Progressive Overload is a Closed-Loop System: The Engineering of Autoregulation

Most lifters run their training like a toaster. You set the timer, press the lever, and hope the bread comes out right. The toaster never checks whether the bread is burning. It just runs the program.

That is exactly how a typical "add 2.5 kg every week" plan works. The plan was written before the first session, and it never looks at what is actually happening inside your body. When you sleep badly, eat in a deficit, or carry a sore elbow into the gym, the program does not care. It adds the weight anyway.

In control engineering, this is called an open-loop system. It works well when the environment is stable and predictable. The human body is neither. In this guide, I'll show you how to turn progressive overload into a closed-loop feedback system using Reps in Reserve (RIR) and RPE, how to model fatigue as system wear and tear, and exactly which data you need to log to keep the loop stable.

Quick answer: autoregulation means adjusting load or volume based on how you perform today, not on a number fixed weeks ago. RIR and RPE act as the sensor, your target effort is the setpoint, and your load adjustment rule is the controller. Research shows this approach is at least as effective as fixed percentage loading, and in some studies better for strength, while making it easier to manage fatigue.

Part 1: Control Systems 101 for Lifters

Open-loop vs. closed-loop

Every control system has the same basic parts. Here they are, translated into gym language:

Control system component Engineering meaning Training equivalent
SetpointThe target valueTarget effort, e.g. "3 sets of 8 at 2 RIR"
PlantThe system being controlledYour neuromuscular system
SensorMeasures the real outputYour RIR / RPE rating, rep count, bar speed
Error signalSetpoint minus measured outputTarget RIR minus actual RIR
ControllerRule that converts error into actionYour load adjustment rule for the next set or session
DisturbancesOutside inputs that push the system off targetPoor sleep, work stress, calorie deficit, illness, joint pain

An open-loop program has a setpoint and a plant, but no sensor and no feedback. The load for week 6 is decided in week 1. If a disturbance hits, the error grows silently until something breaks: a missed lift, a stalled month, or an irritated joint.

A closed-loop program measures the output every set, compares it to the target, and corrects. If the bar moved easier than planned, load goes up. If it moved harder, load comes down. The system absorbs disturbances instead of being knocked over by them.

Why open-loop progression eventually fails

Linear progression works for beginners because their rate of adaptation is so fast that it outruns most disturbances. But as you get more advanced, two things happen:

  • The signal gets smaller. Your weekly strength gain shrinks, so a fixed weekly increase quickly becomes larger than your real adaptation rate.
  • The noise stays the same. A bad night of sleep affects an advanced lifter just as much as a beginner. Day-to-day readiness variation becomes bigger than weekly progress.

In engineering terms, the signal-to-noise ratio drops. A system without feedback cannot tell signal from noise, so it keeps pushing until it fails.

Part 2: RIR and RPE — The Sensor in the Loop

The RIR-based RPE scale

The modern resistance-training RPE scale was developed specifically to measure proximity to failure. Zourdos and colleagues (2016) introduced an RPE scale anchored to repetitions in reserve: RPE 10 means no reps left, RPE 9 means one rep left, RPE 8 means two reps left, and so on. Helms and colleagues (2016) then described how to apply this scale in program design, using it to set daily loads instead of fixed percentages.

This turned a vague feeling ("that was hard") into a numeric sensor reading that can feed a control rule.

How accurate is the sensor?

Every engineer asks this question before trusting a sensor. Here the answer is: good, but not perfect. A scoping review and meta-analysis by Halperin and colleagues found that, on average, lifters guessed they had roughly one rep fewer left than they really did. In practice, your "2 RIR" set is often closer to 3 RIR.

The same analysis found that guesses got more accurate as the set got closer to failure, with heavier loads (fewer reps), and in later sets. Experience alone did not solve the problem: trained and untrained lifters made similar errors.

From a Measurement System Analysis (MSA) point of view, this means:

  • Bias: about one rep, in the conservative direction. You can correct for it.
  • Precision is better near failure: RIR 0–3 readings are more trustworthy than RIR 5–6 readings.
  • Calibration is possible: taking an occasional set to true failure on a safe exercise (for example, a machine or cable movement) recalibrates your sensor. This is the same idea as zeroing a scale before using it, which we covered in our article on why your InBody scan and scale are lying to you.

