Adaptive Training: What the Evidence Supports

What “adaptive training” means

Adaptive training is a broad label. In strength research, the more established term is autoregulation: adjusting a training variable using information available close to the session. That information might include completed repetitions, perceived effort, repetitions in reserve, or movement velocity.

This is different from changing exercises at random. A useful adjustment still serves the longer-term goal. It changes the dose or structure because the current evidence suggests that the original prescription is no longer the best fit.

For Dorsi, the practical translation is wider than load selection. The plan can use a person’s goal, available time, current condition, equipment, and training feedback to arrange today and what comes next. This is a product design choice informed by training principles—not proof that the product produces a specific physiological outcome.

What research says about autoregulation

A 2021 systematic review and meta-analysis compared autoregulated and fixed-loading approaches for maximal strength. It included eight studies and 166 trained participants or athletes. The authors found a possible advantage for autoregulation, while also noting the limited number and size of the available studies. That is encouraging evidence, not a universal rule. Read the review on PubMed.

A newer systematic review and network meta-analysis examined several autoregulation methods, including autoregulatory progressive resistance exercise, RPE-based training, and velocity-based training. Its purpose was to compare their effects on maximal strength, but the methods and populations still vary across the included studies. Read the review on PubMed.

The responsible conclusion is:

• autoregulation is a credible programming approach;
• several methods can be used to implement it;
• the evidence does not justify claims that adaptive training always prevents injury, always improves adherence, or always beats fixed programming for every person and goal.

RPE and repetitions in reserve

Rate of perceived exertion (RPE) and repetitions in reserve (RIR) turn the lifter’s experience of a set into usable feedback. A set that was expected to leave three repetitions in reserve but felt maximal tells a coach—or a planning system—that the prescription and the day did not match.

That feedback is useful because capacity is not perfectly stable. But it remains subjective. New lifters may need time to calibrate it, and motivation, exercise familiarity, and technique can all affect the rating.

A practical system should therefore treat effort feedback as context rather than unquestionable truth. It can combine what was planned, what was completed, and how the session felt, then make the next decision explicit.

Velocity feedback is a separate method

Velocity-based training uses measured movement speed to regulate load or volume. It is often grouped with RPE and autoregulatory progressive resistance exercise because all three can change a prescription using session-level information.

A 2022 systematic review compared velocity-based and traditional resistance training across strength, power, and sprint outcomes. The review found that velocity feedback can be useful, while the quality and consistency of evidence differed by outcome. Read the review on PubMed.

Dorsi should not be described as measuring bar speed unless the product actually performs and validates that measurement. The existence of velocity-based training research does not make every wearable or adaptive app a velocity tracker.

Sleep can matter without becoming a diagnosis

Sleep is relevant to training because sleep loss can affect physical performance and perceived effort. A systematic review focused on inadequate sleep and resistance exercise found that the literature was still limited and varied by the type of sleep loss and performance outcome. Read the review on PubMed.

A broader 2022 meta-analysis also found an overall negative effect of acute sleep loss on physical performance, with the effect depending partly on how and when sleep was lost and when performance was tested. Read the review on PubMed.

These findings support paying attention to sleep. They do not support a rigid rule such as “low HRV means rest” or “four hours of sleep means this exact workout.” Wearable measurements have noise, individual baselines differ, and training decisions still need context.

What should change when the day changes?

An adaptive decision can affect more than intensity. Depending on the person’s goal and constraints, a plan might reconsider:

• session length;
• exercise selection;
• sets and repetitions;
• loading or effort target;
• exercise order;
• whether work moves to a later session;
• how the remaining week stays balanced.

The last two points are easy to miss. Shortening today can be sensible, but an ongoing plan also needs to decide what that means for tomorrow and the sessions after it. Otherwise the product is generating isolated workouts, not maintaining a plan.

Evidence does not transfer automatically to a product

Research on an underlying method and evidence for a specific app are different things.

Autoregulation research can support the idea of adjusting prescriptions. Sleep research can support considering recent sleep. Neither proves that a particular algorithm makes the right adjustment, improves strength by a certain percentage, reduces injuries, or increases adherence.

Claims about Dorsi should therefore stay within what can be demonstrated:

• the inputs the product accepts;
• the plan it produces;
• how a change in an input changes today;
• how the upcoming plan is rearranged;
• what the user can review or correct.

Outcome claims require product-specific evidence with a clear method, comparison, and sample. Until that exists, we should not publish invented trials, user percentages, biomarker counts, or clinical-sounding guarantees.

How Dorsi applies the principles

Dorsi is designed to give each user a personalized workout plan that stays connected to their goal and current conditions. It considers available time, condition, equipment, and training feedback, then arranges today and the training ahead. When those inputs change, the plan can change with them.

This is closer to a continuous planning loop than a one-time generator:

1. establish the goal and current constraints;
2. arrange today and the sessions ahead;
3. record what actually happened;
4. update the relevant conditions or feedback;
5. reconsider the next decision without losing the larger goal.

The quality of that loop should be judged by its visible decisions, not by the word “AI.”

Questions to ask about any adaptive training app

Does it explain why a workout changed?

It should be possible to connect an adjustment to a real input, such as available time, equipment, condition, or previous training feedback.

Does it update the future plan too?

If only today changes, the app may still leave the rest of the schedule inconsistent. Ask what happens next.

Can the user correct the context?

The user should be able to say that the available equipment, time, or condition is different from what the app assumed.

Does it make medical claims?

A training planner is not a diagnostic tool. Pain, injury, illness, and medical concerns require appropriate professional judgment.

Are outcome claims backed by product-specific evidence?

Evidence for autoregulation is not evidence for a particular app. Look for a published method behind any claim about gains, injury reduction, adherence, or recovery.

Bottom line

Adaptive training is a defensible approach when it means using current information to adjust a coherent program. The research supports autoregulation as a useful option, while leaving important questions about populations, methods, and outcomes.

For a product, the honest standard is simpler: show the inputs, show the decision, show what changed today, and show how the plan ahead remains connected to the goal.