Analysing TT performance

[David Martin]

It has bugged me for a while, the question of whether you are really going well, or whether it is the conditions.

I know a number of the regulars at the Fife evening TT's note the times of a few 'markers' to see how they do compared with other riders (if the conditions are good then you would expect everyone to go better). But how do you compare different distances and courses?

I have formalised this in a simple manner.

In race a you race against n other people. They will all have a time that is some percentage of yours.
In race b you race against m other people. A subset of these will also have raced in race a.

If you take the relative times from race a (x) and plot them against the relative times for race b (y) then you get a straightish line. If your dot (at a ratio of 1) is above the line then you went better in race a than in race b compared to everyone else.

If you average this over the whole series (compare race a to b,c,d,e) you can get an idea of where you were relatively going well, and where you were relatively poor.

You can also see clearly whether you are good on the hill climbs, or better suited to dragstrip courses.

Preliminary work up of this is available at Rider list which lists all the riders I have entered dat a for so far (from the fife midweek series for 2008 and 2009) Click on a rider and you get a list of all the events they have ridden and their relative performance in those events. Over 0 is a good day, below is a bad day.

Example: Rider analysis. In event 5 I set a PB, but it was perfect conditions so everyone was going faster, hence only getting a good score of 1.5

This is a first cut. I have to do the pretty bits and some graph plotting too.

[gonzo]

You're ignoring the fact that some people are much better on a given distance/course etc. If you want to compare yourself to others, just race the same category of races and use your finishing position as a good guide. This ignores individual people and takes into account conditions etc.

[David Martin]

You're ignoring the fact that some people are much better on a given distance/course etc. If you want to compare yourself to others, just race the same category of races and use your finishing position as a good guide. This ignores individual people and takes into account conditions etc.

No I am not. If you compare individual races of a certain type that are all compared against the same baseline then you get the effect you describe.

The system is far from perfect but does give an interesting view of performance that is otherwise not readily available.

It would be possible to sort out relative positions but there is probably not enough resolution in the position axis to give anything meaningful (ie you would have to swap positions with others - not so easy when the times are spread over a 33% variation.)

selecting individual races as the baseline should be fairly trivial to do.

..d

[nmcgann]

It has bugged me for a while, the question of whether you are really going well, or whether it is the conditions.
......

Not an issue when you race with a powermeter, the info is there.

[David Martin]

It has bugged me for a while, the question of whether you are really going well, or whether it is the conditions.
......

Not an issue when you race with a powermeter, the info is there.

Not so. As you go through the season your performance will change. As will everyone elses. Are you going better than would be expected for the time of year?  And most (OK all but maybe 2) of the riders in the league do not have powermeters.

[nmcgann]

It has bugged me for a while, the question of whether you are really going well, or whether it is the conditions.
......

Not an issue when you race with a powermeter, the info is there.

Not so. As you go through the season your performance will change. As will everyone elses. Are you going better than would be expected for the time of year?  And most (OK all but maybe 2) of the riders in the league do not have powermeters.

Not sure I see the point you are trying to make. Power is always the key data I look at to gauge performance throughout the season, times are interesting, but there are so many external factors that can affect them.

I don't often have the opportunity of comparing myself against people I know of "similar" ability. I ride almost exclusively opens and usually only know one or two other riders so the competition for me is a personal attempt to improve on my best power output. My best ride last season wasn't my fastest, it was a big PB power output on a slower course.

[mattc]

David,

This sounds really good. I agree with your paradigm - measure against others to eliminate terrain/weather etc.
And I see your point about time-of-year (although of course you assume the other riders are on the same training cycle -  most riders rest in Nov-Dec etc etc ... )

What is not clear, is how you deal with holes in the data. (I gave up trying to do this analysis 'manually' because my local riders only manage 2/3 of the same events.) Could you elaborate?

[David Martin]

Holes in the data are not a problem.

Each race you have done is compared to every other race. You can calculate a bestfit line for comparing all riders who have participated in both races, and then your score for that race is how far off the line you are (bear in mind that your position should be 100/100 percentage wise.

