[Dinarius]

Hypothetically, your club’s President’s Prize for Matchpoint Pairs is taking place over three nights.

Each night there will be N/S 1,2,3 & E/W 1,2,3, etc. obviously.

But, after the three nights, there will only be one winning pair.

I’m asking the question, because I’ve come across numerous approaches to separating the competing pairs.

They have been as follows....

1. Each night, the first six places N/S and E/W are each awarded points 6 down to 1 in descending order. The pair with the most points after three nights wins.

2. Each pair’s % score is averaged over the three nights.

3. Each pair’s % score is aggregated over the three nights.

4. One of either 2 or 3 above, BUT, a pair is allowed to discount their worst night.

a. My instinct is that % is best. You’re playing for a given number of matchpoints over the three nights. So, if you win the highest %, you should win the competition. But, if this is correct, which is better; average or aggregate %?

b. Is there any sense in being allowed to discount a night? If so, should the remaining two nights be averaged or aggregated, or something else?

c. I presume that whatever is best, it would apply over 4/5/6 nights also?

d. Is there an alternative method I’m not considering?

Thanks.

D.

[DozyDom]

Unless my mathematical knowledge has deserted me in the last few weeks since A-levels, average and aggregate scores will give exactly the same ranking no matter what.

As for discounting the worst night, it seems like it wouldn't make a huge difference to anything, but it might lead to less technical bridge being played on the third evening - those lagging behind in the competition might be encouraged to create swings to get ahead, knowing that if the swings don't go their way it won't affect their standings. Not sure if that would actually happen though. Seems more like a matter of taste than anything else.

Giving people points for their placing seems more likely to create draws, which seems undesirable.

[paulg]

My local club awards lots of trophies for competitions (https://berwickbridg.../trophy-winners) and we score them all differently The scoring of the principal pairs event is taken very seriously by the membership and it is an issue often raised at our AGM, but I hope we are in a more stable position currently.

Three evenings is less than all of our events but I suggest that you run single-winner movements on each evening as this will reduce the possibility of ties. We typically use placing points for competitions rather than average percentage because if someone gets 70+% in one session then people tend to lose interest. For our principal event we use what we call reverse placing points: so if 20 pairs play, then the winners get 20 points, second gets 19 etc. This prevents people scoring a normal win on an evening when there is a severely reduced field (normally due to snow in the winter).

If you allow people to discard their worst score when they play on all three nights, this may encourage more people to play on the final night.

Finally, whatever choice you make will be considered wrong by at least 33% of the pairs

[dsLawsd]

Percent makes sense with the following idea- award carry-over points on each of the first two events while doing your best to have everyone play each other pair at least once.

Then on the final day score the event using the carryover in addition to the raw score in order to get overall for the pairs. For a large number of pairs it is likely to require a Mitchell movement, but for a smaller group a Howell type giving one winner would work.

The conditions of contest would require playing all 3 days.
It would be nice to award more than one trophy for a large group up to several places with newer players having their own section winner.
Unfortunately we cannot keep luck from being a factor, can we?

[sfi]

If the event is considered serious, I would expect:

• All sessions to count. Something like best 2 out of 3 can skew the last night actions quite a lot, and throwing some of the results away is more suited to a "fun" event rather than a club championship.
  
• All boards to count equally. So percentage is fine if you play the same number of boards each session, but not if the number changes. In that case, use raw matchpoints and convert to a percent at the end (or factor the results).
  
• To try and ensure all pairs play each other.

Certain numbers work well for this. For instance, if there are 11 tables, you can have a 7-table Mitchell and a 4-table Howell. Each night you play 28 boards. The second night, N-S in the Mitchell plays in the Howell (with one stationary pair who stays in the Howell all event). The third night, the original E-W play.

Another option is to have two weeks of qualifying and a one-week final. The top half (or whatever number is suitable) make it to the final while the rest play off for the plate. You can have a carry-forward if you like - we tend to use 50%, but it does vary depending on the whim of the tournament chair.

[weejonnie]

The trouble about combining averages is that it can give misleading results - see this link

TBH I think any reasonable and consistent method would be OK - you just put it in the CoC.

