reckoner
Reps in reserve is the simplest number in training and it asks one question: when you racked the bar, how many more could you have done? Two reps left is two reps in reserve. Nothing left is zero.
If you have met RPE in a lifting context, you have met this already. The resistance training scale is built out of reps in reserve rather than out of breathlessness: RPE 10 is nothing left, RPE 9 is one rep left, RPE 8 is two, and so on down1. They are the same number read from opposite ends, which is why an argument about which to use is mostly an argument about arithmetic.
The question worth spending time on is not what the number means. It is what the number should be.
Strength barely cares. Size does
This is the most useful finding in the area and it is recent. A series of meta-regressions pooled the training studies and treated proximity to failure as a continuous variable, measured in reps in reserve, rather than as a failure-or-not switch. For strength, the relationship came out negligible: the confidence intervals contained zero in every best-fitting model, meaning strength gains held up across a wide spread of how close people stopped. For muscle size they did not, and size increased as sets were taken closer to failure4.
The authors are careful about it and so should anybody quoting them be. The reps in reserve behind each study had to be estimated from the training descriptions, the models fitted only modestly, and the analysis is described by its own authors as exploratory4. It is the best answer available. It is not a settled one.
Set beside it, a review of 15 studies had already found that training all the way to momentary failure was not superior to stopping short of it for muscle size3. Those two fit together better than they look. Closer is better for size, and the very last rep is not the one buying it.
Train close enough that the set is asking something. Going all the way to failure buys less than it costs.
The count in your head leans in a particular direction
Everything above assumes you can tell how many you had left. You can, roughly, and the size and the direction of the error are both worth knowing.
Pooling 414 people across twelve studies, a meta-analysis of prediction accuracy found people under-predicted the reps they had left by about 0.95 of a repetition on average2. Under-predicted: they thought they were closer to failure than they were. The set you called two reps short was likelier to have been three.
Two things made people better at it. Predicting closer to failure helped, and so did heavier loads: on sets of twelve reps or fewer the error was much smaller than on longer ones2. Both point the same way. Nobody can count what is left from ten reps out, and nobody needs to.
One thing did not help, and it cuts against the received wisdom. Training status made no apparent difference to prediction accuracy in that analysis2. The scale's own originating study had found experienced squatters more accurate than novices at a one repetition maximum1, so the two do not agree, and the honest position is that experience probably helps a little and certainly does not settle it. Nobody should treat their own count as precise on the strength of a training age.
Why a rating beats a percentage
The alternative to rating a set is being told to lift a percentage of your one repetition maximum. Seventy per cent for eight, and so on. It sounds more objective and it is more fragile, for two reasons.
The first is that the same percentage is not the same amount of work for two people. How many reps somebody can complete at a given share of their one repetition maximum varies between individuals, so a table of repetitions allowed at each percentage describes an average nobody actually is5. A rating asks about the set in front of you instead, which puts everybody on the same footing whatever their own numbers look like5.
The second is that it needs a one repetition maximum you have measured, and then it needs that number to still be true. It moves with sleep, food, stress and how the last two weeks went. A percentage of a test you did in March is a prescription for the lifter you were in March.
None of which makes percentages useless. It makes them a planning tool rather than a decision taken under the bar, and the two work together: a load chosen from a percentage with a target rating reported against it is an arrangement the coaching literature recommends outright5.
How to use it in a session
- Count reps, not suffering. The question is how many more you could have completed, not how unpleasant the set was. A heavy set of three and a burning set of twenty can both leave two in the tank.
- Rate the last set, standing there. The earlier sets were done with more left on purpose, and an estimate made on the way home is an impression of the session rather than a count.
- Assume you had one more than you thought. The average error runs that way2, so if your sets keep coming in easy and nothing is moving, the likeliest explanation is that you are further from failure than your own number says.
- Keep the last rep of all for rarely, and never on a lift over your neck. Failure adds little for size and nothing reliable for strength34, and it costs a great deal in fatigue and in technique.
A number you can report honestly every week is worth more than a precise one you can report occasionally. Rough and consistent is the whole job.
What Reckoner does with it
Reckoner asks for the effort on each lift as you finish it, on the same scale, and reads it beside the reps you actually completed. What the rating changes, what it does not change, and the four ways lifters get the number wrong are set out in what RPE actually measures.