[julie5607]

Hi Fred,

I am wondering if there is any way to minimize the distortion caused by crazy results on boards. Right now the results are compared to 16 plays, and if everybody is in a normal contract but one table, where something insane has happend, the results are terribly distorted by that one hand. Although I know that we are only playing here for fun, it is quite discouraging to bid and make a marginal game and find you are blown out of the water by another table playing a vulnerable 6cxx down 5.

Perhaps if there were a way to increase the number of comparisons that would minimize the damage, or to throw out results that exceed the norm by more than 1000 pts in either direction? Just a wish . . .

Julie

[luis]

Maybe depending on the number of results to be compared: N the "M" best results for NS and the "M" best results for EW can be eliminated before calculating the average.
If there're 0-12 results don't eliminate, 13-24 eliminate 2 results, 25-36 eliminate 4 results, etc......

[JRG]

Fred commented on this -- it's in the Online Help under FAQ.

He said that some tournaments get rid of some of the "extreme" scores at both the top and bottom because of the serious effect. At that time, he indicated BBO does not do this.

He said, at that time, that was the reason why it is against the rules of BBO to get ridiculous results on purpose.

Perhaps this thread will persuade him to reconsider implementing this adjustment to scoring. I hope so - I also find it annoying.

[uday]

I don't know that I see the value in this; yes, it will avoid comparisons against (say) 7N-13 at a table where N was annoyed at S, but what if the field was in a flat 3N and one pair tried for 6, generating a swing?

[Cascade]

I vote for keeping all results.

The results caused by 'non-bridge' stupidities are minimal.

And in the long run they will even out.

Wayne

[hotShot]

If we were plaing a team match, i would not bother abaut extreme scores. But if i'm playing with another pair at a table, they play a flat 3 NT and i get a minus score (instead of 0) because one tried a lunatic 6NT and failed 3 times with XX. I don't linke it at all.
Within Butler scoring it is usual to cut of one(or 2 if often played) topscores at both ends and using the rest for the average.
In the case i stated above the average score would be 3NT makin 0 for both sides and the Pair scoring the extreme Result gets it's IMPS against this average too. Which means more IMPS for them.
No result is lost, and everybody gets what he/she deserves.
I don't see how this can colide with BBO rules.
If an extreme score happens more than once it will get into the average this was to.
I can't imagine that this would create more or less extreme results. It's just that the table with the extreme result gets most of their own swing, while others do not.

hotShot

[QQSV]

Here's one vote for keeping things the way they are. Occasional extreme scores are a part of bridge. Usually the extremists will lose big; on occasion they will score a hit. Either way, that is what happens in real life matchpoint events too. Your nice average 3NT may gain a matchpoint over average (or lose a matchpoint) because someone did something silly or brilliant.

That is life.

If you consistently play sensibly, over time you will find yourself accumulating positive IMP scores. Forget about the occasional wild swing. Besides, every once and a while the crazies will swing against you and give you the undeserved 12 IMPs!

One question: with so many people playing IMP pairs, what would be the problem in increasing the number of plays per board (say to 32) for IMP pairs only? I would not object to that method of reducing the effects of one random swing.

--Q

[Erkson]

Quote

Here's one vote for keeping things the way they are.

I join my vote.

Quote

One question: with so many people playing IMP pairs, what would be the problem in increasing the number of plays per board (say to 32) for IMP pairs only? I would not object to that method of reducing the effects of one random swing.

Yes there are an almost continous flow of tournaments but most of the time with different players. Very few play 32 boards in a row. 32-board tournaments would see many players quiting before end for good reasons.

If it is not possible to have 32-board tournaments, at least play 16 boards and only ONE board at each table (not 2 as now) and tournaments limited to 20 tables each.
The problem with 1 board per table is slow play. It has solutions.

Erkson

PS : IMO, Abalucy tournaments go in that direction.

[inquiry]

Hi Erkson

I think the comment deals with the number of comparsions per board not the number of boards played. In tournments, the comparsions per board can be more than 100. But in the club rooms (non tournment), the comparisons are always 16 boards (each hand played 16 times). What was being suggest was increase the total comparisons to 32.

The only slight problem with this is timing. Hands are not moved to myhands until they are closed out (after the 16 comparisons), and so if one required 32, I fear there would be some hands played that never made it over to the myhands site. In tourney play, this is not an issue, as all table play the same hand at the same time. But still, I like the idea of 32 comparisons as being twice as good as only 16...but on the other hand, I don't let the imp score bother me (good thing too, all those huge losses), so getting -3 imps for a perfectly normal 3NT bid and made (which should be 0 imps) does not upset me. I just note that I did the "right thing" and move on to the next hand.

Ben

[Erkson]

Quote

...
I think the comment deals with the number of comparsions per board not the number of boards played.
...

Yes, Ben, and now I understand the "32 plays per board" that I didn't.

Erkson

[DrTodd13]

Right now, I believe that the score deserving of 0 IMPs is determined
by taking the average of all the scores. Remember back to your
statistics courses and you'll remember the "mean, median, and mode"
discussions. It seems to me like some of these ridiculous results
wouldn't matter so much if you used the median to compute the
0 level rather than the mean. The crazy results (especially where
there are only crazy results in one direction, which is very common)
heavily affect the mean but wouldn't affect the median as much.

Thoughts? Has anybody done any simulations on this? Not having
a ratings system, I think such an approach might work whereas if
we had a ratings system you would need someway to keep the
IMPs awarded per board to sum out to 0. I think this is the case
that ratings could continually grow if each board on average had
positive IMPs awarded. Maybe you could scale everyone at the
end of processing ratings to an artificial mean...but I digress.

Todd

[hrothgar]

I've seen plenty of bi-modal distributions in my day
can't recall a bi-meanal, however :-)

BTW, could you resend me the link for the proxy
thnx

[DrTodd13]

http://www.htthost.com