How Sparlift calculates muscle recovery time

A muscle trained with about three sets taken to failure needs roughly 48 hours before it performs at its best again. Quads and hamstrings need about 60 hours; small muscles such as calves, forearms and abs about 36. Sparlift turns every set you tick into a fraction of a set to failure, adds those fractions up per muscle, and runs a recovery clock in hours for each of 19 muscles. A second read-out, the Nervous system, shows whole-body training stress in %, which weeks of hard training build up and a deload week clears (how Sparlift counts training stress and rest days).

Everything is inferred from the log. The app never asks how you slept or how hard a session felt. This page gives every formula and the evidence behind it, with a worked example from a real training week.

#Summary

• Effort, not load, drives recovery time. A set's effort is how close to failure it was: taken from the reps in reserve (RIR) you logged, or estimated against your own best set of the last 8 weeks.
• Hard sets. Effort becomes a hard-set value between 0 and 1 — nothing at half effort, 1 at failure. A muscle gets the full value from exercises where it is a main mover and half from exercises where it assists.
• Hours. Three hard sets leave a muscle its reference time, T_ref: 36, 48 or 60 hours. More sets add less and less, up to 1.5 × T_ref from one session.
• A clock that adds up. Sessions on a muscle that hasn't recovered stack, up to 2.5 × T_ref. The clock counts down one hour per hour and wears the oldest session off first.
• Parts. Shoulders and back are shown in parts (front, side and rear delts; lats, upper back, traps, lower back), each with its own clock and its group's recovery time.
• Whole-body stress. Training stress in %, built over weeks: your last 7 days of working sets set where it's heading, and it climbs there slowly and falls fast (the model). Session effort × minutes, as a 48-hour average compared with a 21-day one, stays as a trend line: your usual always reads 50.

#Recovery time by muscle group

T_ref is the recovery time after about three sets taken to failure. One session can leave at most 1.5 × T_ref; sessions stacked on an unrecovered muscle, at most 2.5 × T_ref.

Muscle	T_ref (≈ 3 sets to failure)	Most from one session	Most on the clock
Quads, hamstrings	60 h	90 h	150 h
Chest, shoulders and back (each in its parts), glutes, biceps, triceps	48 h	72 h	120 h
Calves, forearms, abs, obliques, hips & adductors, neck	36 h	54 h	90 h

Where the numbers come from. Three sets of ten taken to failure in the bench press and the squat kept mechanical performance down until about 48 hours, while the same repetitions stopped short of failure recovered much faster (Morán-Navarro et al., 2017). Eight sets of ten left strength deficits for up to 72 hours; eight heavy sets of three, for 24 (Bartolomei et al., 2017) — hence the session cap of 1.5 × 48 h. Single-joint arm exercises caused larger strength losses and longer soreness than multi-joint ones (Soares et al., 2015), so biceps and triceps sit at 48 h rather than 36. The leg and small-muscle values are extrapolations from those findings, not measured directly.

#Shoulders and back, in parts

Two of the 14 groups hide several muscles, and they are the two where I had blind spots in my own log: shoulders (front, side and rear delts) and back (lats, upper back, traps, lower back). The Regeneration screen shows them in parts, each with its own row and colour, in place of one "Shoulders" and one "Back & traps": tapping one on the body opens a zoom on that side. Every exercise is tagged with the parts it trains, main mover or helper, just as before: an overhead press is a main mover for the front and side delts, a row for the lats and the upper back, and a helper for the rear delts.

• I dropped the pooled Shoulders and Back rows. A clock that adds all shoulder work together reads "Overreached" after an overhead press and a set of face pulls even though no single part is overworked, so I show the parts instead. The engine still computes the group clocks, and the 12 other muscles read as before. The worked example further down keeps the group numbers it was first published with, then gives the same evening in parts.
• A part runs the same clock as its group, fed only by the exercises tagged with it, with the group's reference time. I gave the parts no recovery times of their own: I found no evidence that the heads of one muscle recover at different speeds.
• Which part an exercise trains is an estimate (evidence level: Prior). It comes from a handful of small electromyography studies and from anatomy: in one study of eight trained men the front of the deltoid reached about 70% of its maximum in a machine shoulder press and 55% in a bench press, the side about 55% in lateral raises, and the rear about 90% in a reverse pec deck, well above the other exercises (study); the lats were most active in pull-ups and chin-ups, the middle trapezius in rows (thesis); the erector spinae in every deadlift variant (review). Muscle activity says which part works most, not which part grows most, and the samples are small. Treat a part's state as an estimate of where the work landed.
• Incline presses count the front delts as a main mover (since 13 October). In the first version only a flat or incline bench helped the front delts, and after my push days they read lighter than they felt. The front of the deltoid is more active the steeper the bench: in one study of 30 trained adults it was highest at 60°, and above 45° it was significantly higher than on a flat bench (study). At a gentle 30° incline that evidence is thin, so this is my judgement as much as a finding (evidence level: Prior).
• The lower back gets a floor on deadlifts (since 13 October): see step 3. It is a calibration to my own log, not a finding (evidence level: Experimental).