Then average for all races in which you have participated (ie if you have done n races then you will be averaging n-1 values)

So the averaging is per race rather than direct rider-rider comparisons.

..d

[mattc]

I can see how holes don't HAVE to be a problem, but I still don't understand how you worked round them! I think your brain is just too many steps ahead of me on the problem

Do you have the algorithm on line? Or was that aspect of things done as part of data entry?

Yours, Confused of Oxon ...

EDIT: Actually, there could be a problem - everyone in the analysis has to have raced against someone in common i.e. some sort of chain must exist between all the results. Otherwise aren't you just comparing apples with pears?

[Ian H]

This thread makes me realise that I am just not serious enough about time trialling.

[David Martin]

The data is captured as three tables in a database. Riders, events and efforts. Efforts links rider and event with a time.

Lets take an example with three events.

Riders 1-4 ride event 1
riders 12 45 ride event 2
riders 1345 ride event 3.

We are interested in rider 1. He has ridden all three events so we pull back data for all those events.
We compare event 1 to event 2 and build a best fit line for relative time in event 1 vs relative time in event 2. There are only three datapoints that can be used (1,2,4)
We compare event 1 to event 3 and can again use 3 points, for riders 1,3,4.
Then average 1-2 and 1-3 to get the value for event 1.

I've written it in PHP, importing the participation data (select * from participations p inner join (select distinct event _id from participations where rider_id=?) q on p.event_id=q.event_id) into a multidimensional array. It is fairly straightforward then to walk that array for each event.

So for each event I just have to upload each event data into the database and the results are recalculated on the fly.

..d

[David Martin]

This thread makes me realise that I am just not serious enough about time trialling.

  It is really a way of looking at the data in a new light. I suppose it gets more value out of each event and could be a way to encourage more participation.

..d

[mattc]

This thread makes me realise that I am just not serious enough about time trialling.

  It is really a way of looking at the data in a new light. I suppose it gets more value out of each event and could be a way to encourage more participation.

..d

What is quite good about this approach is that an occasional rider could benefit:
Hullo. I've ridden the late-May HXZ321 for 3 years now, and I always finish mid-field - am I getting any faster?

If they just looked at their personal times, they could get the wrong impression [this may not be a bad thing ... !].

This is no more or less taking-it-seriously than simply trying to beat a PB - it's just using different data!

(David - thanks, I sort of understand your algorithm - I'll try working through your example properly!)

[inc]

This thread makes me realise that I am just not serious enough about time trialling.

  It is really a way of looking at the data in a new light. I suppose it gets more value out of each event and could be a way to encourage more participation.

..d

But surely the time spent on formula or spreadsheets would be more productively spent trying to improve some aspect of your performance. Since TTs are the purest form of an individuals performance, just you alone against the clock why measure your performance against others, the result sheet does that already. There may be some nice data to play with but what is the point. So x was faster/slower on a particular day/course than y  so what, how does that help your performance.

[mattc]

To optimize performance you need to measure it.

So you try a different bar setup - how do you know if it's helping?

If it's blowing a gale, and you go slower, what have you learned?

[TBH I don't go in for this sort of thing, but clearly
I am just not serious enough about time trialling. ]

[MSeries]

[TBH I don't go in for this sort of thing, but clearly
I am just not serious enough about time trialling. ]

Weren't you the chap who stopped for a pint on a time trial a few years ago ?

[David Martin]

This thread makes me realise that I am just not serious enough about time trialling.

  It is really a way of looking at the data in a new light. I suppose it gets more value out of each event and could be a way to encourage more participation.

..d

But surely the time spent on formula or spreadsheets would be more productively spent trying to improve some aspect of your performance. Since TTs are the purest form of an individuals performance, just you alone against the clock why measure your performance against others, the result sheet does that already. There may be some nice data to play with but what is the point. So x was faster/slower on a particular day/course than y  so what, how does that help your performance.

Well, I know that I am more or less always about the same sort of time as X. I change my position/warm up/strategy. I get much closer to his time. Was it my changes or did he have a bad day?

It is a way of doing that comparison against all riders rather than just a select group.

