More hours didn’t make trained cyclists faster.

Training debates have turned into identity statements. Polarised. Pyramidal. Threshold-heavy. Or the fallback advice: just ride more. If you’re already fit and training consistently, you’ve probably tried at least one of these, then hit the same wall anyway.

A 2025 meta-analysis pulled together 41 intervention studies in trained cyclists, covering 81 training groups and 797 riders. These were not beginners. To be included, athletes had to be recreationally trained or better, with VO₂max ≥ 59 mL/kg/min, and they had to complete programmes lasting longer than four weeks. This is the population most people reading this recognise themselves in.

The results create real cognitive dissonance. Training works, clearly. Across studies, VO₂max improved (g ≈ 0.42) and time-trial performance improved (g ≈ 0.39). But the way the training was labelled mattered far less than expected. Polarised and non-polarised approaches produced similar gains, with no clear winner. Longer interventions were linked to larger improvements. Weekly volume and total volume were not meaningfully associated with how much riders improved.

Read that again. Once riders were already trained, adding more hours did not reliably explain who got faster. The idea that the next jump comes from more volume, or from switching to a “better” intensity model, simply wasn’t supported at this level.

The practical takeaway is uncomfortable but freeing. Stop chasing the perfect distribution and start chasing repeatability. Pick an intensity mix that lets you execute the hard sessions well, keep the easy work genuinely easy, and recover enough to train again tomorrow. Then commit to it long enough for adaptations to accumulate. Less weekly tinkering. Fewer resets. More uninterrupted weeks.

If you want something concrete to anchor this to, use a consistent time-trial test every four to six weeks, same warm-up, same conditions as close as possible. Watch the trend, not a single result. The advantage doesn’t come from a new model. It comes from staying in the work long enough for it to show up.

Reference

Cove B, Chalmers S, Nelson MJ, Anderson M, Bennett H. The effect of training distribution, duration, and volume on VO₂max and performance in trained cyclists: A systematic review, multilevel meta-analysis, and multivariate meta-regression. Journal of Science and Medicine in Sport. 2025;28(5):423–434. doi:10.1016/j.jsams.2024.12.005.

On my mindA small moment this week

I had a couple of good calls with athletes this week. One was very much about training and expectations. The other ended up being more of a conversation about AI and how I’m using it, which keeps coming up more often now.

I want to be pretty open about this. I’m not using AI to do the work for me. I’m not handing decisions over to it. What’s happening is more that I’m coming into it with a lot of years of experience and a fairly clear sense of what I think already, and then using it to push on those ideas a bit. To see where they hold up, where they don’t, and where I might be missing something obvious.

That’s the part I find interesting. It helps me say things better. Sometimes it helps me realise I’ve been thinking something for years but never quite put it into words. Other times it just shows me a blind spot faster than I’d get there on my own.

From an athlete point of view, that actually matters. The thinking gets done quicker and there’s less unnecessary chopping and changing. I’m not less involved, if anything I’m more involved, just spending time on the parts that actually move things forward.

If any of this raises questions or concerns, you can always reach out. I’m happy to talk it through properly. Right now, it feels like a really good place to be. Experience, curiosity, better tools. It makes the work more interesting, and it’s genuinely helping athletes, which is the only bit I really care about.

Damian