Your log knew about the injury three weeks before your shoulder did.

The tweak that wasn't random

The shoulder let go on a Tuesday. Bar was a weight you'd overhead pressed for reps since spring, and it went mid-rep like a fuse blowing, no ache walking in, no bad night's sleep, no obvious warning. Random, is the word most lifters reach for.

It wasn't random. Open the log and scroll back a month and the whole thing is sitting there in the numbers. Pressing volume climbing week over week, a ramp you'd never have programmed on purpose. Your logged RPE creeping upward on the same loads, then dropping off a cliff the week before, not because you got fresh but because you'd quietly started grinding and cutting sets short. A deload penciled in and skipped, then a second one skipped on top of it. None of that felt like anything at the time. All of it was written down.

That's the uncomfortable part. The signal was already in your possession. You just never read the log backward.

The four-session autopsy

Here's the method, and it takes about ten minutes. Find the date of your last two injuries, the real ones, the ones that cost you weeks. For each, pull the four sessions immediately before it and read only those, in isolation, for the movement pattern that got hurt.

You're not looking for the session that broke you. The last session is almost never the cause, it's just where the accumulated bill came due. You're looking at the three or four weeks of runway before it, because overuse injuries are a loading story, and loading stories play out over blocks, not single days.

Do it for two separate injuries and something happens that a single case can't give you. The overlap shows up. The same shape of ramp, the same RPE behavior, the same skipped recovery, sitting in front of two different body parts that failed a year apart. That overlap is your personal injury signature, and it's specific to you in a way no generic 'listen to your body' advice can be.

The last session is almost never the cause. It's just where the accumulated bill came due.

Why do I keep getting the same injury?

Because you've been reading your history as a trophy case, not a case file. Almost every 'your training log is a personal dataset' take, and the framing has real momentum across T-Nation, StrongFirst, and the endurance-Outside crowd, treats old logs as a PR archive. Scroll back to admire the numbers that went up. Pull the e1RM curve. Celebrate the volume you survived.

That's the low-leverage read. Your PRs are the outcome you already know. The injuries are the outcome you keep repeating, and repetition is the tell that a pattern, not bad luck, is driving it. A lifter who tweaks the same lower back every eighteen months isn't unlucky twice. They're running the same loading mistake twice, and the log recorded both runs in full.

The lifters who train for decades without stacking up injuries aren't blessed with better joints. Plenty of them have learned, often without naming it, to recognize their own runway before the tweak. The autopsy just makes that recognition explicit instead of intuitive.

The three signatures that keep showing up

When lifters actually run the autopsy, the same three shapes surface again and again. None of them require a lab, a wearable, or a coach. All three are already in a decently kept log.

The first is the volume spike: total hard sets or tonnage for a pattern climbing faster than your tissue's adaptation can keep pace with, usually a jump you made by feel during a good stretch, not one you programmed. The second is the RPE inversion: your logged effort on fixed loads drifting up for a couple of weeks, then dropping suddenly, which reads as recovery but is often compensation, shorter range, more body english, a set called at five reps that used to be eight. The third is stacked skipped deloads: not one missed off-week, but two back to back, the recovery you deleted because you felt fine right up until you didn't.

One of these on its own is noise. Any given block has a heavy week. It's the convergence, two or three of them stacking in the same three-week window on the same pattern, that's the readable signature.

One lifter, two autopsies

Take a lifter three years into 5/3/1, squat and deadlift on the same block. Left knee flared last October, hip on the right side flared this June. Two different joints, two different times, filed under bad luck.

Run the autopsy on October first. The four sessions before it show squat volume he'd bumped himself, adding a fifth and sixth back-off set because the weights felt easy, so his logged squat tonnage over three weeks was well above anything in the prior block. RPE on his top singles read 8, then 8.5, then a suspicious 7 the week before, the week he later remembers rushing between sets. His scheduled deload sits in the log with every set logged as completed at full load. He trained through it.

Now June. Different lift, identical shape. Deadlift frequency crept from once to twice a week during a program he was 'auditing.' RPE inverts the same way, up then down. The deload before it, skipped, and the one before that, also skipped. Two joints, one signature. The knee and the hip were never the story. The runway was.

What changes when the log is a case file

If the autopsy holds for you, and the only way to know is to run it, then injury prevention stops being a vague virtue and becomes a specific, personal watchlist. You stop trying to avoid 'overtraining' in the abstract. You start watching the two or three numbers that actually preceded your tweaks.

That's a different relationship with the log. Forward-looking deload detection, catching fatigue accumulating in real time, is its own worthwhile read, and a separate one. The autopsy is retrospective and diagnostic: it tells you which signals to trust in the first place, because it shows you which ones cashed out in injury when you ignored them last time. One builds the watchlist. The other watches it.

And it reframes what a training history is for. Four hundred logged sessions aren't just a record of what you lifted. Kept honestly and read backward, they're the closest thing a lifter has to a controlled study of one, and the injuries are the most expensive data points in it.

The instrument only works if it kept the details

The autopsy has one requirement, and it's brutal: the data has to actually be there. You can't read RPE drift across the four sessions before a tweak if you never logged RPE, or if your history got re-zeroed when you switched apps last year, or if the log rounded your accessory work down to 'shoulders, done.' Most lifters discover mid-autopsy that the exact detail they need is the one they didn't capture.

That's the quiet argument for keeping a log that respects the details and keeps them yours across years and across apps. Per-set loads and reps, RPE where you tracked it, deloads marked as planned versus actually taken, all of it preserved as native data when you bring a history in rather than flattened into a summary. Platepusher is built for the lifter who's going to want to read backward five years from now, so the instrument records what the autopsy will eventually ask for and gets out of the way while you train. The signal was always going to be in your history. The only question is whether your log kept enough of it to read.

What we're reading next

The open question is how few sessions you actually need. Four is a workable default, but for high-frequency lifters the runway may be shorter and for low-frequency strength work it may run six or eight sessions deep. The autopsy is the same either way. The window that best predicts your signature is the thing worth finding, and the only place to find it is your own history.