The sub-three dream, take two

I should be honest about the target before anything else, which is that a sub-three marathon is probably beyond me, and I know that. My only marathon was Chicago back in 2009 in 4:21:49, and I am not one of those runners who arrives at this sort of goal with years of natural speed behind them. But I have always found that reaching for something slightly ridiculous carries you further than setting a target you already know you can reach, because if I aim at three hours and end up at three fifteen I will take that gladly, whereas if I aim at three fifteen I will find perfectly reasonable excuses to settle somewhere the wrong side of three thirty.

Nine weeks that genuinely worked

The plan was twenty-six weeks starting at the end of April and pointing at October, and I got nine weeks into it, which is roughly a third of the build and enough to see what was actually happening.

When I started, my Garmin thought I was a 4:09:57 marathon runner, and by week nine it had me at 3:26:08, which is forty-three minutes and forty-nine seconds faster and around two-thirds of the way to sub-three. What that prediction actually means is 7:52 per mile, which happens to be precisely the pace of my half marathon personal best from September 2025, so the watch was telling me I could hold my half marathon pace for the full distance, and I am still not entirely sure I believe it even now.

The physiology underneath moved with it, which is the part that mattered more to me than the prediction. VO2 max climbed from 50 to 54, threshold pace came down from 7:29 per mile to 7:06, and my heart rate at threshold dropped from 169 to 161, which is the detail I was most pleased about, because it meant I was running faster at a lower effort rather than simply gritting my teeth and hurting more. Resting heart rate went from 58 down to a best of 48. I completed thirty-four of thirty-seven scheduled sessions, ninety-two per cent adherence, which is considerably better than I would have bet on beforehand.

More than any of the numbers though, I felt genuinely good. Runs that had been hard in April were comfortable by June, paces that needed concentration became conversational, and I finished sessions with something left rather than emptied out. That is the thing nobody tells you about a proper training block, that the real reward is not the splits on the watch but the growing sense that your body is doing what you ask of it without complaint.

The five miles that cost me everything

In early July I went out to do five miles of tempo at under seven minute pace in thirty-five degree heat, and I knew it was wrong somewhere around the second mile, and I did it anyway. That is the whole lesson in one sentence, so let me explain why I did it, because I suspect a lot of people reading this will recognise themselves.

The session was written in the plan. It said five miles at that pace on that day, and I had ninety-two per cent adherence that I was quietly proud of, and I did not want to be the person who let the number slip. So I overrode what my body was telling me and what my head was telling me and finished the prescribed session in conditions that any sensible person would have recognised as unsuitable, and I got heat exhaustion for my trouble.

This is where I think training plans, including the good ones and including the ones built with clever technology, have a fundamental limitation. A plan can adjust for how you slept, for what you ate and when, for your recovery score and your heart rate variability and your training load over the last twenty-eight days. What no plan can do, and I do not think any of them ever will, is know how you actually feel before a run, during the third mile, or in the hours afterwards. Only you know that, and the plan is not going to ask.

The temptation, when you are the sort of person who likes completing things, is to treat the written session as the authority and your own judgement as the excuse. I did exactly that, and the cost was not the five miles. The cost was nine weeks of genuinely hard work, the whole remaining block, and for a while, more importantly, my enjoyment of running altogether. Once that goes everything else follows, because nobody sustains four or five sessions a week at something they have stopped looking forward to.

Looking back at the data, the warning was there before the heat got me. Through week seven I had missed one session out of twenty-four, and by week nine that had become three out of thirty-seven, while my resting heart rate had given back half of its improvement and drifted from 48 back up to 53. The block was still getting faster on paper while the foundations underneath were already going, and I did not read it at the time.

After that, work took over. Run Love Sprint needed everything I had and running became the thing that moved rather than the thing that got protected, which is an easy trap and one I walked into without much resistance. The app was a real demand on my time and it was also a very convenient excuse, and both of those things can be true at once.

Starting again, and starting ahead

So we go again, and the first run is already done, which felt good in a way that I was half expecting not to. Better still, I am starting this block a little ahead of where I was at the end of April, because the work from the summer has not entirely disappeared even after the time off, so the hope is that things come back to me reasonably quickly as the miles build.

For the next week or two I am running whatever Garmin suggests, without pressure attached, just to put some consistent weeks underneath me. Once that is ticking along I will pick the proper plan back up, the one I built with Claude, and find out whether AI really can train me to a sub-three marathon.

Two things are in my favour this time. The weather is turning, and cool autumn mornings are an entirely different proposition to tempo work in a July heatwave. And I now know precisely what my failure looks like, which means I have a decent chance of seeing it coming, because the warning was never the heat exhaustion itself but the missed sessions and the creeping resting heart rate in the fortnight before it.

The bigger change is how I intend to treat the plan. It is a guide rather than a contract, and if a session is wrong on the day, for the heat or the sleep or simply because my legs have nothing in them, then it gets moved or dropped and the week carries on. Ninety-two per cent adherence is worth nothing at all if the missing eight per cent is the session that ends the block.

What I expect

I will probably not run a sub-three marathon, and I am at peace with that. But nine weeks of proper training took forty-four minutes off my predicted time and I stopped with seventeen weeks still to run, so if I can keep this going and stay honest with myself about what my body is telling me, I should end up somewhere close to where I want to be, and that is worth the early starts.

I will keep publishing the numbers either way, the poor ones included. That approach worked well enough last time, right up to the point where I had to publish a letter explaining why the plan had stopped.