Running MoJo: Different things to do!

healy1835 wrote: »

How about everyone runs a 5k or a 3k time trial. We take the fastest time and then everyone else's handicap is their percentage of the winners time, and we use that for the following 'event' and so on.

Eg. I run the fastest 5k in 20:00, AMK runs it in 21:00. AMK's time is 105% of the leading time, so for the next event he gets a 5% time handicap. Everyone can have a rattle then, and we don't need to worrying about 16:40's

How is this better than Skyblue’s far simpler method that doesn’t call for a ‘sandbagged TT’. :pac:

healy1835 wrote: »

I would posit that this would give a far greater % of posters a chance to get near the top of the rankings than an age based table. Not all posters are of similar age/standard/experience. No need to over complicate this though just a thought....

+1 on simpler the better, cause I can’t make head nor tale of skyblue46’s approach and that’s after getting a broken down lesson on it!!

SuspectZero wrote: »

Age grading is all over the shop as well. it's not as fair as it sounds and is really only accurate in comparing the same event within the same age group. For instance, If a 20 something woman were to take part in a 3k race, they would age-graded against a full-time pro athlete who was juiced to the gills on EPO like Wang Junxia.

Look at the 5k. If you are 44, you are marked against Bernhard Lagat running 13:06, one year older and you are marked against 14:11(Mainly because someone with Lagats talent hasn't run a 5k at that age). In the 10k, you compete against 27:49 as a 44 year old, 29:44 as a 45 year old

Handicapping is fairer and more balanced if you want everyone to have a shot.

Where are you getting those figures from - perhaps you’re looking at the age category records that are based on 5-year intervals (ie M40, M45 etc)? Age grading is based on single age records - ie the record for an actual 44 year old. The factors are not updated every year however - latest set is from 2015. But AFAIK it is also adjusted to filter out outlier times (such as what might be produced by an outlier juiced up athlete).

More here: https://www.runscore.com/Alan/AgeGrade.html

Its obviously not an exact science, but it’s not made up either! I would still think its a better indicator of relative ability than winning a handicapped event.

The weakest thig about AG is that the single age records are probably softer the older you get, as the pool of runners producing the data is smaller and therefore less competitive (arguably).

On the other hand, handicapping is possibly a better way to allow people of different AG standards to compete.

Why not both? Binary options are so limiting.

A lot of interesting analysis here but I can’t follow your points because I don’t know what you are looking at. The page I referred to has many different tables and curves, and links to a great deal of data, as well as detailed explanations of how the AG curves were drawn. Also a lot of people involved in doing this work - don’t think it can all be dismissed that easily.

Lagat may well force a rethink but that hasn’t happened yet because his 5k 13:06 is from 2016, whereas the latest AG tables date from 2015. No doubt this will be factored into the next version of the tables, if anyone ever produces them. I am not a statistician and could be very wrong, but I don’t think the design forces a redrawing of the 10k curve due to a 5k time, as you are suggesting for Lagat. The 10k curve from which the AG factors are derived are based on 10k times only, and adjusted to filter out obvious outliers, as clearly explained on the page. Same goes for all distances. (Unless I’m missing something very obvious, which is entirely possible!)

Anyway this is a discussion that probably should take place in a separate thread - I’ll happily engage if anyone sets it up.

Murph_D wrote: »

A lot of interesting analysis here but I can’t follow your points because I don’t know what you are looking at. The page I referred to has many different tables and curves, and links to a great deal of data, as well as detailed explanations of how the AG curves were drawn. Also a lot of people involved in doing this work - don’t think it can all be dismissed that easily.

Lagat may well force a rethink but that hasn’t happened yet because his 5k 13:06 is from 2016, whereas the latest AG tables date from 2015. No doubt this will be factored into the next version of the tables, if anyone ever produces them. I am not a statistician and could be very wrong, but I don’t think the design forces a redrawing of the 10k curve due to a 5k time, as you are suggesting for Lagat. The 10k curve from which the AG factors are derived are based on 10k times only, and adjusted to filter out obvious outliers, as clearly explained on the page. Same goes for all distances. (Unless I’m missing something very obvious, which is entirely possible!)

Anyway this is a discussion that probably should take place in a separate thread - I’ll happily engage if anyone sets it up.

I didn't want to start a thread for just you and me to hammer this out so just waited for this thread to kind of fall off.

Firstly, I'm not dismissing the work of any of the people who done it, As I said, this is by far the fairest age grading tables I've ever seen and I actually like the way they have gone about it. I actually haven't disagreed with any of the formulas used, the data points are the limiter as Lagat is proving(nothing they can do about that and they are obviously aware of that too which is why the tables are so theoretically based).

