No love for triple jump in the calculator?
Unfortunately at this point TJ has not yet been processed, so alas currently, there is no “love” for the TJ. I mentioned it here http://duncanscore.com/take-me-to/ but perhaps you did not see it.
should the age lists not be ranges? Eg: M 70-74 etc
They actually ARE age ranges (5 year) as per WMA. So 70-74 would be classified as M70 (for men) and W70 (for women)
I just put in a time one second faster than the world record time for W60 800 meters and it only gave a 97%.
It needs some work.
Thank you for your comment.
Actually, your example shows that this “system” works quite well in this instance. The W60 WR (2:33.09 as of when the 800s were processed) is in the dataset, and yields a SCORE of 967 with a percentile of 97%. A 1 second improvement on that yields a very slight SCORE improvement to 969 and also a 97 percentile. Were you expecting a percentile of 100%? The duncanSCORE is different from Age Grading (which often yields AG % over 100% for new world records). The duncanSCORE is based upon the number of standard deviations from the “average” (in this case the W60 800m average is 3:34) and will never be greater than 99%. So it is a statistical calculation and the 97% can be interpreted as “the probability of every W60 800 performance being as good as or poorer”. 97% is a very, very high percentile. (Actually a better description would be “the likelihood of a better W60 800m performance is 3%)
Excellent, Duncan! That’s a big task you gave yourself. Quite an accomplishment.
Pls excuse the length of my somewhat rambling comments below.
A number of questions and observations. The first being that the reference data is from 2013/4/5/6. Will that be updated yearly or something? Every 5 years? Does it matter much if its updated? Only if the average performance and the spread in an age group changes a lot, I imagine, not the world records which will surely change. Such changes in the averages would then be very slow I would think and might even go down a little as well as up a little.
One of the problems I had with World Record based Age Grading is the change in my own performance becomes muddled if the reference changes. So I could be getting “better” and my AG gets worse. So I have used the WMA AG 2006/2010 on-line calculator by Howard Grubb as a fixed, “absolute” reference point. Not so much to compare with my peers but to monitor my own performance and set goals.
Being somewhat obsessed with data I have a record of my yearly “best times” and AG% back to the year 2000. I plugged my best for each of the age groups from 60 to 75 100M and 200M into DuncanScore and the results were very interesting, sometimes surprising, sometimes encouraging.
A couple of things I noticed with my own data:
When I was 60 I did 13.08…DS=894/89%….WMAAG (Howard Grubb)=89.4%
When I was 65 I did 13.16…DS=867/87%….WMAAG (Howard Grubb)=91.6%
Does this mean that my 65 performance was better on an “absolute” scale but on an “average” based scale relative to my theoretical peers over the specific years 2013/6 (DS) it was not? It was worse even though 13.16 is quite close to 13.08? This does not look right.
I then checked what would happen for the same time, 13.08, at 60 AND 65 which would have been amazing. ….13.08 gives DS=894 as above at 60 but at 65 only gives DS= 874. Checked it 10 times. This apparent anomaly does not occur at much faster and slower times. (I only checked 12.5 and 14 at 60 and 65 respectively and they behave as one would expect). By chance this anomaly occurs around my personal 100M time. Worth looking into.
Regardless, I like my 13.16 at 65 much better than my 13.08 at 60!
100M 70/75 and beyond.
While I’m getting slower my DS is getting higher. In 2020, just 2 years away, I will (hopefully) enter the 80’s age group. Today the absolute time to achieve a 100M sprint DS=930+ seems like a walk in the park…..but we’re not there yet. This also must have something to do with the data for the older age groups being “different” from the younger. Fewer competitors for sure, so less accuracy, and the average dropping faster than the outliers records drop perhaps. So the more I succeed in maintaining my absolute time performance the greater DS age grade reward relative to the younger groups?
5 year age groups
While I acknowledge the issues, and that the WMA now uses 5 years for Age Grading, a 5 year age group age grading is still difficult to live with as time goes by except for comparison with everyone in the Age Group (peers) as opposed to tracking one’s own performance year on year. These are two different views of mostly the same information. How do I compare to others, how well am I maintaining my performance relative to some “reference”.
The performance drop between 65 and 69, 70 and 74. 75 and 79 is a lot, and increasing with age, so you can get the transition anomaly at the switch over point with a jump of 7 or 8% for similar absolute times. The above mentioned Howard Grubb WMA calculator gives 1 year (or fractional year) figures. For my own interest I will assume a linear change during the 5 year step for DS. Obviously crunching actual data for each year of an age group is both a lot more work and would lead to more anomalies and inaccuracies due to insufficient data. Linear interpolation over the 5 years is sufficient to get an idea.
