The previous post in this series ended at the result. Across 56 D2 matches in 2025, Kintetsu won the closing twenty minutes by a margin (plus 8.1 per match, net) that no other team came close to. The natural follow-up question, from anyone who programmes for a living, is how the training year was set up to support that.
I have access to weekly squad-level GPS data, position-group splits, and the medical sheet across the full club year. This post walks through what the training programme actually looked like, week by week, and ends with one finding I went back and forth on before deciding to publish.
The annual curve
Squad-mean weekly distance, club week 2 through 43. Pre-season, in-season, late season, off-season. The first thing to notice is that the pre-season-to-in-season drop is shallower than the textbook prescription. Most professional programmes drop weekly volume to 60 to 70 percent of pre-season values once the season starts. Kintetsu's in-season squad-mean was 85 percent of pre-season.
In raw numbers, pre-season averaged 17,709 metres per active player per week across weeks 2 to 7. The in-season mean across weeks 8 to 30 was 15,123 metres. That's a 14.6% drop, not the 30 to 40% drop that a "match weeks are recovery weeks" model would predict.
The peak in-season week (wk 16, 20,149 metres) sits within 33 metres of the highest pre-season week. Match weeks weren't recovery weeks. They were maintenance-volume weeks with a quality bias.
This is the part of the data that makes me think the closing-twenty number isn't an accident. If you want to hold high-end aerobic and repeated-effort capacity into late-game minutes, you need to keep training those qualities through the season. The programme did.
Position-group retention
The squad-mean number hides a useful detail. Different position groups run different distances and accumulate different work, and the question of "did we maintain volume" is really four questions, one per group.
All four groups retained 80 to 91 percent of their pre-season volume through the in-season block. Halves and outside backs sit at the top of the absolute distance numbers, as expected. Tight-five forwards run the least in absolute terms (15,227 metres pre-season, 13,523 in-season), but their 89% retention rate is on par with the rest of the squad. The forwards weren't training a different programme, in volume terms; they were training the same programme scaled to their match demands.
The high-speed-running retention numbers (67 to 83%) are slightly lower than total-distance retention. Speed-specific work peaks in pre-season and settles into a maintenance band through the season. That's the expected pattern, and it's the right way around if your goal is to be physically prepared at week 30 rather than week 4.
For a head of performance evaluating a programme, this is the chart I would lead with for "did we hold our work through the season." All four groups, comparable retention, retained in the right direction. The programme did what it was designed to do.
Deload weeks: when and how often
Across the in-season block, five weeks dropped to under 60 percent of typical squad-mean distance. Club weeks 8, 9, 17, 24, and 28. Two of those (weeks 8 and 9) sit back-to-back; the remaining three are spaced roughly a month apart through the middle of the season.
If you ran one deload every six matches, on a fourteen-round season, you'd expect roughly two to three deload weeks. Five is on the higher end of the spec. The programme was built to deload more often than the league average, which is consistent with the closing-twenty thesis: the cost of deloading is some training stimulus; the benefit is recovery to support late-season output.
That's the textbook reading. The data doesn't quite support it.
The deload paradox
I went back and forth on whether to publish this part. The textbook expectation is that deload weeks reduce injury rate. The Kintetsu data, in 2025, shows the opposite.
Across the 23 in-season weeks, deload weeks (5 weeks) had a higher new-injury rate than non-deload weeks (18 weeks). Per player per week, 0.072 against 0.046. A 56 percent relative increase in injuries adjacent to deload weeks.
There are three plausible readings of this number, and I want to walk through each because the honest version of "what does this mean" is "we don't know yet, here are the three things it could be."
Reading one is that the deload was reactive, not planned. If the coaching staff added deload weeks in response to clusters of injuries that had already occurred, the apparent association reverses cause and effect. The injuries don't follow the deload; the deload follows the injuries. This is the most likely explanation for at least some of the five weeks. Looking at the medical sheet around weeks 8 to 9, there is a cluster of soft-tissue events that pre-dates the deload. The deload was, in part, a response.
Reading two is that the bounce-back from deload caused soft-tissue injuries. When a squad returns from a 60-percent-of-typical week to full training volume, the rate of change in workload week-on-week can be high enough to spike injury risk on its own. The classic acute-to-chronic ratio model would predict exactly this: a sharp drop in chronic load makes the next week's acute load look disproportionate. If the team came back from deload to full training without a graded ramp-up, soft-tissue injuries in the week or two following the deload would be the expected outcome. This reading would mean the deload concept is right but the post-deload ramp was too steep.
Reading three is that the sample is small enough to be noise. Five deload weeks out of 23 in-season weeks is a small denominator. Seventeen events versus 46 is the kind of count where a single match week's events change the rate by 5 to 10 percent. The 95% confidence interval on the rate ratio is wide enough to include "no effect." With one season of data on one club, the data alone cannot rule this in or out.
I think the honest answer is "some combination of all three, in proportions we cannot estimate from one season." Reading one is partly true for at least the early-season cluster. Reading two is the most actionable if it's true; a graded ramp-up from deload weeks is cheap to implement and easy to test in 2026. Reading three is the right stat caveat to keep on the table.
What I would change
A few things, with the closing-twenty thesis intact.
Track the post-deload ramp explicitly. Set a rule, not a target, on the percent change from deload week to the following week, and watch it across the season. If the post-deload week is itself an 85-percent-of-typical week with a graded return to full volume the week after, the ramp problem is gone.
Keep deload-week count where it is, and avoid clustering them. The two back-to-back deloads (weeks 8 to 9) coincided with the first injury cluster of the season. I do not have enough data to tell whether they caused, contained, or merely accompanied that cluster, but a single-week deload followed by a normal week is easier to reason about and easier to recover from than two weeks of low load back to back.
Hold the in-season retention number. Eighty-five percent retention is the right neighbourhood for a programme that wants to be physically prepared at week 30. The closing-twenty result is downstream of that retention, and the data supports continuing it.
What this post can't tell me
This is one season at one club. I do not have league-average periodisation data to benchmark against, and I would want at least two more seasons before treating these magnitudes as stable. I also do not have direct injury-risk modelling on the GPS variables; the rates I report here are unadjusted for player position, age, prior history, and match minutes. A proper injury-risk model on this dataset is a piece of analytical work I'd want to run before recommending a programme change.
What I can defend is the descriptive picture. Kintetsu's 2025 programme was a high-retention, multi-deload, in-season-heavy programme. It produced the closing-twenty net result reported in the previous post. And it produced one finding that pushes back on the textbook deload story. All three things are in the data, and all three deserve to be in the post.
Coming next: the position-group splits in detail, and the case for evaluating in-season programmes by what they produce in the closing twenty rather than by what they save in load-monitoring totals.