You are missing something though which is that you have taken the words very literally even though the math and numbers show there's more than one thing going on.

On the point 10k times been based on 10k times only, that's incorrect, times are are not just based on times ran in the same event . let me show you an example from the open times:

1500m 3:25.8(no one ha ever run this fast)
Mile 3:42.6(no one has ever run this fast)
2000m 4:43.3(no one has ever run fast)
3000m 7:20.6(Daniel Komens World record)
2 mile 7:54.x(no man has ever run this fast, actual world record is 4 seconds slower)
4000m 9:58.0(No man in history has ever run 9 laps under 9 minutes nevermind 10, actual world record is 10 seconds slower)

So the question is where the heck did the other times come from? they are all the theoretical equivalent time of daniel komens 7:20.67 when you put them in a running calculator so times in other events effect times from other events so Lagats 5k will effect the 10k, the 3k etc etc. its happening in every event at every age group. These might seem like small differences at open level where times are super optimised already. But with Lagats 13:06, It means a drop of the 10k time from 28:04 to 27:17 because it means if someone can run 13:06 at 41, the bar of performance at other distances is theoretically much faster too. It's outlined why this happens down the main page where the age-factor graph is.

I wish I could draw this as it would be much easier but it's shown that you don't lose the ability to maintain the 1.08 coefficent runing calculators use to calculate equivalent performances between 5 and 10k as you age. So when someone destroys a 5k time, it means that the 10k is weak, the tables and formulas used by him and backed up by his own words say that these tables are used to sniff out weak times(which is why I like it so much as a stat fan).


On outlier times, your assumption is not correct. He has only excluded doped up Ma's Army, Eastern Bloc and African under 18's due to rampant age cheating as well as a 50 year old Ukrainian running 2:29 in the marathon, every other time is there no matter how far ahead it is of their peers. The Bell curve is actually based on outliers because if it wasn't, the graph would look like a saw blade but it's smooth and progressive as to sniff out weak times, look at how many single age records fall above the curve(to signify weak and are therefore completely ignored in the tables in the same way as El Guerrouj middle distance records are ignored as weak as illustrated above)

Single age records are used but only very little where the line is drawn through it on the graph(none of the single age records that are above the curve are used, this is called an inflection point in math), it's used so you can theoretically fill in the gaps between points to reduce the need to over rely on real time data that isn't avaiaible. A way to show this would be using a hypothetical using three data points:

39 year old 5k world Record: 13:30
40 year old 5k world record: 14:00
41 year old 5k world record: 13:46

when a 41 year old WR is faster than a 40 year old, you can asume that the 40 year old is a weak data point so completely ignore it and to get a theoretical 40 year old time, you draw a line from 39 to 41 and can say 13:38 is a better representation of what it should be from the evidence of the other two points. When you stretch this out say 60 years into a bell curve, you might only use 4 single age records that are outliers from the other 56 data points to make your theoretical representation of how aging deterioates performance at those ages in between defined as weak. this is what he has done and its a good way to do things as real data can be extremely flakey. Hence why the the tables are so smooth in there fall off and not all over the shop like the real world is.

But then, sometimes a new data point like Lagat comes along and scan show just how inaccurate your data is, it doesn't mean your method is flawed, it just means that the old real data points used to get the inflection points is out of date and it means a complete overhaul of your plot points at every age, distance and level because it shows the effects of aging are overestimated currently.

When you now go to draw your curve through lagat, you notice the line has changed its path through every age and event. like this for just a hypothetical example:

old data

30 year old 5k world Record: 13:30
40 year old 5k world record: 14:00
50 year old 5k world record: 14:30

if I want to find the theoretical ability of a 35 year old, my line would pass through 35 at 13:45, 32 woould be 13:36, 45 would be 14:15

new data

30 year old: 13:30
40 year old: 13:45(new lagat type data)
50 year old: 14:30

Now the line passes through 35 in 13:37, and 32 at 13:33, 45 would be 14:07

This is a simplified version of how inflection points work using a straight line between points(less accurate), bell curves with exponential variance points like used in the tables would be much more complicated and more accurate but the above illustrates just how impactful one data point can be, especially one as impactful as Lagat. Add in running calculator equivalences mentioned earlier and it means just not a rewrite of single age, single event graphs and tables but a rewrite of every table.

