Improve Your Strength Using Data
Most lifters have more training data than they have decisions. The problem isn't a lack of metrics — it's that raw logs don't answer questions, and chasing every number produces noise rather than progress. Five rules fix that.
The short answer
Log completely, derive one number per lift that normalises for rep count, read it in blocks rather than daily, treat volume as a separate axis from strength, and stop when the numbers stop changing your decisions. That's the framework. Everything below is elaboration on those five points, plus the metrics worth tracking and the mistakes that make data worse than useless.
Rule 1 — Log completely, or don't bother
The first rule is about the input, and it is unglamorous enough that it gets skipped. A session log where you recorded only your top set is not a partial record of your training — it is a misleading one, because it will make your best-set estimate drop over time as your logging lapses, which you will then read as a plateau.
Every set, every session, including the bad ones and especially the ones you failed. The failure sessions carry more information about your programming than the successes do, and a log that omits them will systematically flatter your progress.
This is the single highest-leverage thing you can fix, and it costs nothing but the discipline of logging on the days you don't want to.
Rule 2 — Derive one number, not twelve
Logged weight cannot answer whether you are getting stronger, because it cannot distinguish 90 kg for 4 from 85 kg for 8. Both look like "the weight went up" or "the weight went down" depending on the week, and neither tells you about capacity.
Derive one normalised figure per lift per session — an estimated one-rep max via weight × (1 + reps ÷ 30) — and you have a line instead of a list. This guide covers the method, and the 1RM calculator will do the arithmetic.
Why only one? Because each additional metric multiplies the number of decisions you could make, and most of them are decisions you shouldn't be making. One derived number plus session volume is enough to run a training cycle. The four records apps typically show overlap heavily, and only estimated 1RM survives comparison honestly.
Read in blocks, not days
Session-to-session variation on a main lift is a few percent in either direction for any trained lifter. Daily review therefore shows you noise roughly half the time, and the correct response to noise is to do nothing — which is a decision most people find unsatisfying and therefore stop checking.
Compare the mean of your last three sessions against the three before. Three is the smallest unit that reliably separates a real change from variation. Once a week is enough; once a session is actively counterproductive.
The corollary matters more than the rule: do not change a working programme because of a single data point. Programmes get changed at mesocycle boundaries, using blocks of data. Changing weekly on the strength of one bad session is the most common way training data makes people worse at training.
Rule 4 — Treat volume as a separate axis from strength
Most stalled progress is diagnosed as a strength problem when it is a volume or fatigue problem. These move independently, and the combinations are informative:
| Estimated 1RM | Volume | What it usually means |
|---|---|---|
| Rising | Rising | Progressing well. This is the target state. |
| Rising | Falling | Often a deload, or a shift in rep range. Normal and frequently desirable. |
| Flat | Rising | Fatigue accumulating without added capacity. The pattern that precedes a stall. |
| Falling | Falling | Recovery problem or a genuine overreaching block. Check sleep before blaming the programme. |
| Flat | Flat | A real plateau, or a logging problem. Rule out incomplete logs first. |
The row that matters most is the third. Estimated max holding flat while volume climbs is the signature of fatigue outrunning adaptation, and it is visible weeks before you would notice it by feel.
Stop collecting when it stops changing decisions
There is a point at which additional metrics produce no better decisions, only more opportunities to over-interpret. For most lifters that point is four numbers:
- Estimated 1RM per main lift — the primary signal
- Session volume per main lift — the fatigue axis
- PR drought per lift — days since that lift last improved, which makes stagnation visible
- Sleep, weekly average — because it explains variation that otherwise looks like a plateau
Everything else is context you consult when a question arises, not a dashboard you maintain. Sleep is on that list because it is the one non-training variable that reliably affects output and that you control — consistency there beats duration. Recovery scores, on the other hand, are a weaker signal than most marketing implies: the relationship to output is real but small. Our read on that evidence.
What you actually need to log in
Worth being concrete, because this is a common sticking point. You need:
- Date
- Exercise name, consistently — the single most common data failure. "Barbell Bench Press" and "Bench Press (Barbell)" are two exercises to a spreadsheet, and you will lose the history.
- Weight and unit — pick one unit and never mix
- Reps and set type — you cannot compute an estimated max from a rep count you didn't record
- RPE — optional, and one of the more useful optional fields. It distinguishes "moved the same weight and it felt awful" from "moved it and felt fine."
That is it. A dedicated strength logger is convenient but the data shape is trivial, and a spreadsheet or note-taking app will serve anyone willing to keep naming things consistently.
Mistakes that make data worse than none
- Comparing across exercises. Your bench press in January is not comparable to your bench press in December if the rep scheme changed. Estimated 1RM handles some of this; nothing handles a changed exercise.
- Comparing across lifters. Your estimated max is yours. It is not a benchmark and should never be used as one.
- Chasing recovery scores. Small, noisy relationship. Fine as context, poor as a decision rule.
- Optimising sleep stage percentages. Estimated from movement, not brain activity. Two metrics matter; the rest is context.
- Reading daily and feeling something. Daily variation is noise. Weekly review is signal.
- Logging inconsistently and concluding you have plateaued. Incomplete logs look like declining performance. Check your set counts before you change anything.
One thing we deliberately don't give you. You will find articles claiming specific percentages — that match-intensity lifters out-progress others by X%, that pushing through bad days predicts regression within Y days. We have removed those figures from this site. The effect sizes in the research are small, they come from small samples, and they describe populations rather than you. Presenting them as actionable targets would be the exact error this framework exists to avoid.
Frequently asked
Which metrics actually matter for tracking strength progress?
Estimated one-rep max per lift, read in blocks of three sessions rather than daily, is the primary metric. Session volume alongside it catches the fatigue pattern, and a per-lift PR drought counter catches stagnation. Four numbers is enough, and logging consistency matters more than how many you track.
Do I need a wearable to use my training data?
No. The useful numbers come from your session log — weight, reps, dates — which a spreadsheet handles. A wearable adds recovery and readiness context, which is genuinely valuable, but it is not what turns a log into progress data. Derived records are arithmetic on what you already logged.
How often should I review my training data?
Weekly for the trend, and only at a mesocycle boundary for programme changes. Reviewing daily produces anxiety rather than information, because session-to-session variation is larger than the weekly signal. Checking more often leads you to change a working programme based on noise.
What should I do with my data when I plateau?
Check whether it is real first — a plateau is usually four to six sessions without meaningful change in estimated one-rep max, and often it is incomplete logging or accumulated fatigue rather than a ceiling. If it is real, deload before changing anything.
Is more data always better for training?
No. Additional metrics have diminishing returns once you can answer your real questions — is this lift progressing, is the programme balanced, am I accumulating fatigue. Beyond that, more data mostly adds opportunities to over-interpret noise.
Related
- Estimated 1RM calculator — free, six formulas
- Build an estimated 1RM trend from WHOOP logs
- How to spot a lifting plateau in your data
- Peak Load vs estimated 1RM vs Top Set Output
- Sleep consistency and your training data
Sources
- ACSM position stand on resistance training progression models.
- Schoenfeld, A.J. et al. on dose-response relationships and resistance training volume.
- Mann, T.L., Lamberts, R.P. & Lambert, M.I. (2014). High responders and low responders: factors associated with individual variation in response to standardized training. Sports Medicine.
- Fullagar, C.J.K. et al. (2015). Sleep and athletic performance. Sports Medicine.
- Epley, B. (1990). Weightlifting predictors of one-rep max performance. Journal of Strength and Conditioning Research.