[Deanrover]

I'd like to see a form of IMP scoring that I am sure has an official name, maybe Butler, not sure of that.

When comparing your score in order to give the IMP result, I'd lke to see the "highest" and "lowest" scores removed in order to give an "adjusted mean". I think it would give a more accurate representation of the hand, and would remove from the equation some of the more eccentric results we see. I appreciate that over the long run, you are as likely to be lucky as unlucky in your comparisons, but this would be beneficial in the short run, thus ideal for tourneys (depending on the size of which more than 2 boards would be removed).

[1eyedjack]

This has been discussed a few times in these forums. I for one prefer the existing method.

Have a look at

http://bridgebase.lunarpages.com/~bridge2/...=2835&hl=Butler

to name just one thread (I only picked that one because I had some input to it).

[Cascade]

I think that is called Butler.

My understanding is that the Statisticians prefer cross-IMPs as the best method.

[ciscokid]

What about having more comparisons - 30+ vs. the 15 that are done now?

-CK

[pclayton]

This is one of my pet peeves too; and something that I have debated a lot with a very active member of BBO that does not visit the site anymore.

BBO only compares 16 scores. OKB is a lot more - +/- 50-60 if I remember. You should never have a situation where you reach the optimal contract and lose IMPs, only because a few of the pairs post numbers like -2800 or +2320. Playing the board more will even out the real 'swingy' results.

Either have increase the number of times a board is played (shouldn't be that difficult as hundreds of tables are in play during a given hour). 16 might have been a good rule of thumb when BBO had 20-30 tables in play at once, but not now.

Or (an idea I like less) is to either cap the effective IMPs on a hand, or cap the maximum point score for a certain result, as it relates to IMPs.

[paulhar]

Just out of curiosity, if you're playing in the Main Bridge Club, do the other 15 results all come from the Main Bridge Club or can some of them come from the private and public clubs?

[Dwayne]

Ideally you should be removing the top and bottom scores

And more sensically in a tourney, only taking scores from, say, the top 10 tables after round 1.

Dwayne-man.

[helene_t]

Cascade, on Jul 26 2004, 11:51 PM, said:

My understanding is that the Statisticians prefer cross-IMPs as the best method.

Probably true. If you get +100 while a single +3200 "disaster" pertubates the mean from 0 to +200, you get -3 IMPs instead of +3. On the other hand, if you make +1100 it makes a much smaller difference if the mean is 0 or +200. Removing the extremes solves the problem to some extend but it also means that the mean IMPs made in the E-W direction is not necesarilly zero. Cross-IMPs is a more elegant solution to the problem.

[hotShot]

The IMP scale should make sure, that playing 3NT+1 or 4Major makes an imp difference of 0. An overtrick should make not much of a difference but a missed game or an undertrick should cost.

But if some major desaster moves the mean it can suddenly make a difference. This is why scoring to a mean requires cutting the extremes off.

With CrossIMPs the IMP's with every other result are calculated und you get the mean of them. This is quite fair but, if you play 10 boards with the same result that most others have, your CrossIMP score will not be 0. Usually it will be negative, because someone playing your side scored big.

How could that be improved?
Perhaps by shifting the calculated crossIMPs. e.g. If i have 16 results, i calculate the mean of the ordered crossIMPs 8 and 9 and shift all IMPs that way that this gets 0. That way we could have the "average result" with 0 IMPs and differences from cossIMPs.

[TimG]

hotShot, on Aug 12 2004, 09:03 AM, said:

With CrossIMPs  the IMP's with every other result are calculated und you get the mean of them. This is quite fair but, if you play 10 boards with the same result that most others have, your CrossIMP score will not be 0. Usually it will be negative, because someone playing your side scored big.

Usually negative? Every time it is negative for you, it is positive for your opponents.

[hotShot]

TimG, on Aug 12 2004, 02:06 PM, said:

hotShot, on Aug 12 2004, 09:03 AM, said:

With CrossIMPs  the IMP's with every other result are calculated und you get the mean of them. This is quite fair but, if you play 10 boards with the same result that most others have, your CrossIMP score will not be 0. Usually it will be negative, because someone playing your side scored big.

Usually negative? Every time it is negative for you, it is positive for your opponents.

Hi Tim!

Take a look at this simple tourneyresult:

Example Tournament

It shows 39 +IMPs to 45 -IMPs sum tham up and you get -6 IMPs.
Meaning a pair having the average result at every board would have a score of -6 IMPs.

You can look at other tourneys if you like, but it's the same (but don't get a survivor ...)

Playing the average result each of e.g.10 boards will result in a negative crossIMP score.

[inquiry]

Uhhh.. hotshot....

what you are talking about is a tounrments online where people fail to finish a board in time. When this happens, the software assigns both sides average minus, and as a result, the net imps will be minus... (the size of minus depends upon frequency of people not finishing a round).

If you look at an untimed tourment, or a team match, you will find, that in fact, Tim is absoluately correct.. For every imp one side gets, the other side doesn't get... and it averages out perfectly.

Here is a link to an random team game http://bbo.bridgebase.com:81/tourneyresult...17242%7E744.lin

So, if you find an untimed tourment, or a team game, or a tourment where no procedure penalties were issued... Tim would be 100% correct...