WMA Toronto 2020
So based on the above and my current and projected DS, Toronto 2020 is looking very promising except for one thing. The event start date is 4 days too early. 4 days before my 80th birthday. On the day of the 100M first round I will be running with the 75 year old youngsters while I will in fact be 80 and likely the oldest in the field. I will have to compare my results with the 80 age group to see “the truth” if not the reality. My DS score will recognize the fact that I am 80 but the competition will not.
I have spoken to Doug Smith on a number of occasions about this most unfortunate choice of date for Toronto 2020 but so far without result…..
Thank you for your detailed comment … I appreciate it immensely.
First let’s clear up the M60-M65 questions if I may (and then I will reply to your other comments). THERE IS AN ERROR on my part that screws up the M65 100m calculation!! In the processing, I apply a standard error “filter” to try and weed out what would appear to be “unacceptable” or erroneous results (data entry errors by those who submit to mastersrankings.com, people who may “walk” the event, perhaps para athlete results etc.) I did not apply these filters to the M65 100m dataset! I missed this in my data checks. Revising that data would give your 13.16 a SCORE of 891/89 percentile (a statistical “tie” with your M60 13.08). For M65, 13.08 yields a revised SCORE of 898/90%. This makes a lot more sense than the SCORE you received. I’m not sure when I can get a corrected table uploaded (my web developer is overseas), but I hope soon. (By the way, I’d suggest your WMAAG% is probably a statistical tie as well).
Now, I hope you will allow me to move on from my boo-boos. I started out on this data quest late in 2015. For several months I played around with different concepts with a mixture of events and Age Groups, trying to see if any of these concepts would “stick”. A little over a year ago I basically finalized the methodology and went to work. At that time, John Seto had just 4 “complete” years of data … that’s why it’s 2013-2016. Obviously complete 2017 is now available, but I still have several events to finish processing (some Throws, and Jumps, and RW). So far I have processed over 340,000 performances. My plan is to update the data either annually or biennially (assuming people want me to). This will also depend on whether I jump in to INDOOR data.) I believe you are quite correct in that updates will yield only minor changes … it’s not likely 1 additional “normal” year would change the average or standard deviation very much. This is a big advantage over AG, I believe.
You are also right about the robustness of the datasets as the AG gets older. In the 100m, for example, there are 1996 M65 performances, 1544 for M70, 1236 for M75, 798 for M80, and only 286 for M85. This is definitely one of the weaknesses for the duncanSCORE … smaller datasets for AG 80+. The dropoff in average times becomes more pronounced, too … M75 17.5456 seconds, M80 19.4400, and M85 21.5524.
I doubt Individual Age/Year duncanSCOREs will ever happen … it’s 5x the processing and I doubt I’ll live that long. But I have thought about it. As part of the processing I keep individual ages, so I can calculate average performance by each Age. I’m thinking some sort of performance ratio against this average may be an acceptable compromise, but it’s only a concept and I haven’t even begun to tinker with the data.
As for TO2020, I know Doug likes his beer. That may be a good place to start.
Mike ,,, the M65 100m data is fixed
This method has merit as it is based on more data than just the best performance.
However, I have 2 comments:
1. This score is for a 5-year age range, but if, for example, the same time run is by a 64 year old and a 60 year old, the 64 year old should earn a higher score
2. In many track and field events, all performances are done at meets. But in the distance events, many performances are on the road, and often it is the ‘elite’ who race these in the track. So the DuncanScord includes only a subset of performances for 5 and 10k.
Let me respond to your 2nd point first.
Track and Road Racing are really apples and oranges. I fully recognize that the Track is a very marginal section of racing. But calibrating and comparing 5000m and 10000m times to 5k and 10k road times is really a fool’s errand. Generally road times tend to be slower than track performances for a host of reasons, so the roads would need its own set of tables. I have seen standardized scoring for the Roads, but these are based on “all” times, not on an athlete’s best annual performance, and as far as I know these have not been restricted to “certified” courses, which I believe, is the only way to get quality data. So the Roads is a huge mega project.
On the 64 year old vs 60 year old scoring, I have briefly delved into the single year analysis. The problem is the sample sizes. By definition, they are cut in 5. In fact, it’s a bit worse than that. There appears to be a big bump in participation at the start of a new age group, and then each year becomes less until the beginning of the next AG. Because of the smaller samples, there can be a considerable “wow” in the data. I am hoping when I incorporate 2017 results, this will be considerably lessened (ie overall samples will then be 25% larger by adding another year).
So yes, you have pointed out some limitations on the current situation. But hopefully in the future these can be mitigated somewhat
Indoor vs outdoor track would interest me.
My hope is that once I finish Outdoors, I can begin the process of adding Indoors
Your email address will not be published. Required fields are marked *