Posted: 10 December 2007, 9:52 PM
1. Medeltiden (Tm) berÃ¤knas pÃ¥ de tre bÃ¤sta tiderna gjorda av rankade lÃ¶pare. FÃ¶r att fÃ¶rhindra orealistisk lÃ¥g segrarpoÃ¤ng vid alltfÃ¶r "tunna" klasser tillÃ¥ts dock medeltiden ligga hÃ¶gst 10 % efter segrartiden.
The Average Time is calculated from the three best times done by ranked runners. To prevent unrealistically low winning times in "thin" classes, the winning point is only allowed to be 10% after the winning time.
So it is the three best times but thy have to be ranked runners. This is actually relevant because you have to pay to be on the ranking list so not everyone does take part in it. You only do if you think you'll want to run elite in races which will have more entered than the 60 or whatever is normally allowed. And sometimes there's really popular events where it's hard to get onto long without a ranking.
Posted: 11 December 2007, 6:57 AM
Thanks for the translation - my interpretation was correct.
The NZ system is slightly different from the Swedish one - the EWT of the class is used. The scaling factor as shown in the detailed description on the website is actually redundant as the scale factor is cancelled out and the formulae ends up as independent of the scale factor.
I will amend the website so that scale factor is not shown any more.
Statistically speaking, in New Zealand we always have 'thin' classes so we have to allow for these - the rankings will not be perfect - that's why we have selectors to select teams.
Posted: 11 December 2007, 10:32 AM
That difference (top 3 place rather than rank) moves the points calculated by quite a bit. What system are you thinking of using Bryan?
Posted: 11 December 2007, 11:09 AM
The program I used for all previous rankings actually uses 'top 3 place' rather than 'top 3 rank' for calculating Tm (average theoretical time) even though the website says otherwise - this was picked up by Jeff who queried me about his manual calculations of ranking points for some events compared with what was generated by my program. (Thanks Jeff - I will amend the description and keep the program the same).
The difference would usually not be much as top ranked runners usually win (unless one or two of the top ranked runners run badly which is why calculating a theoretical time average based on the first 3 is better than the top 3 ranked as the points could get skewed).
Posted: 11 December 2007, 11:31 AM
It makes a massive difference in points over the past 18 months, have yet to see if the ranking places change.
Posted: 12 December 2007, 12:01 AM
In NZ I think there would be a large difference between the 3 top ranked runners and the 3 top places. When you only need 4 good runs to get a good ranking it doesn't require the top ranked to be very consistent.
Posted: 12 December 2007, 7:28 AM
Now let's try to clear some things up.
- the NZ ranking system as it stands is almost the same as the Swedish (the only difference is in the use of different EWT)
- Both systems use the top 3 ranked in a class to calculate points.
- Both systems calculate a theretical average time called Tm based on the top finishers (not the top ranked)
Here's an example:
Rankings M40 from 2005C
1 0 Peter Swanson
2 2.8878 Jeff Greenwood
3 8.7114 Andrew McNeill
4 9.4252 Hamish Goodwin
2006 Nationals long distance
1 Peter Swanson 53.45
2 Paul Dalton 59.0833333
3 Hamish Goodwin 63.21666667
5 Andrew McNeill 70.56666667
Current NZOF system:
Pm = 6.0455
Tm = 58.5833
1 Peter Swanson -0.179
2 Paul Dalton 6.651
3 Hamish Goodwin 11.664
5 Andrew McNeil 20.578
An alternative system would be if Tm was calculated using average time of the top 3 ranked:
Pm would still be 6.0455
Tm would now be 62.4111
Points would now be:
1 Pete -4.15533, 2 Paul 2.257, 3 Hamish 6.96, 4 Andrew 15.539
resulting in a difference of about 3 to 5 points.
Remember that average of best 4 is required over the time period
so the overall difference is approximately 1 point which is not much.
Let's try a worst case scenario with the alternative system:
Rank A 0 points, Rank B 5 points, Rank C 10 points
1 A 100.00
2 B 110.00
3 D 120.00
last C 200.00
Pm for both would be 5.
Tm (NZOF and Swedish system) 110.0000
Tm (alternative) 136.6667
Resulting in huge negative scores for the alternative system.