Part 3: What the Research Says About Autoregulation

Does closing the loop actually produce better results? Here is the evidence:

  • Graham & Cleather (2021, JSCR) compared RIR-based autoregulation with fixed percentage loading in 31 trained men over 12 weeks of squat training, with volume matched. The autoregulated group gained more squat strength, and it also ended up training at higher average intensities, because the feedback loop let the load rise as they got stronger.
  • Helms et al. (2018, Frontiers in Physiology) compared RPE-based loading with percentage-based loading in periodized programs matched for sets and reps. Both approaches worked well, with results slightly favoring the RPE group for strength.
  • Mann et al. (2010, JSCR) showed that an autoregulatory progressive resistance exercise (APRE) protocol, where load is adjusted based on reps achieved, produced greater strength gains in college athletes than linear periodization.
  • Hickmott et al. (2022, Sports Medicine – Open), a meta-analysis on load and volume autoregulation, gives the balanced view: overall, autoregulated and fixed (percentage-based) loading led to similar strength gains.

Engineering interpretation: closed-loop control is not magic. Its main benefit is robustness. When the environment is stable, open-loop and closed-loop give similar outputs. When disturbances appear (and in real life they always do), the closed-loop system keeps you on target while the open-loop system drifts.

How close to failure should the setpoint be?

Your setpoint (target RIR) is a design choice. A meta-analysis by Refalo, Helms and colleagues found no clear hypertrophy advantage for training all the way to momentary failure compared with stopping a little short, and suggested that the relationship between proximity to failure and muscle growth may not be linear.

Practically, this means most working sets can sit in the 0–3 RIR range for hypertrophy, with true failure used selectively. You do not need to run the machine at 100% load every cycle to get the output you want.

Part 4: Fatigue Accumulation — System Wear and Tear

The fitness–fatigue model: a two-state system

In 1975, Banister and colleagues proposed a systems model of training. It treats each workout as an input that creates two outputs at the same time:

  • Fitness: a positive effect that builds slowly and decays slowly.
  • Fatigue: a negative effect that builds quickly and also decays quickly.

Performance today ≈ Baseline + Fitness − Fatigue

This is why you can feel weak in the middle of a hard training block and then hit personal records after a deload: fatigue disappears faster than fitness. The fatigue is real, but it is masking the fitness you built.

Tolerance limits and the stress–strain analogy

Mechanical engineers know that materials can handle repeated loading below a certain limit almost forever, but loading above that limit causes cumulative damage, and eventually failure. Muscles, tendons and joints behave in a similar way:

  • Muscle recovers relatively fast (days).
  • Tendons and connective tissue adapt and recover more slowly (weeks to months).
  • The nervous system and overall stress state are affected by sleep, nutrition and life stress, not only by training.

Research on the time course of recovery supports this. Morán-Navarro et al. (2017, European Journal of Applied Physiology) found that sets taken to failure required noticeably longer recovery of neuromuscular performance than the same work stopped short of failure. Training to failure is not "bad," but it raises the maintenance cost of each session.

Load spikes, not load itself, cause breakdowns

Gabbett's (2016, British Journal of Sports Medicine) work on the "training–injury prevention paradox" showed that athletes who build high chronic workloads gradually are well protected, while sudden spikes in workload relative to what they are used to are associated with higher injury risk. In engineering terms: it is the rate of change of load that kills the system, not the load level itself. Open-loop programs are especially prone to creating these spikes when they ignore missed sessions or accumulated fatigue.

Monitoring fatigue: more than one sensor

Halson (2014, Sports Medicine) reviewed methods for monitoring training load and fatigue and recommended combining external load (what you lifted) with internal load (how hard it felt) and simple wellness measures such as sleep and soreness. This is sensor fusion: no single sensor is perfect, but several imperfect sensors together give a reliable picture.

Of all these wellness inputs, sleep has the biggest effect on both your recovery capacity and the accuracy of your RPE readings. We cover the details, plus a practical protocol, in Sleep as System Maintenance.

Your first sensor readings come from the warm-up. Before the controller has any working-set data, your ramp-up sets already tell you how the system is running today: a 70% set that moves slowly is an early disturbance signal. Learn how to structure those startup sets without wasting fatigue in Warm-Up as Machine Ramp-Up: The Engineering of Startup Curves and Joint Preparation.

