RPE is a count of reps in reserve, not a vibe. Here is how a half-point of drift compounds across a block, and what reading it honestly looks like.
The single that wasn't a 9
Three reds on the third top single and the lifter racks it, chest heaving, and writes 9 in the log. The bar tracked at the same speed as the opener. Two more were there, maybe three.
Nobody lied. He felt a 9. His heart rate said 9, his grip said 9, the chalk cloud said 9. But reps in reserve is a count, and the count was two. He just bought himself an extra deload he didn't need, because next week's percentages get calculated off a number that was half a point too high.
This happens on the overwhelming majority of logged RPE sets, and it happens to people who have run RTS, 5/3/1, and a full RP hypertrophy block. Knowing the scale cold does not make you accurate on it.
The half-point tax
Call it the half-point tax. A serious lifter, asked to rate proximity to failure on a hard set, will land about half an RPE point above where the bar actually puts him. He calls the 8 a 9, the 7.5 an 8. The error is small per set and it is directional, which is the dangerous combination: it does not cancel out across a block, it accumulates.
The reason it survives years of training is that RPE feels like the one honest number in the log. Load is given by the plates, reps are given by the count, but RPE is the lifter reporting on himself, and self-report is exactly where the quiet, consistent bias lives.
RPE is a measurement, and most lifters treat it as a confession
The dominant way RPE gets taught is as a feeling: how hard was that, on a scale of one to ten. That framing is why the tax exists. Effort is a sensation, and sensation spikes near the top of any heavy set whether or not you are actually close to failure. Adrenaline, breathing, and the psychological weight of a grinder all read as high effort while you still have clean reps in the tank.
Reps in reserve is the opposite kind of question. It is not how hard did that feel, it is how many more could you have made with good bar speed. That is a prediction about a physical fact, and predictions can be checked. The lifters who rate accurately stopped asking the feeling question years ago.
What the reps-in-reserve research actually found
Mike Tuchscherer brought the scale into powerlifting through Reactive Training Systems and anchored every number to reps in reserve: a 10 is a grinder with nothing left, a 9 leaves one, an 8 leaves two. Eric Helms and Mike Zourdos did the validation work. Zourdos and colleagues published the RIR-based RPE scale in 2016 and tested how closely lifters' ratings matched the reps they actually had left.
The pattern across that literature, and across the field reports Greg Nuckols summarizes at Stronger By Science, is consistent. Novices are noisy in both directions. Trained lifters get sharper as they close in on failure, but a meaningful share still stop a rep or two short of where they believe they are. The direction of the error tends to run one way: lifters more often overestimate how close to failure they were than underestimate it, meaning sets end earlier than they needed to. Mike Israetel's RP autoregulation guidance leans on the same reality from the coaching side: RIR targets only work if the lifter's RIR estimate is honest, and the common failure mode is calling sets harder than they were and leaving prescribed volume on the table.
Watch it compound across a block
Take a lifter on a 5/3/1-style top set plus RPE-capped supplemental work, autoregulating his back-off volume to an RPE 8 stop. Week one, he hits the cap at five reps and racks it, calling 8. The bar said 7. He left two clean reps, which across three supplemental sets is six reps of stimulus he programmed for and skipped.
Week two, same load feels the same, same call, same shortfall. By week four his log shows a tidy, disciplined progression of RPE 8 sets, and his actual proximity to failure has been drifting around 7 the whole time. The numbers in the journal look like textbook autoregulation. The training underneath them was a notch too easy to drive much adaptation, and he never saw it, because every individual rating was defensible and the bias only shows up in aggregate.
The other direction exists too and bites harder. A lifter coming back from a layoff overrates the wrong way, calls a true 9 an 8, keeps grinding past his intended stop, and walks into junk fatigue he can't explain.
Honest RPE is a logging habit, not a personality trait
If the tax is directional and small, the fix is not willpower, it is calibration against your own record. The single most useful drill is the AMRAP audit: a few times a block, take a set you would normally cap at a given RPE and instead push it to true failure with a spotter or in the rack. Count the reps you actually had past your stop. That gap, the reps between where you called it and where it ended, is your personal half-point tax, and it is usually steadier than you would guess.
Once you know your bias, you read your own ratings through it. You stop trusting the feeling at the top of the set and start trusting the bar speed and the rep count you logged the last three times you saw this weight. Honest RPE is less a moment of self-awareness on the platform and more a thing the log tells you afterward, if the log kept enough to tell you anything.
Where the number has to live
A logged RPE only earns its keep when something can hold it against the reps you actually hit. Write 9 next to five reps at 315 this week, and 9 next to five reps at 315 four weeks later, and a record that keeps both rows shows you the rating stayed put while the load stopped moving. That standstill is the half-point tax made visible: same effort score, no progress underneath it.
Platepusher logs the RPE in the same row as the load, the reps, and how the bar behaved on the set, so the drift surfaces as a pattern across the block instead of a feeling you have to trust in the moment. The math runs while you re-rack. It does not tell you what number to write. It just keeps every number you have written where the next one can be checked against it.
What we're tracking next
The open question is whether the tax shifts with bar speed data in the loop. If a lifter can see velocity drop-off next to his RPE call over a block, does his self-rating tighten toward the bar, or does he keep deferring to the feeling at the top of the set. That is a measurable thing across enough logged blocks, and it is the next piece worth writing once the records are there to read.
If you log RPE every session, the honest version is worth keeping where it can argue with your past sets. Run a block through Platepusher and let the number get checked.
Every claim here traces to the people who built and validated the scale: Tuchscherer's RTS, the Zourdos and Helms reps-in-reserve work, Nuckols' summaries, Israetel's autoregulation guidance. This is the reading a programming-literate lifter does anyway, gathered into one place and pushed one step further into what the bias does to a block.
