Submarine Smashed

Project “Submarine” is done. On July 26th I crossed the finish line of a 10k race in 37:18 and with that closed up the project I started back in February. Time to break down how the final weeks played out, settle the score on both goals and announce what’s next.

If you missed the first post, Submarine, So Far covers the background: the goals were running 5 kilometres in under 18 minutes and 10 in under 38, following an LLM-generated training plan with roughly 11–12 hours of weekly training. That post ended mid-May with a 19:55 benchmark over 5k, paces recalculated and an effective VO2max climbing at about +0.8 per week.

From benchmark to benchmark

At the end of May I ran the first official 10k of the project. It was planned as a controlled competition to test how the 5k-focused training transfers to the longer distance. I finished in 40:31, right at the fast end of the projected 40:30–41:30 window. Controlled it was (and hot too) and it confirmed the transfer worked.

Then came the June benchmark, the last one before the racing block. My timeline projected 19:00–19:20. I ran 18:23. That’s 92 seconds off my April time in seven weeks and the moment the timeline properly broke in my favor. The training paces derived from that run put my threshold sessions at 3:57–4:05/km — faster than what used to be my all-out 3km pace not that long ago. Funny how quickly “impossible” paces become training pace. Also, this was the time I figured I could go on vacation in August xD.

The 5k: sub-18, with an asterisk

In July I started at a local 5k race to attack the sub-18. This is where it gets a bit messy. My Apple Watch clocked 17:56 at exactly 5.00km — sub-18, boom. The official net time however read 18:30 and my watch measured 5.2km in total. That’s a 4% deviation from the certified distance. I asked Claude how that happens and the best it could offer was “not running the ideal line”, which I can’t quite picture being worth 200 metres on a 5k. My best remaining explanation is: GPS, as it was quite crowdy and the course led through a forest.

So officially, no sub-18. Personally? I clocked 5 kilometres in 17:56, one month ahead of the timeline I set in February and I’m counting it. The asterisk stays in this paragraph and doesn’t make it into my head :).

The finale

With the 5k goal ticked off, I switched my entry for the final race from the 5k to the 10k. One goal left, one race left.

I tapered into race day following my competition-week protocol and finished 12th overall in 37:18 — sub-38, smashed. The race itself was a mental rollercoaster though, thanks to my watch once again. It read a pace of around 3:50/km throughout and at the finish line it showed 9.74km, telling me I still had almost 300 metres to go. This time the GPS drifted in the opposite direction. So mid-race I genuinely believed sub-38 was out of reach, next to feeling some limits in my legs that day. Then the official results came in above my expectations. Maybe the watch had been messing with my perception the whole time.

Looking at the splits (with due GPS respect): out in 3:52, hovering between 3:45 and 3:55 through the middle, the slowest kilometre at 3:58 coming at kilometre 9 right when the doubt peaked and then closing at 3:42/km with my heart rate maxing out at 96%. Not the negative split masterpiece from April, but an honest, evenly-paced fight.

Ahead of schedule

Here’s the part I’m most proud of: my plan reserved sub-38 for the peak phase in August/September, with a target window of 37:20–38:00 and a main race plus a backup. I ran 37:18 in July, during what was supposed to be peak preparation. Both goals fell a month early and the peak phase became obsolete before it started.

At this point ending in July wasn’t purely a choice of triumph though — I had planned a vacation in August anyway (or rather, a different kind of running project involving trails, but that’s a story for another day). So July 26th “had” to be the finale. Sometimes external deadlines are the best coaches.

Submarine in numbers

Five months of project “Submarine” sum up to 130 runs, 1,221 kilometres, 4 days and 11.5 hours of pure running time and 11,364 metres of climbing. 3 pairs of running shoes and 31 strength sessions.

The number I find most satisfying closes a loop from the very beginning of the project. Back in February I used the Daniels/Gilbert Formula to calculate the effective VO2max I would need: 56.4 ml/kg/min for the sub-18 5k, 55.1 for the sub-38 10k. I started the project at around 44. My effective VO2max now sits at 56.5, having peaked at 57.3 — almost exactly where the math said it needed to be. The plan wasn’t just motivating, it was numerically on point, too.

Since that finish line I have not been street running at all, deliberately giving my body time to release the stress of five months of continuous training. Seeing my weekly structure without three quality sessions feels strange, but the break is part of the plan too.

Some thoughts

For me personally this “project” is not only about numbers. While for the average runner I’d consider my results pretty great, I am far away from national levels. What counts for me was the challenge after some absence of serious endurance training and competition. Most of the mornings I looked forward to my run and running in different environments (location, route, climate) contributed to that. Working with an LLM on this was also quite fun and as expected continuously sticking with the plan worked out great.

Mentally I felt great after each run. Physically I can’t say that confidently :). Running without music, focusing on my breath and combating the one or the other demon in my head mid-run, helped me winding down the mind without forcing or distracting it.

What’s next: “Revenge”

In the first post I hinted at the next project and it’s time to announce it. Remember that half marathon from 2024 — targeted sub 1:45, finished in a catastrophic 1:51? Project “Revenge” is exactly what it sounds like. Starting in September, running three races across Europe. The goal: run a half marathon under 1 hour and 17 minutes (3:39min/km).