How to judge reps in reserve: a practical method for rating sets honestly

Why the estimate matters

Reps in reserve turn a set into a comparable number. Method folds the RIR you log into the estimated 1RM, compares that estimate across sessions for records, and uses your last set's RIR to suggest the next load. If your ratings wander, those numbers wander with them. None of that needs you to be perfectly right. It needs you to be consistently you.

A calibration routine

1. Pick a safe, stable movement. A machine press, a dumbbell row, a leg press. Something you can take to failure without a spotter and without technique collapsing.
2. Take one set to true failure at a weight you would normally rate around 2 RIR. Count the reps. Notice how reps 1 to n−3 felt, how n−2 and n−1 felt, and what the final, failed rep felt like.
3. Write it down. "I rated this as 2 RIR and got 3 more" is the whole point. Do this once every few weeks per movement pattern, not every session.
4. Rate working sets against that memory. When you stop a set, ask "how many more like that last one could I have done with the same form." Answer in whole or half reps.

The evidence says this is the normal pattern. In 25 trained men squatting to failure at 70 percent of 1RM, the error in predicted reps in reserve was about 2 reps when they called 1 RIR, about 3.7 reps at 3 RIR, and about 5 reps at 5 RIR, and more total reps in the set made predictions worse [1]. In 81 adults across a wide range of experience, error was about one rep when 0 to 5 reps remained and more than two when 7 to 10 remained, and experience did not change accuracy [2]. A meta-analysis of 12 studies found an average underprediction of about one rep, with accuracy slightly better closer to failure and in sets of 12 reps or fewer [3]. The fix is exposure, not trying harder.

Cues that predict failure

Bar speed on the concentric. The most reliable cue for barbell and machine work. Average bar velocity and RIR-based RPE were strongly and inversely related in the squat in both experienced and novice lifters [4]. A clear slow-down relative to the first rep usually means two to three reps left; a grind with a sticking point usually means one or none.

Form drift. Elbows flaring, hips rising early, the bar path bending forward. Form changes are a sign of proximity to failure; a set you stop because form is going is often at 1 to 2 RIR.

Rep-to-rep change. If the last rep felt like the one before it, you likely had more than two left. If it felt clearly different, you were close.

Breathing and urge to stop. Useful but personal. Some people want to stop at 4 RIR; some do not notice until 1. Calibration tells you which you are.

Where ratings go wrong

• Rating by plan instead of by set. The prescription says 2 RIR, so the set gets logged as 2 RIR. Rate what happened.
• Rating by fatigue instead of reps. Feeling tired after a set is not the same as being out of reps. The question is how many more reps of that quality you could have done.
• Changing the standard across sessions. A hard day rated generously and an easy day rated harshly produce noise. Keep the standard even and let the weight tell the story.
• Ignoring half values. "Maybe one, maybe two" is 1.5. Method accepts half values, and the suggested-load calculation uses them.

What Method does with the number

• Estimated 1RM uses reps plus reps in reserve as effective reps. Logging 5 reps at 2 RIR is read as a set of 7 to failure.
• Records compare estimated 1RM, so an overrated RIR can create a false record and an underrated one can hide a real one.
• Suggested loads read the RIR on your last eligible set and the RIR target on the next set. If you rate sets consistently, the suggestion is consistent.
• Display scale is your choice. Log in RIR or RPE; Method stores one value and shows either. See RIR vs RPE.

When not to bother

Warm-ups do not need a rating. Technique work at light loads does not need one. Very high rep sets, above 12 to 15, are hard to rate; both the squat study and the meta-analysis found accuracy falls as the total reps in a set rise [1][3], and the estimate formula is rough there anyway, so a coarse number is fine. Save the precision for the working sets that drive the numbers.