I will still keep the system as it stands unless someone can point out a flaw or a refinement.
Posted: 12 December 2007, 10:25 AM
No flaw in the system, just not doing what is actually listed.
Here is a real example of the difference the 2 different methods produce (using a ranking of up to date points at time of the event)
Waitangi Loops Race-M21E (ranking points)
Ross Morrison 75.13 (4.380547348)
Michael Adams 75.23 (13.54367361)
Mark Lawson 79.59 (8.750701441)
Tane Cambridge 82.22 (23.55832276)
Carsten Jorgensen 82.48 (0)
Nick Harris 108.43 (33.220985)
Rob Jessop 121.53 (5.11413599)
Jeff Greenwood 123.33 (28.93698443)
Points produced by taking top 3 ranked
Ross Morrison -14.89225283
Michael Adams -14.72582751
Mark Lawson -10.13248867
Tane Cambridge -7.752606582
Carsten Jorgensen -7.319900749
Nick Harris 18.5592366
Rob Jessop 31.70683692
Jeff Greenwood 33.37109013
Points produced by taking top 3 times
Ross Morrison 6.775852894
Michael Adams 6.990290857
Mark Lawson 12.90877865
Tane Cambridge 15.97524152
Carsten Jorgensen 16.53278023
Nick Harris 49.87788354
Rob Jessop 66.81848265
Jeff Greenwood 68.96286229
Thats a massive 21+ swing in points and I guarantee that the top 4 would be using their points gained by the top 3 ranked in their best 4 results for ranking, but not necessarily if the top 3 times method was used. This happens when a top ranked runner limps around the course.
Are you keeping the system as you current use, or currently state? As this shows I think we would get more consistent ranking results from the top 3 times.
Posted: 12 December 2007, 1:01 PM
I'm keeping the system we currently use - it will use the top 3 times (actually better described as top 3 ranked, top 3 times). I won't use the top 3 ranked (better described as top 3 ranked, top 3 ranked times) as from your excellent example with Rob limping around, it produces huge negative results.
I can't see how you got the ranking points above though, on the web site it shows differently eg Ross (3.4368), Rob (4.6916) from the most recent rankings.
Posted: 12 December 2007, 2:04 PM
Used a system of updating after every race, going back to the last issued list and the system stated on the website (top ranked)
Posted: 12 December 2007, 4:29 PM
I only update after every list is calculated not after every event.
The system stated on the website is wrong and will be changed soon - the scalefactor is not required,
and the statement:
'Set Time1, Time2, Time3 to be the finishing times (in seconds) for the 3 ranked runners.'
'Set Time1, Time2, Time3 to be the finishing times (in seconds) for the first 3 finishers.'
The ranking software already does this.
Posted: 12 December 2007, 11:53 PM
too intense for me
Posted: 13 December 2007, 12:07 PM
Likewise I don't follow the maths but I appreciate that we have a method which scales not only the times but their spread, so that different performances can properly be compared.
Some multiday events have used this sort of maths for overall scoring, and its come in for a bit of stick. Mostly fear of the unknown I suspect.
Posted: 13 December 2007, 12:22 PM
The "when do we update the factors" is interesting. I would think that in an ideal world we would use the very latest information about rankings in order to scale the results (unless this caused the rankings to swing about wildly when some damping might be needed).
Practicality suggests this might be un-necessary, but we really need to go through the process 2-3 times a year don't we? Otherwise we could be scaling results on the basis of out-of-date abilities?
Posted: 13 December 2007, 2:52 PM
I agree, the ranking list should be created 2-3 times a year. This year unfortunately most of my spare time was spent creating the Hapua Wetlands map (which I'm pleased is going to be used for the relays at the nationals next year).
Problems with 'updating the factors' are:
-Some events have very few competitors in many classes
-Some competitors only compete in the South or the North.
-Got to be careful to have enough counting events (but not too many)
-there is a lot of drift of competitors upwards and downwards with some classes completely changed at the beginning of a new year
-When to remove someone from the rankings
So I have decided on the practicle solution of just using the rankings of the most recent published list for all the events in
the next list.