Part 5: Designing the Controller — Your Adjustment Rules

Within a session: set-to-set correction

A simple proportional controller works well. After your first working set, compare actual RIR to target RIR and adjust:

Error (actual RIR − target RIR) Meaning Action on next set
+2 or moreLoad far too lightIncrease load ~5%
+1Slightly too lightIncrease load ~2–3% or add 1–2 reps
0On targetKeep the load
−1Slightly too heavyReduce load ~2–3%
−2 or more / form breakdownLarge negative error or disturbanceReduce load ~5% or more, check sleep, pain and nutrition log

These percentages are practical starting values, not laws. Tune them like you would tune the gain of a controller: too aggressive and your loads oscillate; too passive and you under-train.

Across sessions: Statistical Process Control (SPC)

To see whether you are actually progressing, track an estimated 1RM (e1RM) for your main lifts every session. A common estimate is the Epley formula, adjusted for RIR:

e1RM ≈ Load × (1 + (Reps + RIR) / 30)

Then treat your e1RM like a process variable on a control chart:

  • Calculate the average of your last 4–6 sessions and the normal range of variation.
  • A single low session inside the normal range is common-cause variation (noise). Ignore it.
  • Two or three sessions below the lower range, especially with poor sleep or rising joint pain, is a special-cause signal. That is your trigger for a deload or a volume reduction.

This is how factories avoid reacting to every small fluctuation while still catching real problems early.

The controller works inside a plan. Autoregulation adjusts load day to day, but it should operate within a block that has a clear goal. See how session-level control fits into the bigger schedule in Periodization as Production Planning: Aggregate Planning, MPS, and Mesocycles for Lifters.

Part 6: The Data Logging Framework

A feedback loop is only as good as its data. Below is the minimum dataset a lifter needs to run a stable closed-loop system.

Variable Unit / scale Frequency Role in the control loop Action threshold (example)
Load (weight)kgEvery setController outputAdjust per RIR error rule
RepsCountEvery setOutput measurementHitting top of rep range at target RIR → increase load
RPE / RIRRPE 1–10 or RIR 0–5Every working setPrimary sensorError of ±1 or more → adjust next set
Estimated 1RMkg (calculated)Each session, main liftsProcess variable for SPC2–3 sessions below normal range → review fatigue
Sleep quality1–5 rating + hoursDailyDisturbance sensorUnder ~6 h or rating ≤2 for 3+ nights → reduce volume
Joint pain0–10 pain scale, by jointEach sessionSafety sensor (tolerance limit)≥3/10 or rising over sessions → swap exercise or reduce load
Muscle soreness0–10 by regionBefore each sessionRecovery sensorHigh soreness in target muscle → move to lighter day
Bodyweightkg, weekly averageDaily weigh-inEnergy balance sensorUnplanned loss + falling e1RM → check calories
Calorie intakekcal vs. TDEEDailyEnergy inputDeficit >~500 kcal → expect lower performance, adjust setpoint
Machine settingsSeat / pin / pulley positionWhen setup changesMeasurement reproducibilityAny change → do not compare loads directly

Two of these rows link directly to earlier articles. Machine and cable settings matter because, as we showed in our biomechanics analysis of the leg press, hack squat and Smith machine, a small setup change can change joint torque and make loads non-comparable. Calorie intake matters because energy is a system input: if you don't know your maintenance level, start with our TDEE Calculator.

Sensor calibration tip: on machines, an undocumented change in seat height or back pad angle adds noise to every number you log, so your controller ends up correcting for setup error instead of real fatigue. We break this down in Machine Seat Settings as Jigs & Fixtures: The Engineering of Setup Error.

Part 7: Putting the Closed Loop Into Your Split

Assigning setpoints by day

Autoregulation fits any split, but the structure of your split decides how often each loop runs. In a Push/Pull/Legs split, each muscle group usually gets feedback once or twice per week. In an Upper/Lower split, the feedback frequency per muscle can be higher. We compared both systems in our article on Push/Pull/Legs vs. Upper/Lower workload distribution.

A practical setpoint design:

  • Heavy compound lifts (squat, bench, deadlift variations): 1–3 RIR. Higher fatigue cost, so avoid failure.
  • Machine and cable isolation work: 0–2 RIR. Lower systemic fatigue and safer to push, and a good place to calibrate your RIR sensor with an occasional true-failure set.
  • Early in a training block: start 1 RIR more conservative than your target, then tighten the setpoint each week.

