Seventeen years ago I wrote "never run on a treadmill again." The central argument was wrong.

You don't even care, do you. I taught this. In rooms. To other trainers, who then went and taught it to their clients, so somewhere out there is a third generation of this nonsense wearing my fingerprints. And you're already scanning ahead to find out what the right answer is. The audacity of you, dear reader.

Fine. Here's what I said.

The belt drags your foot backwards, so your hamstrings and glutes never have to pull your body forward, so you're training a run with a third of it missing. Great story. It has a mechanism, a villain, and a diagram. I drew that diagram with roughly the confidence of a man explaining to a room full of people that the best defence is, in fact, no defence at all.

Here's the problem. During stance your foot is planted and still relative to the belt, and your body travels over it — which is precisely what happens on the ground. The frame moving with the belt is an inertial frame; the mechanics are the same. Somebody wrote that down in 1980 and I did not read it before getting confident.

And when people actually measured it: no significant difference in peak hip extension, across thirty-three studies and four hundred and ninety-four runners. Then a study of muscle synergies found treadmill running had greater activation in the swing-phase modules — the opposite of the machine doing your work. Then a sprinting study found the treadmill produced a faster extending hip.

Three for three, all against me. The references, since somebody always asks and quite right too: van Ingen Schenau in Medicine & Science in Sports & Exercise, 1980. Van Hooren and colleagues, Sports Medicine, 2020. Oliveira and colleagues in PLOS ONE, 2016, for the synergies. McKenna and Riches, Scandinavian Journal of Medicine & Science in Sports, 2007, for the sprinting.

And yes, alright. Feel free to take that one to the group chat. If you'd like to spend a Sunday under a glaring LED light foraging for meaty study chunks in a haystack of gristly, poorly structured faux-studies to check my working — be my guest. I'll leave the light on.

In 2025 a group published data on 5,205 runners across 588,071 individual running sessions, asking a very specific question: what happens when a single run is longer than anything you've done in the past month?

More than doubling your longest recent run carried more than twice the injury hazard.

Look at what that measures. Not weekly mileage. Not your monthly total. One run, against your own recent history.

While we're here — the ten per cent rule. Everybody's rule. I've said it out loud, with a straight face, more times than I want to count. It's been tested: 532 novice runners randomised to a thirteen-week graded programme built on it against a standard eight-week one. Injuries: 20.8 per cent versus 20.3 per cent. A second trial, 839 runners, compared building volume against building intensity and found no difference either.

Both things are true at once, and anyone telling you otherwise has read one of them: big single jumps look genuinely risky, and no tidy weekly rule for avoiding them has ever been shown to work.

Start descending before you need to. Small doses of downhill, or loaded lowering work if the trails aren't open — step-downs, controlled eccentric work. This is the single most specific thing on the list and almost nobody does it.

Strength train. Across sports, a meta-analysis of randomised trials found strength training reduced injury risk to roughly a third of the control rate, dose-dependently. That's across sports rather than runners specifically, and I'll say so rather than quietly upgrading it — but it's the best-evidenced thing here and it's the thing runners skip.

Change one variable at a time. Longer, or hillier, or rougher underfoot. Not all three on the same Saturday because the weather finally broke.

Two things I'd steer you away from paying for first: your running surface, which the systematic reviews have never established as a risk factor, and the acute-to-chronic workload ratio, heavily promoted for years and since fairly comprehensively taken apart.

More on rebuilding in return to running, and the descending argument in longer form on the hikers page.

One footnote I enjoy more than I should: the landmark study everybody cites about increasing your cadence to reduce knee load — the one that found a five per cent increase cuts energy absorbed at the knee by around twenty per cent — was conducted on a treadmill.