Damn lies and spreadsheets.

I was entering my commute into the bikejournal tonight and wondered whether I'm getting faster or slower over time. One wonders. There's opposing forces at work here: firstly I'm riding more, secondly I'm getting older. Which one wins? (I'll leave aside the fact I moved from flat Victoria to hilly Tasmania halfway through.)

As it turns out, the question is more complex than I would have thought. I assembled all my ride data into a single spreadsheet. There lies the details of more 500 rides totalling more than 1,000 hours (two more milestones just recently passed!). OK, how to proceed? I'll graph the buggers!

The first graph is ugly. It doesn't show anything much at all. The regression line for my average speed is flat. Odd. Ok, so I chuck out the 20 slowest rides, clean the data up a little. Now the curve slopes down. It's just barely detectable, but it slopes down. Hooray: I'm getting slower over time.

But wait. The graph shows a bunch of slow rides in July and August last year. Hang on that's when I was fanging around Beijing on the folding bike. Surely those rides don't count? Out comes July and August 2008. And guess what? The curve is upward. Just slightly, but upward. Grotesquely enlarging the graph makes the slope steeper. Cool. But by how much?

Very close analysis reveals the regression line through my rides slopes upwards from 21.85km/h at the left hand end to 22.0km/h at the right hand end. Over four years I'm 150 metres faster an hour. So if the old me raced the new me for an hour on an Olympic velodrome, he would be be close enough to yell abuse but wouldn't have lapped me. Still, any progress is good.

In the end, there is no real answer. Five years after getting back on the bike I'm fitter, happier, healthier and more handsome - all things numbers can't measure. The moral of the story is as always: ride for fun. They haven't yet made a cycle computer that can measure that.

711km so far this year.