Wednesday, February 20, 2013

Give me a break! Plea from ordinary mug


In the hope that one day an ordinary mug like myself could win something other than a lucky dip, how about a real handicap calculation which doesn’t only adjust for age and gender, but also ability?

I originally thought about this years ago and may have even written about it in your old feedback columns. I thought it might be time to revisit and develop something.
 
Thinking back to the Dezzie idea from a while ago where times were compared to the event winner, there are a number of anomalies:
  1. Was the winner be particularly fast/slow? Was it an Olympic gold-medal standard or just a Commonwealth Games gold medal standard?
  2. Were the conditions that the winner swam in significantly different from the rest of the field? This could make your Dezzie look better or worse, and you don't have any  idea if you have improved against the field based on your previous efforts.
  3. On longer swims, the winner is always significantly faster than those finishing further back in the pack, probably because they’re toiling for a much shorter period.  This gives you a worse than expected Dezzie.
So, I reckon we need to come up with a way to adjust for these problems and develop a world-first oceanswimming handicap system.
 
Where to start???  I’m not a golfer, but, in golf, payers of all abilities are rated and there are similar problems and a well-developed handicapping system.  So, I did some golf reading:
  • Different courses are assigned course ratings and slope ratings. The course rating is how difficult a course will be for the scratch golfers and the slope rating adjusts according to how difficult the course is for those further back in the pack.
  • Scratch golfers are those who’d score “Par” when playing to their ability on a course, so they’re not necessarily the pros, just very good golfers. It seems there are teams of scratch golfers who visit different courses, play them and rate them.
  • In Australia, there’s even a daily scratch rating to adjust for conditions, but that’s really getting complicated.  Some interesting reading... click here and here
In oceanswimming, having a daily scratch rating may be impractical, so to get started, I thought we should find the “scratch” swimmer in each race and use their time as the basis for the handicapping. My thought was that the swimmer finishing 12.5% of the way through the field in each swim could be the one who provides the time base for the handicapping. So, if there were 504 people racing, then the time of person who finishes 63rd is the course rating time. I came up with this a while ago and can’t quite remember where it came from. I think it was from the now defunct Australian Golf’s CCR (Calculated Course Rating) which was determined by taking the 15th percentile score (ie assuming that 15% of players played to their mark or better, with 85% worse). Not exactly the same concept, but something to work from.

The next step would be to compare each swimmer’s “Event Handicap” (for the swim just completed) with their “Handicap average” taken from their last five swims. The swimmer with the highest “Event performance compared to Handicap average” (the greatest improver in the particular swim) would be the week’s handicap winner. The idea is that the handicap award would be restricted to those swimmers who completed at least five swims over the previous 12 months.

Summary of steps

  1. Calculate finishing time for the placegetter who finished 12.5% of the way through the field (our scratch swimmer)
  2. Calculate each swimmer’s time compared to the scratch swimmer. This is the “Event handicap” time for each swimmer.
  3. Calculate each swimmer’s “Event performance”. This is their event handicap compared to their average handicap from their last five swims.
  4. The swimmer with the best “Event performance” is the handicap winner.
(Rob attached a spreadsheet with his results since 2004 showing this event performance handicap system: os.c)
 
With all the data (oceanswims.com) must have, this, or something even better, must be achievable. Anyway, I’m not a statistician, or a great mathematician, just a guy who, even after eight years and 70-odd swims, is still trying to figure how he’s going in this oceanswimming caper.

Anyway, it would be interesting for ideas/feedback from your resident stats guru oceanswimmers.

Rob Salomon

12 comments:

  1. I'm glad someone has raised this. I was interested to see that 9 of the top ten current dessies are >55yo, which being neither > 55 not particularly fast rules me out of any prizes :)

    In any case. Judging off the top 10 / 12.5 / 25% turns out to be good for fields of consistent depth -ie excluding your smaller swims, cole classics, bondi-bronte etc - this doesn't work well for any swim where there are a large number of tail-enders, nor where the field is especially tight (eg SLSA champs!)

    Unfortuntately basing results off proportional time (eg x% above) the 1st / 5th/ 10th swimmer depends also on the depth of the field and the individual swimmers at the top of the pack - for example it won't work well when you have either a freakishly fast winner, or a smaller field where tenth position is two minutes behind the winner.