It is also a reasonable way of identifying the most improved rider over the season.. best times may not be an accurate record of performance given variation in courses and weather.

Obviously you don't have to look at the data. And as for time taken - no more than writing out the results as the code generates the results from the database.

..d

[mattc]

[TBH I don't go in for this sort of thing, but clearly
I am just not serious enough about time trialling. ]

Weren't you the chap who stopped for a pint on a time trial a few years ago ?

That was the 24H - you can stop for a pint and not lose any places.

And it wasn't me - it was FixedWheelNut, IIRC. I stopped by a nice pond to rub my feet in the shade at about the same stage of the event. I still broke the club record. Admittedly I might have gone further by taking it more seriously

[Greenbank]

[TBH I don't go in for this sort of thing, but clearly
I am just not serious enough about time trialling. ]

Weren't you the chap who stopped for a pint on a time trial a few years ago ?

That was FixedWheelNut, but it didn't affect his time at all.

It has bugged me for a while, the question of whether you are really going well, or whether it is the conditions.
......

Not an issue when you race with a powermeter, the info is there.

Power is a good thing to measure, and compare between rides (not even time trials), but it's affected by various things including clothing (leg warmers will sap about 5W of power allegedly) but it does sort out problems of comparing hilly windy rides with flat still rides (200W average a 200W in all weathers).

[gonzo]

Power is a good thing to measure, and compare between rides (not even time trials), but it's affected by various things including clothing (leg warmers will sap about 5W of power allegedly) but it does sort out problems of comparing hilly windy rides with flat still rides (200W average a 200W in all weathers).

Power is the amount of power you produce; it can't be reduced by wearing leg warmers*! What it will do is to add drag.

Most of your power goes into overcoming drag. Because drag increases as speed increases until it is equal to the power you put out, adding drag just means that you reach equilibrium at a lower speed.

*I'm very dubious about that claim as it shouldn't make any real difference.

[Greenbank]

Powermeters measure the power you are managing to get to the rear wheel (or cranks). They don't measure the work you are doing to create that power.

It takes more effort to produce 200W of power wearing leg warmers than without.

As a thought experiment: Put 10 pairs of leg warmers on, so that it's harder to bend your legs, then sit on the bike on a trainer and produce 200W of power. Are you doing more or less work than without the 10 pairs of leg warmers on?

Given that, which is better based purely on power outputs, producing 200W wearing leg warmers or 201W without them?

[giropaul]

I can see how David Martin's proposal could be a useful guide to form, but only in serious Open Events when everyone is likely to be giving it full gas.

For example, in days gone by I rode the occasional time trial (well, when we're young we often do things on our own that really are much better done in company!)

In a typical week I might ride the club 10 on, say, a Tuesday. In this event there was nothing to win; it was training. I usually rode out and rode on my normal road wheels.

On Saturday I might ride an Open. Best wheels, proper warm up, full gas.

One could often hear conversations such as "he only put a minute into me on Tuesday, and he's done a 21 today, therefore I've really done a 22"

WRONG - you did what you did. My effort isn't a standard comparitor.

I guess it's part of the time rather than position mindset in time trialling (that also leads to everyone looking for "fast" courses and standard distances).

In Belgium, France etc if you ride a (rare) time trial the only thing that matters to anyone is finishing position.

[David Martin]

Your point is valid. However it is easy to look along the list and you can then see who was taking it easy on any particular event.

I have a few target events during the year - typically these are the two local 10's and this year should be the club Open 25. But that doesn't preclude the TT series that runs every week from being taken seriously. The results are what they are. The analysis shows a comparison between races and racers - some common sense in seeing which races is required.

..d

[gonzo]

Powermeters measure the power you are managing to get to the rear wheel (or cranks). They don't measure the work you are doing to create that power.

It takes more effort to produce 200W of power wearing leg warmers than without.

Sorry, I'm used to people only talking about aero drag!

[David Martin]

One other area where it would be useful is in assessing the effectiveness of e.g. a different warm up, or position, or a new bit of kit.

Assuming you put the same power into the powermeter, that doen't relate to the same time/relative position if other factors have changed.

..d