Forum   |   Links    


Forum Home   Start New Topic   Edit Profile   Register  

1   2   3  

New Ranking list

Show Profile  predragz Posted: 16 February 2005, 6:22 AM  
Why not use a national database (at least with M21E and W21E competitors), where each person is assigned a unique id? Then, Claire v Clare or Robert vs Rob would be avoided.

Show Profile  PaulS Posted: 16 February 2005, 10:01 AM  
Yeah and then I wouldn't get a Paul Smith from the early 80s appearing in my results.

Show Profile  Bryan Posted: 17 February 2005, 12:10 AM  
Here are the ranking points calculated for Marquita, Rachel, Neil and Darren. Basically, even though Rachel and Darren many times
beat Marquita and Neil, the others had events where they got better rankings because of how they ran against their fellow competitors for that race.

Date Class Event Ranking Best 4 Ranking
3/01/2004 W21E S2004D1 23.53
4/01/2004 W21E S2004D2 10.16
8/01/2004 W21E S2004D4 -4.99 -4.99
10/01/2004 W21E S2004D5 12.48
11/01/2004 W21E S2004D6 14.35
21/02/2004 W21E DT2004D1 9.66
22/02/2004 W21E DT2004D2 11.34
10/04/2004 W35A NZ2004L -17.6
12/04/2004 W21E NZ2004M 14.15
5/06/2004 W21E QB2004 8.48
6/06/2004 W21E QB2004 8.04 8.04
7/06/2004 W21E QB2004 6.42 6.42
30/10/2004 W21E AOA2004M 16.50
31/10/2004 W21E AOA2004 17.23
20/11/2004 W21E CDOA2004 6.82 6.82 4.0725

Date Class Event Ranking Best 4
3/01/2004 W21E S2004D1 6.48 6.49
4/01/2004 W21E S2004D2 9.31
8/01/2004 W21E S2004D4 65
10/01/2004 W21E S2004D5 11.05
11/01/2004 W21E S2004D6 9.26 9.26
20/03/2004 W21E SIC1_2004 6.13 6.14
21/03/2004 W21E C2004 4.84 4.84
9/04/2004 W21E NZ2004S 12.10
10/04/2004 W21E NZ2004L 11.69
12/04/2004 W21E NZ2004M 15.26 6.6825

Date Class Event Ranking Best 4
3/01/2004 M21E S2004D1 26.70
4/01/2004 M21E S2004D2 23.61
8/01/2004 M21E S2004D4 6.47 6.47
10/01/2004 M21E S2004D5 21.25
11/01/2004 M21E S2004D6 12.12 12.12
9/04/2004 M21E NZ2004S 20.62
10/04/2004 M21E NZ2004L 20.96
12/04/2004 M21E NZ2004M 25.97
5/06/2004 M21E QB2004 25.72
6/06/2004 M21E QB2004 18.79
7/06/2004 M21E QB2004 21.52
30/10/2004 M21E AOA2004M 32.17
31/10/2004 M21E AOA2004 22.12
6/11/2004 M21E WOA2004 13.12 13.12
7/11/2004 M21E WOA2004M 11.96 11.96
20/11/2004 M21E CDOA2004 32.66 10.9175

Date Class Event Ranking Best 4
3/01/2004 M21E S2004D1 21.15
4/01/2004 M21E S2004D2 20.05
10/01/2004 M21E S2004D5 30.72
9/04/2004 M21E NZ2004S 17.79
10/04/2004 M21E NZ2004L 16.37
12/04/2004 M21E NZ2004M 22.29
5/06/2004 M21E QB2004 28.37
6/06/2004 M21E QB2004 11.78 11.78
7/06/2004 M21E QB2004 15.25 15.25
30/10/2004 M21E AOA2004M10.41 10.41
31/10/2004 M21E AOA2004 10.04 10.04
20/11/2004 M21E CDOA200415.45 11.87

Marquita vs Rachel: 4.07 vs 6.68 (or 3.82 vs 6.43 when the best is zeroed)
Neil vs Darren: 10.92 vs 11.87 (or 6.32 vs 7.27 when the best is zeroed)

Show Profile  Bryan Posted: 17 February 2005, 12:27 AM  
To answer Micheal, the new rankings after the Oceania carnival will
be based on events only in that carnival (and maybe Otago champs if I can get the results). I used to produce rankings that spanned years but this caused problems with juniors/seniors who
changed classes at the beginning of the year. Usually I produce a new ranking list when there is enough events in a year to compile a meaningful ranking.