So in short, this does not devalue or dismiss their method(which I really like), it just shows how lack of data can completely skew advantages, often massive ones towards masters athletes and even moreso the older they are. This is not the statisticians fault but it just illustrates the how falliable and the limitations of these kind of tables when it comes to make comparisons across generations and events. The advantages are inverted from the young to the old. Turn a graph upside down and you'll get a sense of this.

The math and world records are with me on all these things I'm writing I promise that, I may not have a perfect understanding of every formula or be as mathimatically gifted as the people who created the table but I know enough to work back and see the patterns.

The whole table uses a combination of Single age records used as inflection points, Running calculators and a formula for age factor(i.e loss of speed and gain of endurance through age as in you are closer to peak at 800 at 18 than the 10k etc).

Murph_D wrote: »

That's all very interesting. However as this whole discussion came up in the context of a 5k road race, I don't know why you continue to focus on the Track tables (which date from 2010).

I was talking primarily about the 2015 Long Distance Running tables, which cover distances from 5k to 200k. These tables are based on data drawn from single age records for 5k, 10k, Half Marathon and Marathon. Records for those distances were plotted on individual graphs. Curves were drawn using various (no doubt questionable) criteria to exclude obvious outliers, most of which seem to be at the very young / very old ends of the spectrum. In most cases the graph predicts better times for these outlier values (e.g. the tables use 20:44 for an 80-year-old male 5k runner, when the record (when the graphs were drawn in 2015) was closer to 23:00. So the tables certainly seem to err on the hard rather than soft side.

Standards for events other than 5k/10k/HM/M were derived (using interpolation) from data from these four events, as the researchers felt there was not enough data for less popular distances to generate reliable standards. I'm happy to trust the way these mathematicians have treated the data.

Lagat doesn't come into it - both track and road tables predate his 2016 and later performances. Any new track tables would have to take this into account however. Some of his road times (e.g. his 13:48 at Carlsbad at 41) are also outside the scope of the long distance tables, other aren't (e.g. his 13:32 at 39 in Philadelphia).

I don't think you and I are talking about all the same things, but I certainly appreciate your analysis!

It's not about the track or the road, the track is just my choice because I'm more familar with the times(both masters and Open),I'm looking at this holistically, the same problems will still exist with the tables. If you want me to point out a problem with the road times, I'll point to Sammy Kipketers 5k World Record in carlsbad at "17" etc.

The same problems will still exist, the data is limited as Open cat times are super optimised, I'm showing the flaws here, for instance if Lagat's 13:32 at 39 never happened, that data point wouldn't exist and the tables would be completely different. For instance, his 13:48 you refer to at 41 been absent doesn't have as huge an effect as his 27:49 road 10k at 41 which is another formula breaker, he nearly ran as fast for 10k as he did for that carlsbad 5k, this would also mean the 5k the 5k would have to be set to 13:21 at 41 due to equivalence factor(faster than Lagats time at 39) and we are right back to exactly the same problems I outlined above in my last post, the whole tables have to be rewritten again.

All this once again with the exponential nature means much softer times as the comparison get older. There is a lot more going on than single age records(your point about the 80 year old record for 5k proves the point I have been making but you have been disagreeing with) and the graphs actually use outliers(a bell curve is impossible without using outliers). you can say that the graph errs on the side of harder towards older because of that 5k been harder(more likely down to less data due to very few people that age competing) but when someone like Lagat with world class talent and a pro contract comes along and blows away the standards at every single age, it actually shows they are soft on aging, how much the further away you get from Lagat where optimisation is occuring is guesswork.

If you ook up the age records and compare them to calculator equivalents and you will see most of the event times are derived from using a calculator. the pattern is there, i.e .005 seconds per 400 slower at every distance one year above prime, .05 two years above prime at every distance(this is only partly based on age records as real life is not this perfect). We are talking about the same thing, I'm showing how limited by data the tables are for cross generation comparison and using Lagat(a dream data point) as my reference.

It's impossible to hold both the opinion of trusting the tables(road, track, 2005 and 2010) and aknowledge the existence of Bernhard Lagat post 40 as he shows the completely outdated nature of the data and how flawed it is now. It's not their fault, they done the best they could with the data they had at the time but things have changed since making the tables obsolete, Lagat has proved they were never fair

Run around the green in my housing estate 100 times?

Something different and probably something that not to many of us ever consider;

Why not do a jumps conditioning workout

https://www.youtube.com/watch?v=MVrbbOMMPNE

There is huge transfer between neuromuscular stimulation and performance but most of us confine it to simply doing the odd max velocity hill sprints (i.e 6 sec)