    Correcting for the above would ideally

    * be based off the top times (eg your time vs 1st or 5th) combined with some other measure of the field (eg your time vs 10%)

    * take into account the depth of the field - are the fast people at a different swim that week? This can probably be controlled for by a offset based on swim size eg scale based on number of participanst - ie penalize smaller swim results, as comparing your times to the winner's or tenth place will on average make you look better than you are

    * control for slower (larger) fields in the comparison to 10/12.5/25th percentile by adusting for weak fields - this could perhaps be 25th percentile time divided by 10th percentile time - so then when you're comparing your own result against the 10th percentile you can then reduce it by the (for example) sqrt(25% time/10 % time) ratio to account for weak fields

    This being the first time I've actually really thought about the above, the formula I've previously used - that also seems to agree with how I actually felt about the swim is:

    performance = (my time / winner's time) ^ 2 * (my time / 25th percentile time) * (my time / 10th percentile time) ^ 2

    As this varies a fair bit (generally being much greater than 1) relative performance for each swim is then usful to calculate as (performance for that swim / max performance for the season)

    Additional utility can be derived from dividing the winner's time by 12 minutes (to obtain the ostensible km distance) then working out your own per-kilometre pace in km/h, this is always interesting to compare.

    ReplyDelete
  2. One final comment too - faster swimmers will benefit in currents running against them, by virtue of their higher relative speed. So if the distance of the swim is known and the current is high then performance can also be adjusted to suit, for example

    Bondi-Bronte is 2.4km nominal

    Winners will do it in ~28 minutes with no current and minor swell (assuming 12min/km), i.e. 5km/h velocity

    25th percentile will do the same distance in about 1.5 times the winner's time, i.e. 42 minutes = 3.4km/h

    Surface currents are reputed to be principally influenced by wind, which can easily be 15-20km/h - so if there's a resulting 2km/h current from the south then the resulting absolute speeds are

    Winner : 5 - 2 = 3 km/h

    25th percentile : 3.4 - 2 = 1.4 km/h

    Suddenly - as you allude to - the winner has a significant advantage from being in the water for less time, as they're going 3/1.4 = 2.14 times faster than the 25th percentile swimmer, as opposed to the original calculation of 1.5 times the speed.

    The opposite of course applies for a following current - with a 2km/h tailing advantage the winning swimmer is doing 7km/h while the 25% swimmer is up to 5.4 - a remarkable reduction of the time take to complete the swim to 7/5.4 = 1.3 times

    These combined are the reason why it's worth pushing harder through rips / upwind legs (ie improving your relative speed to the winner) - especially if you can then take it easy on downwind/with-current legs, where you won't be as disadvantaged (this being said, the nature of human physiology is likely to mean that your technique will probably suffer and you'll do worse by pushing yourself to 125% of your continuous capacity for a while then falling back to 80% for the rest)

    Anyway, some ideas to think about - as you say it would be interesting to isolate each of these into a specific factor to control for each variable and give a combined rating factor for each swim

    ReplyDelete
    Replies
    1. Thanks for your thoughts.
      I think this shows we need to adjust individual results from each swim for 2 factors... The difficulty of the swim based on (1) swim length and (2) the day's conditions.

      Delete
  3. Handicap system a great idea. Gives everyone something to strive for.

    ReplyDelete
  4. As an old fart, all the above is far too complex for me I'm afraid. I'd like to suggest a simpler approach.

    Here's how I look at it. The point of a handicap system is to give everyone a chance to win, right? So let's look at what it would take to give me a chance to win.

    Thanks to the "what's your age?" question all swims ask, everyone who cares - which I estimate as being only my wife and my parents - knows I'm 61. So advanced age is obviously one thing to be taken into account. But that's already included, as is sex, and I still came about 350th out of 550 at Maroubra.

    So what about my other handicaps?

    Well, first of all I'm a Pom and therefore a) didn't learn to swim until I rocked up in Oz in 1979 at age 28, and b) I'm scared of water in case anyone accidentally drops some soap in it. I reckon being a Pom should be worth subtracting 1 minute per kilometer per decade living in England, which gives me 3 minutes per k.

    Next, I'm a short arse. We all know the longer you are the better you glide and the bigger your hands the better you pull. So I reckon anyone less than 6 foot (being 61 and a Pom I really don't get these metre things) should subtract 30 seconds per k per inch under 6 foot. At 5ft 6 inches that's another 3 minutes per k.

    Thirdly, I'm a tad hairy and there's no way any sensible beautician is going to try to wax my overall covering of fur even if I wasn't allergic to pain. I reckon that's worth 30 seconds per k for a hair covering of more than 75% of your body (my head excluded).

    Fourth, after more than 40,000k running and many marathons and half marathons I have runners' calves. Even though I know where to put apostrophes my calves are less than optimal for things like bending and kicking and this is a major handicap. I suggest 30 seconds per k per 10,000 kilometres lifetime running would be a fair compensation, which is 2 minunes a kilometre for me.