The ranking explanation can be a bit detailed but I think I've also already explained it in this forum before. Anyway, here's a quick explanation:

Rankings are calculated in a number of steps:
- Process A grades only
- Find the 3 lowest ranked runners running in this event
for each class where each person in a class is checked against the ranking points file.
- Calculate the rankings for each competitor using statistical formulae.
- Calculate an average for a ranking period
- Set number of counting events to be equal to 4
- Minimum of 2 events
- find the lowest average ranking points for each class. This may be negative. Subtract this value from all ranking points for the class. This will mean that the highest ranked person in a grade will have a ranking of 0.

Show Profile  Bryan Posted: 17 February 2005, 12:43 AM  
To answer predragz and PaulS, your suggestions are useful but difficult to implement. To attach a unique identifier to each competitor requires a match from each event against a valid list of correct names (this has to be compiled first which would take a lot of time) and then an exception list generated every event which shows new names or possible mispellings. A lot of work. I almost have something like this already.
In every year in New Zealand Orienteering, approximately 20% of competitors are new or are lost from our sport. This would make maintenance of a unique name database difficult.

Until the NZOF require registration with firstname and secondname
(instead of firstname only), the problem with different Orienteers with the same name will remain unsolved. I've fixed this before for some names by changing the name of the older competitor (not competing anymore) to a nickname or something slightly different upon personal request from the newer Orienteer. In events in the first few years, I only had an initial and surname from the results and it was a difficult exercise to match a firstname to these names. Another problem with matching names is that Orienteers change classes, clubs, nicknames frequently (and sometimes surname), and also organisers mispell names frequently and its difficult sometimes to determine the correct name or the correct unique person.

Show Profile  Jamie Posted: 17 February 2005, 1:29 AM  
Hey Bryan,

Would it be possible to have the elite ranking list on a continuous basis, rather than a seperate one each year?

The problems you suggest don't seem to apply to the elite grade.

Having a continuous ranking list would allow for greater accuracy particularly at the beginning of each year.

I also understand that at present only the best four results count, if the ranking list was continuous we could perhaps lift this to 8 over say a two year period, ie after two years a result would drop off.

Anyway, just suggestions.



Show Profile  mark Posted: 17 February 2005, 2:04 AM  
you elites are always asking for special treatment.

Show Profile  Michael Posted: 17 February 2005, 3:37 AM  
Bryan wrote...
> The ranking explanation can be a bit detailed but I think I've also already explained it in this forum before.

Yes I know. But unless you put it on the website you'll be forever answering this question:-)) The simplified answer was quite enough for me, I don't need the actual maths.

Spanning years: is it really such a problem, you already cope with people running different classes in the same year, and you give them a ranking in each. I'm in 3 classes for example.

Number of events to count: I would speak against Jamie's 8. The bigger the number the longer it takes for a change in an orienteer's form to show up. While 4 is arbitrary, I note the IOF ranking scheme uses 4.

Show Profile  Chris Forne Posted: 17 February 2005, 6:43 AM  
How about placing a weighting on the ranking points depending on how old they are? So, rather than calculating an average over each year, calculate a weighted average on a continuous basis.

Ie, you could reduce the weighting by 5 percent per month, so after about 1 1/2 years ranking points will no longer contribute towards a persons overall ranking. This way an event 1 year ago will still contribute, but only at about 40%.

