible way to take results out using this method. This method is sensitive to the size of the sample (the more, the merrier. An exterme result would have less weight).

 

[opinion]

Having a computer to do the calculations, CrossIMPs seems a much better method than Butler.

However, the sample size in BBO is very small (in MBC).

[/opinion]

[helene_t]

I don't see why there's no sensible way of taking extreme results out of XIMP-calculations. Just IMP the table result with all aother tables except for the top two and bottom two. What's wrong with that?

 

But because XIMP is more robust than Butler, the gain from robustifying it further by taking out extreme results is smaller and it's not clear if it would be an improvement.

 

[Opinion]

I'm not so concerned about crazzy results in XIMPs. After all, the law of large numbers applies, and there's virtually no bias.

 

MPs is different. MPs is supposed to reward overtricks, improving 6M to 6NT etc. But if a substantial part of the tables got crazzy results, safety-play becomes more attractive. I would prefer BAM with four GIBs at the other table.

[/Opinion]

[barmar]

If you exclude the extreme tables when cross-IMPing, how do you assign scores to those tables? Do they still get IMPed against the other tables? If you do this, the scores won't add up correctly.

[hotShot]

If you exclude the extreme tables when cross-IMPing, how do you assign scores to those tables? Do they still get IMPed against the other tables? If you do this, the scores won't add up correctly.

The idea would be to to use a weighted average.

 

BBO XIMPS are:

 

Sum of all IMPs / (number of scores -1)

 

If you use (16 scores / board):

If your score ranks 3-14

Sum of IMPs( position 3-14) / 10 + weight * (Sum of IMPs (Position 1-2, 15-16) / 4 )

 

If your score ranks 1-2 or 15-16:

(Sum of IMPs (Position 1-2, 15-16) / 3 ) + another_weight * ( Sum of IMPs( position 3-14) / 11)

 

Now you select "weight" and "another_weight" in a way that the scores add up correctly.

[helene_t]

That's true but you have the same problem (if it is a problem, I don't really see why) with butler.

[hotShot]

That's true but you have the same problem (if it is a problem, I don't really see why) with butler.

If you play a team match and both tables play the same contract with the same number of tricks, both sides get 0 IMP's.

 

There is some sort of justice and beauty in this form of scoring, this justice is lost when using crossIMPS. CrossIMPs have the same sort of injustice that MP's have, you get a score that significantly depends on seating and the skill of players competing at different tables.

 

If you take the above board into the MBC and get 16 scores, although N/S and E/W played best possible, usually one side gets + imps and the other side - imps.

Which side you are on, depends only on your seating relative to pair responsible for the eccentric score. Of cause this noise, matters less if you play enough boards. BBO tourneys are way to short for this, and in the MBC there are often more than 1 or 2 strange results.

 

In a tourney you get about +-0.5 IMP's/board randomly, so playing an 8 board tourney you will usually win or loose about 1-2 imp to this randomness. The inaccuracy of this form of scoring means that in tourney results all positions in the midfield are completely random. Even among the winners and the loosers, the actual lineup of the top positions could just depend on seating.

[Al_U_Card]

In the MBC we usually have 16 results shown. Are those all the results played? Why 16? Is it a server thingy? If, say 32 results were used, how would that affect the "fairness"?

[uday]

It is a bbo-client thingie. This client can't show more than 16 results for a given board.

[hotShot]

The weight of a freak result out of 16 is 1:15.

The weight of a freak result out of 32 is 1:31.

So the effect of the freak score is only half of what it was.

 

So using more results would definitely be an improvement.

 

These links won't work forever, but they show the effect.

 

A traveler with 55 results, the average 2♠ contract gets 0 IMPs.

 

A traveler with only 12 results The average 4♥ contract gets -1.3 IMPs generating a difference of 2.6 IMPs.

[sceptic]

It is a bbo-client thingie. This client can't show more than 16 results for a given board.

 

if that is the case (and I could be talking out my backside here) why not score it over 32 hands and only show 16 of them?

[Gerardo]

With more hands, what's lost is the ability to see all comparisons in one look.

There is also a decision on how to show the results (if the best NS are shown, your own result might not be there at first look). Best would be to center it on your own result, unless it is at the very top or the very bottom, and let people scroll for better and worse results.

 

What's gained is accuracy, when the extreme results lost weight.

 

Don't know how hard would this be to implement, particularly taking in consideration Fred's very limited time (assuming it is desirable at all).

[GMarco24]

I think there is enough space to show results in two columns, if that is a problem. Let's calculate 20, and show 16 results, with top, and bottom two - out. Look, I don't know too much about XIMPs, but I'm convinced we can improve it on BBO. It's still same question: why I'm losing 3-4 imps on board where opps made 6 with 38 hcp?