Multi day scoring systems

[andy]

Oh look, this old chestnut.

Under the Scottish system the racing for the first 2 places in M60L would have been over after the first 5 days. Under the Croeso system it was still possible to win overall and therefore the race was still open till the end.

The Scottish system is clearly flawed because it does not reflect the quality of wins or keep the competition open. However, it is simple to understand which is useful. The Croeso system is obviously better because it does reward exceptional runs and keeps the competition open, but it is more complicated. Another flaw of the Croeso system is that it can not differentiate between days where the spreading of the runners is different, for example forest days and urban days. A system using standard deviation would succeed in solving this problem, however it would add yet more complexity.

Virtually every statement in the above 2 paragraphs should be preceded with 'In my opinion'.

The problem with scoring systems is there is no right answer and it's entirely down to personal preference so there will always be anomalies and arguments/discussions. Pick one and roll with it.

[Jon Brooke]

NIce idea - maybe 10 prizes and I might get one....

I prefer the idea of 11 prizes.

[Jon Brooke]

It's certainly simple, but it's difficult to argue that it's "fair" in the situation where different days are of widely different styles and (crucially) expected winning and average times (for exactly the reasons you've pointed out elsewhere in your post).

But every course has legs that are widely different - the reductio ad absurdum of your point is that all courses should have the same number of similar legs on the same sort of terrain. In fact - it should just be a running race.

Despite the fact that it seems a bit odd to see some discarded days with better placings than the ones that were being kept, I liked the scoring system precisely because it left open the (faint) hope that a really good run could turn things round right until the end. I have to say that most of the results I've taken an interest in look "about right" - maybe some anomalies, but no travesties.

[babs f]

I believe that the lowest score of a finisher is capped at 100.

Either that or a surprising number of calculations came out at 100.0.

See M55L and M75L for instance.

I was puzzled by some scores until I saw that the "par time" is based on courses not classes.

[Ali Wood]

As an M60L beneficiary of a late start on Day 3 I would like to sympathise with Eddie re. the exaggerated scoring on that day.

Perhaps a standard deviation calculation would have been slightly fairer, but I would favour a Scottish 6 day points add-up (yes, even if it allows overall results to be predicted prior to Day 6.) OR adding up the total times (yes, despite some days being longer than others).

Finally I would like to express my gratitude for any system of scoring that allowed me the luxury of a wonderful run up Cader Idris on Day 6 instead of running around blocks of student flats!

[sketchweppers]

I was puzzled by some scores until I saw that the "par time" is based on courses not classes.

Which is why it would be great to be able to see the results by course as well as by class.

[PhilJ]

Finally I would like to express my gratitude for any system of scoring that allowed me the luxury of a wonderful run up Cader Idris on Day 6 instead of running around blocks of student flats!

Try the cader fell race, usually held end of May.

http://www.cader-race.co.uk/

[SIman]

I was puzzled by some scores until I saw that the "par time" is based on courses not classes.

Which is why it would be great to be able to see the results by course as well as by class.

Once the results are all on the BOF site (only Day3 so far) you will have them by course.

[david_rosen]

Eddie's post spurred me into recalculating the M60L scores based on
a) the "Stats" method which uses the mean and standard deviation
b) the Croeso method just using the mean
c) the current Scottish method which ignores times and just uses positions

I can send the spreadsheet to anyone who is interested.

As is usually the case, there is quite good agreement between all three and with all of them Eddie is still 3rd.

The best multi-day scoring method (in terms of both fairness and simplicity ) is definitely just to add all the times. But of course that doesn't work if you are going to use best 4 out of 6.

The Croeso method is exactly the same as the small classes formula of the Stats method. The Croeso formula can be obtained by replacing the standard deviation in the Stats formula by a fixed proportion of the standard deviation. So the Croeso formula works just as well as Stats formula if the standard deviation and mean are about the same ratio each day. That was true for Days 1, 2, 4 and 5 at Croeso where the ratios were were 16.7%, 15.8%, 16.1% 17.8% respectively. However, Day 3 was a tougher forest with a ratio of 19.6% and Day 6 was an easier urban race so the ratio was 11.2%. Hence, with the Croeso system (as Eddie has highlighted), the winners scored comparatively too much on Day 3 and not enough on Day 6 which disadvantaged Eddie with his mp on Day 3 and win on Day 6. The Stats system would have given Eddie his highest points on Day 6 rather than his lowest.