An approximate way to implement this would be to simply add 1 to all event ranking points every month. So after 1 year an event ranking of 0.0 would become 12.0. Other than this the rankings would be calculated as before. In this way if say Carsten was not to compete for a year he would slip back to about 9th place, and after 2 years of not competing would be back at about 18th place based on the current ranking points. A slightly larger number could be added if this fall off was to be increased.

Show Profile  Jamie Posted: 17 February 2005, 3:44 PM  
Michel - form is temporary class is permanent. As the ranking list stands it is too affected by individual events, eg oceania will completely change it. Someone shouldn't be able to go from bottom to top in one carnival. Having 8 events counting would change this as well as rewarding those athletes that are consistent rather than having 4 good results out of say 12.

Show Profile  Bryan Posted: 18 February 2005, 12:36 AM  
Lots of feedback and I'll try to answer all.

I've put a quick explanation on the website.

I'll modify a report tomorrow so people can see the points earned for an event.(use:

Spanning years for young and old classes: It is a problem with a very good runner moving up, or down. Eg A few years ago, when I spanned years, Ross Morrison (who was now competing in a higher class) was always appearing at the top to the disadvantage of the current runners in the younger class.

Number of events to count: I agree for a ranking list elite classes need special treatment. I will now create a modified ranking system just for the elites. I could implement any ranking system but it
all takes time and I want to keep it simple and not stray too far from the current system for now.

I can make it continuous, and I can make the criteria stricter. Any objections on the following (to be implemented before the next ranking comes out:
- average of best 6
- minimum of 4 events
(current criteria is average of best 4, minimum 2 over the ranking period).

On the question of the ranking period we have two flavours:
1. A specific time period (eg last 2 years)
2. Weighted ranking - 1 is added to all ranking points every month.
I'd like feedback on the pros and cons before I decide.

Show Profile  Neil K Posted: 18 February 2005, 2:57 AM  
Excellent suggestions Bryan and others. I agree with 6 (min 4), 8 over a year would be too many (as most elites didn't do that many last year).

On the ranking periods. I see there are positives and negetives to both. I think a mixture would minimise the negetives of both. I suggest a period of 6 months were the ranking is worth 100% then once it is six months old a declining weighting is introduced reducing it to zero after two years.

Show Profile  addison Posted: 18 February 2005, 4:08 AM  
Hi Bryan.

The weighted ranking seems ok, but it does not take into account the fact that our orienteering season is not set, and therefore all of a sudden at big carnivals the rankings will all be low then before the next event everyones will let slip. Smaller events that not everyone can go to will severly alter the rankings. I reckon stick with a set period as it is fairer.

Any idea on when the oceania results will be up on stats site?

Show Profile  Michael Posted: 18 February 2005, 4:54 AM  
I'm a bit nervous about "decay" of results with time. Lets say we have some meaningful rankings, a period with no events will cause them all to decay in unison, then an event with low attendance will cause some "fresh" rankings to overstate the ability of the few who ran relative to the others. I think it would be sounder to have a moving "window" of (I think) 12 months.

Re the number to count, I would like a lowish number so the ranking is a better indication of recent change. I don't distinguish "form" and "class", a classy orienteer who is not in form should not be rewarded. How can an orienteer develop such tremendous form that (s)he can go from the bottom to the top in a single carnival, without demonstrating "class"?

On the other hand a reason for having a larger number is to iron out statistical anomalies which are always going to be there. I will defer to Bryan on this, and urge that the decision be made not on popular vote but on what you know about the stability of the numbers.

Show Profile  Jamie Posted: 18 February 2005, 5:10 AM  
Hey Bryan,

Excellent! i agree with the arguments re decaying ranking although think that statistically the argument is exaggerated anyway as the new rankings are always relative (I could well be wrong on this). I would go for the two year period but maybe we can settle for 18 months?

Michael - as it stands an orienteer with particular strengths be they technical/physical favoured re a specific kind of terrain ie woodhill, farmland may achieve a ranking that perhaps doesn't reflect their capabilities, having a larger number of events counting would minimise this

Anyway good discussion

1   2   3  

Ruffneck Productions © Ruffneck Productions