Planned maintenance: the deload

Factories do not wait for a machine to break before servicing it. They schedule preventive maintenance. In training, this is the deload: a planned week (often every 4–8 weeks) with reduced volume or effort. Your log decides whether you take it early (special-cause signal) or on schedule.

For the full maintenance-engineering breakdown (TPM, MTBF, a scheduled-vs-indicator decision matrix and a deload specification table), see The Deload Week as Preventive Maintenance.

Key Takeaways

  • "Add weight every week" is an open-loop system. It works until disturbances (sleep, stress, diet) become larger than your adaptation rate.
  • RIR/RPE is your sensor. It has a bias of about one rep on average, and it is more accurate near failure. Calibrate it occasionally.
  • Autoregulation's main benefit is robustness. Research shows it is at least as good as fixed loading, and some studies show greater strength gains.
  • Fatigue builds faster than fitness and decays faster. Manage load spikes, not just load.
  • Log the full dataset: load, reps, RIR, e1RM, sleep, joint pain, bodyweight, calories and machine settings. Without data, there is no feedback loop.

References

  1. Zourdos, M. C., Klemp, A., Dolan, C., Quiles, J. M., Schau, K. A., Jo, E., Helms, E., et al. (2016). Novel resistance training–specific rating of perceived exertion scale measuring repetitions in reserve. Journal of Strength and Conditioning Research, 30(1), 267–275.
  2. Helms, E. R., Cronin, J., Storey, A., & Zourdos, M. C. (2016). Application of the repetitions in reserve-based rating of perceived exertion scale for resistance training. Strength and Conditioning Journal, 38(4), 42–49.
  3. Helms, E. R., Byrnes, R. K., Cooke, D. M., et al. (2018). RPE vs. percentage 1RM loading in periodized programs matched for sets and repetitions. Frontiers in Physiology, 9, 247.
  4. Graham, T., & Cleather, D. J. (2021). Autoregulation by "repetitions in reserve" leads to greater improvements in strength over a 12-week training program than fixed loading. Journal of Strength and Conditioning Research, 35(9), 2451–2456.
  5. Mann, J. B., Thyfault, J. P., Ivey, P. A., & Sayers, S. P. (2010). The effect of autoregulatory progressive resistance exercise vs. linear periodization on strength improvement in college athletes. Journal of Strength and Conditioning Research, 24(7), 1718–1723.
  6. Hickmott, L. M., Chilibeck, P. D., Shaw, K. A., & Butcher, S. J. (2022). The effect of load and volume autoregulation on muscular strength and hypertrophy: a systematic review and meta-analysis. Sports Medicine – Open, 8, 9.
  7. Halperin, I., Malleron, T., Har-Nir, I., Androulakis-Korakakis, P., Wolf, M., Fisher, J., & Steele, J. (2022). Accuracy in predicting repetitions to task failure in resistance exercise: a scoping review and exploratory meta-analysis. Sports Medicine, 52(2), 377–390.
  8. Refalo, M. C., Helms, E. R., Trexler, E. T., Hamilton, D. L., & Fyfe, J. J. (2023). Influence of resistance training proximity-to-failure on skeletal muscle hypertrophy: a systematic review with meta-analysis. Sports Medicine, 53(3), 649–665.
  9. Morán-Navarro, R., Pérez, C. E., Mora-Rodríguez, R., et al. (2017). Time course of recovery following resistance training leading or not to failure. European Journal of Applied Physiology, 117(12), 2387–2399.
  10. Gabbett, T. J. (2016). The training–injury prevention paradox: should athletes be training smarter and harder? British Journal of Sports Medicine, 50(5), 273–280.
  11. Halson, S. L. (2014). Monitoring training load to understand fatigue in athletes. Sports Medicine, 44(Suppl 2), S139–S147.
  12. Banister, E. W., Calvert, T. W., Savage, M. V., & Bach, T. (1975). A systems model of training for athletic performance. Australian Journal of Sports Medicine, 7(3), 57–61.

Disclaimer: This article is for educational purposes only and is not medical advice. Persistent or worsening joint pain should be assessed by a qualified professional.

Comments

Popular posts from this blog

Caffeine as a Turbocharger: Adenosine Antagonism, Half-Life Decay, and System Performance

Warm-Up as Machine Ramp-Up: The Engineering of Startup Curves and Joint Preparation

Periodization as Production Planning: Aggregate Planning, MPS, and Mesocycles for Lifters