The Scottish method, while simple, is very prone to coming up with obviously erroneous results. Orienteering races in multi-day events are time trials so you can't engineer a 2 second win in the way that you can in head to head racing. For example Alastair Wood's 1st place by 2 seconds on Day 1 is given the same score as Jean Dermine's 90 second win on Day 3. There is actually nothing to choose between Alastair and Jean Dermine over the six days (Alastair won by just 5 Croeso points) yet the Scottish method would have given Jean 8 points overall to Alastair's 5; a large margin which didn't reflect reality.

It should be quite possible to print the current score based on the results so far on the results sheets as they are posted during the day. Just as people understand that the positions on those sheets will change, people will also understand that the scores will change as the results come in.

The Swiss 5-Days organisers this year wanted to have best 3 scores out of the first 4 days and then a chasing start on the last day. To do that they absolutely had to use a system which converted times linearly to points. They successfully used the Stats method.

Since Croeso are happy to use a relatively complex formula, I would urge them to go the whole hog and use the Stats method next time which will reduce anomolies arising from terrain variations such as we saw with Days 3 and 6 this year.

[EddieH]

David, I am not disputing for one second that I didn't deserve to win having failed to complete 2 days, the whole point of my thread is that knowing how the now know that should day 6 be urban in future and one is behind as I was there is no point whatsoever in believing that just conceivably one could turn it round. Similarly should there be middle distance forest amongst a lot of open terrain it is vital to do well on that day.

In fact I already thought this, as I was determined to run cleanly and hard on day 3, and on day 6 I had already told my club mates that to make the top 2 was possible only in theory.

Of course the position system would not hve been any different, but personally I like it, and whatever else, it is really simple and encourages a racing mentality rather than a safe time getting performance. The difference was clear to me when on day 2 I turned round in the start lanes to see Ali Wood behind me Every step behind me I thought might be him (even before he'd probably started

[Mrs H]

If Croeso had used the S6D system instead I would have finished second, not third on count back - that would have been very unfair on Oonagh who was definitely traveling faster than me but we both had 1,2,2,4, followed by my 4 and her 6. Both of us only got 5 results.

I'd like to explain the NGOC mini-league scoring system which seems very satisfactory but I can't find any description of it. It's got something to do with the average time of the top 3 being used as the base score.

[Jon Brooke]

The best multi-day scoring method (in terms of both fairness and simplicity ) is definitely just to add all the times.

Unless you bash your head on branch sticking out above a really stupidly placed control and are forced to retire on one day (?)

Don't know if Clive Hallet would have won overall had he not done that on the last day, or even if he would agree with me and disagree with you, but I like the idea of having at least one discard to at least mitigate against the odd bit of genuinely bad luck.

[HarryO]

I don't think there exists a scoring system upon which everyone would agree. If there was, we would be using it

[awk]

Given that one needs to have a system where days can be dropped, I like the simple way the Scottish 6-day now uses. One point for 1st, two for second, and so on. (Or one where you score something like 100, 99 for second etc.). Simple but effective, and focuses on what matters: orienteering is about your position in a race not about your time (you don't get medals for finishing close behind the winner).

I have no time for systems which use complex maths to try and 'be fair'. Just takes away from the racing.

[Red Adder]

Just watched Ben Ainslie win the Finn class at the Olympics. Perhaps we could consider something like their system - 1 pt for first, 2 for 2nd etc. On the final day double points and in event of an overall tie then last day position decides it.

Personally I'm all for something pretty simple - either cumulative time (means that everyone must compete in all days with large penalties for DQs / DNS) or some sort of positional points summation. The statistical methods may seem fairer (and are loved by those of us with a slight geekish tendency) but always seem to attract tons of argument, are complex and time consuming to calculate and are pretty opaque to most people.