    Finally, I'm a coward and therefore scares of any wave bigger than a sizeable ripple. This sends adreneline coursing through my body and exhausts me - well worth an allowance of 30 seconds per k per metre of swell.

    Lets see how that would work at Maroubra.

    The swim was (ostensibly) 2.4k. So my handicap would be;

    Pom allowance - 3 minutes per k x 2.4 = 7.2 minutes
    Shortarse allowance - 3 minutes per k x 2.4 = 7.2 minutes
    Body hair allowance - 30 seconds per k x 2.4 = 1.2 minutes
    Runners allowance - 2 minutes per k x 2.4 = 4.8 minutes
    Cowardice allowance - not applicable, no waves

    That comes to 20.4 minutes. Add 30 secs per k for every year over 30 = 3 minutes x 1.2 = 3.6 minutes, giving me a grand total handicap rating of 24.0 minutes.

    My time was 47.45, less 24 minutes and that makes my time after handicap 23.43. And that makes me a clear winner, with Jarrod Killey crawling out of the water more than a minute behind me in 24.54.

    Send my winner's cheque in the mail please, to

    Steve Hall, c/o deluded fantasy land, Leichhardt, NSW 2040





    ReplyDelete
  5. I agree a handicap system would be great.

    Personally, I use the "Kelso System" - I compare my time to John Kelso. I was 114.4% Kelso at North Bondi and 112.3% Kelso at Malabar - I'll take that as improvememt. My goal is 99.9% Kelso

    ReplyDelete
  6. Steve I would support your Pom Allowance. I agree it's a massive disadvantage to be born in the UK.
    Don't agree with the rest of your handicaps though. Suggest you shave down and get your short-arse down to the beach to learn some surf skills if you want to be competitive.

    I just pick my bunnies to chase on each swim - a few regular swimmers in my age category whom I perceive to have roughly the same lack of swimming ability as myself. A race within the race.


    ReplyDelete
  7. Great idea, count me in when it comes to working something out, or playing with the data. Can I see the spreadsheet? (os.c has my details)

    Most of the issues in doing this type of handicapping are outlined above. It is the beauty of ocean swimming that even if you swim the exact same course twice, environmental conditions will be different and so will your times. You can't compare your time to the winning (or last place) time each week as they are too variable (if Thorpe or Hackett show up, we have a problem...). Which is why picking the 12.5% person is actually not a bad idea. You could also do something with the mean/median/ st dev of the times if you really wanted to.

    I feel I swim pretty consistently, so my result usually reflects the strength of the field rather than my own performance. At the big popular events, with more people simply giving it a go, I can crack the top 20% whilst for the Nth Beaches club swims I'm usually closer to the top 30%.

    So I'm not yet convinced we can do this in a rigorous way, but it's worth a shot and data is something us maths people like playing with. If you want a good data source, see www.oceanswimseries.com - I'm not connected to it, but it's a good way of seeing all the results of NSW swims on one page.

    I like the Kelso system - there's another West out there (no relation) who I compare myself to, she beats me 9/10 times.


    ReplyDelete
    Replies
    1. Thanks for the feedback. It will be great to have someone with a bit of statistical nous try to find the pattern in the data to devise a handicap system.
      I’ve also followed oceanswimseries over the years, but it’s not designed as a handicapping system.
      I've emailed my spreadsheet. As you’ll see, there’s not much raw data, just the winner’s time, the 12.5% placegetter’s time & my time.
      Interestingly, my results in the short 1km swims I’ve done this year are much better compared with the 12.5% placegetter. I think this is a result of an overall weaker field in these swims. So, “field strength” is another factor which needs to be adjusted for in addition to the “swim length factor” (where, on a longer swim, slower swimmers’ times are significantly slower than faster swimmers’ times).

      If you haven’t read this golf link, it’s pretty interesting: http://www.golfaustralia.org.au/site/_content/document/00011802-source.pdf

      Delete
  8. Great work Steve ... there should also be a handicap for volume of water and frequency googles fill up. Everytime I'm standing on the beach ready to lauch into a race, I say to myself, better get some new googles this week. Even a calm flat Malabar I twice had to bail out.

    ReplyDelete
  9. Hi Richard,

    Thanks for your Pom allowance support, I agree it's a massive disadvantage being born in the UK - for a swimmer that is.

    There are, however, advantages - being able to support winning rugby and cricket teams, seeing our flag in the corner of the Aussie one, being able to spell and punctuate correctly and of course understanding decent beer.

    Swings and roundabouts I suppose.

    ReplyDelete
    Replies
    1. Ouch.
      Might only be worth buying into this one after the Lions series...?

      Delete

Please use the drop down menu, Comment as, to attach your name to your blog.