Ben

[TimG]

hotShot, on Aug 12 2004, 09:28 AM, said:

Take a look at this simple tourneyresult:

Example Tournament

It shows 39 +IMPs to 45 -IMPs sum tham up and you get -6 IMPs.
Meaning a pair having the average result at every board would have a score of -6 IMPs.

I'm not convinced that's what it means. Can you provide a link to the travellers for this tournament?

[hotShot]

inquiry, on Aug 12 2004, 03:10 PM, said:

...


Here is a link to an random team game http://bbo.bridgebase.com:81/tourneyresult...17242%7E744.lin

So, if you find an untimed tourment, or a team game, or a tourment where no procedure penalties were issued... Tim would be 100% correct...

Ben

Uhhh Ben

how can you score a team game with crossIMPs ?
(But you are right taking a tourney like that was wrong.)

The problem occurs only if you have more than one result.

Take a look at this board:
Example traveler

We have:
6 times 6♠ +1 1460 2.21 IMPs N/S
12 times 6♠= 1430 1.126 IMPs N/S
once 5♠+1 680 -11.63 IMPs N/S
once 7♠X-1 -200 -16.79 IMPs N/S

So every of the 12 EW pais that joined the 6♠= club leave the board with -1.26 IMP's for doing everything right.

If you cut of the 2 extreme scores they would have been rewarded with about +0.5 IMP for not allowing the overtrick.

Hope i made myself clear this time.

[TimG]

hotShot, on Aug 12 2004, 10:40 AM, said:

The problem occurs only if you have more than one result.

I think you mean more than one comparison. It's hard to IMP a single score.

Quote

We have:
6  times 6♠ +1 1460 2.21 IMPs N/S
12 times 6♠= 1430 1.126 IMPs N/S
once 5♠+1 680 -11.63 IMPs N/S
once 7♠X-1 -200 -16.79 IMPs N/S

So every of the 12 EW pais that joined the 6♠= club leave the board with -1.26 IMP's for doing everything right.

Every one of the 12 NS pairs that joined the 6♠= club leave the board with +1.126 for doing what was routine. I think this example is a perfect counter to your claim:

Quote

Usually it will be negative, because someone playing your side scored big.

This time the NS pairs who did the routine thing got a positive result and the EW pairs who did the routine thing got a negative result. I'm still at a loss for how it can be "usually negative".

[hotShot]

Well you see my point.

1) It is usually negative, if someone on your side scores big.
And you notice!

2) It is usually positive, if the big score is at the other side.
And you don't care why it's good.

[TimG]

hotShot, on Aug 12 2004, 12:37 PM, said:

Well you see my point.

Actually, it seems you have seen my point.

[1eyedjack]

hotShot, on Aug 12 2004, 10:40 AM, said:

So every of the 12 EW pais that joined the 6♠= club leave the board with -1.26 IMP's for doing everything right.

And each of the 12 EW pairs have scored the same as each of the other EW pairs who did the same thing. So no individual EW pair has gained or lost against any other EW pair who has done the same thing. So what is the problem?

In real life you do not score all your results against par. You score them against what happens at the other table. If you were playing a simple teams of 4 match, and you bid and make a routine 4S when your opponents go overboard for no reason and bid a ridiculous 6S - 2 at the other table, would you suggest that at your table you should just score a flat zero because you did the routine thing? No, you would would take the 10 imps from scoring your routine result against the other table's ridiculous result.

That is all that is happening in a cross-imped pairs scoring. You are just playing 15 or so separate teams matches.

[hotShot]

Well 1eyedjack,

do we have seperate results for EW and NS in tourney results here at BBO?

No, we don't.

24 Pairs played the "par contract" at this boards, but half of them gets +1.26 and the other half -1.26.
Now look at any tournament what a difference 2.5 IMPS does in the middle reagion.
Imagine a swiss movement, each group of 12 is seated against each other.
If they score the "par contract" there,
half of them will be rewarded with the "plus score"
- the other half wih the "minus score".
Pairs of the same strength will get different IMP-Scores.

This is unfair!

So one would have to create 2 results, as in a full Mitchel-Movement
or adjust the scale, so that the par of each board is leveled to 0.

[paulhar]

hotShot, on Aug 12 2004, 04:43 PM, said:

24 Pairs played the "par contract" at this boards, but half of them gets +1.26 and the other half -1.26.
This is unfair!

No, not unfair at all. The side that got +1.26 IMPs reached the par contract, something that some others their way didn't do. So, in effect, they risked being one of the pairs who didn't, and if they had been one of those, they would have lost mucho imps. This pair had to do something right to reach par, something that not every pair would do. So they should be plus IMPs since other pairs did not do that something right.

The other side lost IMPs for playing against somebody who did something right, the way it is in any bridge game. The total avaliable IMPs available to their direction is zero. Just because their opponents bid the par contracct, they lose a couple IMPs. If they had happened to play against one of the pairs who didn't, they would have won many IMPs.

Would you prefer a system where the side that bids par gets nothing while if they don't reach it, they lose a bunch (heads, I break even, tails I lose bigtime), while the other side gets +0 IMPS with a shot to win the lottery if their opponents mess up? Now, THIS system seems unfair to me.