[barmar]

weejonnie, on 2018-July-10, 08:19, said:

The trouble about combining averages is that it can give misleading results - see this link

Simpson's Paradox tends to occur when the sample sizes are significantly different among the results that are being combined. If you have a decent size field and they're about the same size each night, it shouldn't cause too much trouble here.

If the attendances are very different, maybe there's some way to weight the games to mitigate this effect.

[Vampyr]

Can you not separate the players into three lines and run the competition as an all-play-all?

[P_Marlowe]

Hi,

I am not a TD, but my feeling is, that your first option is the most practical.
If you have a club event running several days, chances are, that several pairs wont show
up every day, this problem gets larger, if the number of days increases.
The method allowes, that pairs, that did not make on the 1st day, can still enter, and
may even fight for the win.
But I would go with something like

10 for 1st,
7 for 2nd,
5 for 3rd
4 for ...
3
2
1

This gives a bonus a finishing on top.

With kind regards
Marlowe

[fromageGB]

paulg, on 2018-July-09, 13:18, said:

..so if 20 pairs play, then the winners get 20 points, second gets 19 etc. This prevents people scoring a normal win on an evening when there is a severely reduced field (normally due to snow in the winter).

Or if Scotland gets to the world cup final. (For those not in the know, Berwick is part of Scotland, not England.)

This method seems to me the best of those here, and is akin to the local points earned. Split any tie by total percentage.

[dokoko]

dsLawsd, on 2018-July-10, 03:01, said:

Unfortunately we cannot keep luck from being a factor, can we?

Is it really desirable to eliminate luck? IMO if luck wasn't a factor many ppl would drop out as they would constantly be confronted with the fact that they are bad players.

With luck being a factor, even bad players sometimes have at least so-so results and you can accredit good results to your skill and bad results to your luck. Wouldn't bridge be sad for many without that?

[paulg]

fromageGB, on 2018-July-21, 09:27, said:

Or if Scotland gets to the world cup final. (For those not in the know, Berwick is part of Scotland, not England.)

This method seems to me the best of those here, and is akin to the local points earned. Split any tie by total percentage.

Berwick-upon-Tweed is in England, although the bridge club and football team are affiliated to the Scottish authority across the border despite the home club/stadium being comfortably in England. It is for historical reasons for the bridge club and such 'poaching' is not allowed these days by the EBL.

The scoring method is similar to master points earned, save that they are only awarded to the top third of the field.

[shevek]

I run a few of these events. In fact I am doing one now as I post this.

This particularly one is easy enough, since I know everyone will turn up for every session.
I have 15 tables, running Sat arvo, night, Sunday morning.
(It's a country congress with people driving for a few hours to get here)

I had 2 choices.
1. Round robin of 3 x 9 boards. Each pair plays 27 out of 29 opponents. Need a twin set.
T 1-9 play a Mitchell, can arrow-switch R9 but I don't bother.
T10-15 play a ¾ Howell, 3 stationary.
(If you have one set of boards, T1 can share with T10, etc)
The 3 stationary pairs stay there for the whole event, while the other 27 rotate through NS 1-9, EW, Howell.
(In fact I have keyed movements like these as one section, "user" moves)
This is what I chose. More social, 3-board rounds throughout. Going well ...

Best numbers for this approach are 10, 14 & 20 tables, pure all-play-all.
Each extra table means 2 pairs missed.
With 13 tables, 7T (7x4) + 6T ¾H means 21 of 25 opponents played, 84%. Even worse if 12½ tables, plus 4-board sit-out.
Likewise, 18-19 tables makes for a poor round robin.

2. A better approach - if numbers are uncertain - is to do 2 session qualifying, again aiming for all play all movements.
Take 13 tables, They can play a 13T mitchell, then a 13T interwoven Howell - all move, one repeat.
Top 10 pairs go to a 5T Howell final, rest to a "Plate".
Carry-forward as you wish, I do a third.
With 15 Tables I would play a web (Bowman-Ewing), then a twin Howell. 2-boards throughout.
Always keen to have all boards played by all pairs.

The problem comes with uncertain numbers for week two. Then you would need to adjust the movement but no big deal.