#Step 1 — how hard was each set

Each working set gets an effort q between 0 and 1, where 1 means the set was taken to failure. Warm-up sets and sets you didn't tick are ignored.

RIR logged (or RPE, as RIR = 10 − RPE):  q = reps / (reps + RIR)
No RIR, a best set in the last 8 weeks:  q = reps / possible
                                         possible = 30 × (e1RM_best / weight − 1) + RIR_best
No RIR and no history:                   q = 1   (assumed taken to failure)
Heavy sets (weight ≥ 85% of e1RM_best):  q ≥ weight / e1RM_best − 0.1

possible inverts the Epley formula for the estimated one-rep max: the reps your best set says you could do at this weight. RIR_best is the RIR you logged on that best set, or 1 if you didn't — most lifters stop a rep or two short of true failure, so an unmarked best set is assumed to have had one rep left. The heavy-set floor exists because a set at 90% of your best is never far from failure, whatever the arithmetic says.

RIR is the best input. The RIR-based RPE scale is well validated in experienced lifters (Zourdos et al., 2016; Helms et al., 2016). Predictions aren't perfect: across studies, lifters under-estimated their remaining reps by about one on average, and got a little better close to failure (Halperin et al., 2022). Logging RIR 0 on the set you took to failure also sharpens every other set of that exercise, because that set becomes the yardstick.

#Step 2 — the hard-set value

h = clamp((q − 0.5) / 0.5, 0, 1) ^ 1.5

Effort q	0.5	0.6	0.7	0.8	0.9	1.0
Hard-set value h	0	0.089	0.253	0.465	0.716	1

The shape follows the evidence that proximity to failure, more than the load itself, sets how long recovery takes. Squats at 80% of 1RM stopped far from failure (20% velocity loss) recovered within about 24 hours, while sets carried close to failure recovered slowest (Pareja-Blanco et al., 2019).

#Step 3 — hard sets per muscle

H(muscle) = Σ h × role        role = 1 for a main mover, 0.5 for an assisting muscle

• Joints and tendons. A heavy multi-joint lift counts × 1.25 on its main movers when one of its sets reaches 80% or more of your best. "Heavy multi-joint" means a systemic load of 0.5 or more in the exercise catalogue: deadlift (1.0), back squat (0.9), bench press (0.7), overhead press and barbell row (0.6), pull-up (0.5). Tendon collagen synthesis peaks about 24 hours after exercise and is still raised at 72 (Miller et al., 2005).
• The lower back on a deadlift. On an exercise that has the lower back as a main mover, every set at half of your best e1RM or more counts at least 0.6 hard sets for the lower back, whatever its effort, and the ×1.25 above then applies. Without it a deadlift day of three ramp-up sets, one heavy triple stopped a rep short and a back-off set — my own — counted 0.4 of a hard set for the muscle that held the bar through all of it; with it, the same day leaves the lower back about 48 hours. I found no study that measured how long the erector spinae need after deadlifts. In one study of well-trained men, four sets to failure at 80% of 1RM in the deadlift recovered no slower than the squat or the bench press (Belcher et al., 2019), so the floor is my calibration, not a result: I set it from how my own lower back feels after a deadlift day, and will move it if the log says so. It applies to the lower back only; the hamstrings, glutes and everything else on a deadlift are credited by effort as before.
• One-arm and one-leg exercises. A left/right pair at the same position is one set for the muscle; the harder side counts.

#Step 4 — from hard sets to hours

D = T_ref × v(H)
v(H) = H / 3                  for H ≤ 3
v(H) = min(1.5, √(H / 3))     for H > 3

#Step 5 — the recovery clock

Each muscle has a clock in hours. It counts down one hour per hour, and a new session adds its debt D on top of whatever is left, up to 2.5 × T_ref. A second session on a muscle isn't discounted: the repeated-bout effect that would justify a discount was absent in resistance-trained men (Falvo et al., 2008).

The clock keeps each session's hours as a separate part and wears the oldest part off first — it arrived first. The parts left always add up to the clock, which is what the app shows when you open a muscle: the sessions behind its hours, with the exercises and the sets that counted.

Light touches. A session that leaves a muscle less than 15% of T_ref — for the upper back, a couple of face pulls where it only assists — still adds its hours, but it isn't counted as the last time you trained that muscle on the clock. In the app a muscle's row still counts every workout of the last 48 hours that credited it, however light, and lists it when you open the row ("Fresh · 1 workout", "last workout today"), so a muscle you did work never reads as untouched.

#Step 6 — states and "ready in"

Hours left ÷ T_ref	State	What it means
under 0.15	Fresh	Ready to train
0.15 to under 0.5	Recovering	Almost ready
0.5 to 1.25	Worked	About one normal hard session's worth
over 1.25 to 1.5	Taxed	As much as one session can leave
over 1.5	Overreached	Sessions stacked on an unrecovered muscle

Ready in = hours left − 0.15 × T_ref: the time until the muscle reads Fresh. The app shows it in whole hours; the sessions behind a muscle split that number in proportion to what each still has on the clock (largest remainder, so the parts add up exactly).

#Combat sports: rounds and classes

Rounds and timed blocks — pads, bag work, sparring, clinch, grappling rolls — aren't sets. Every 3 minutes count as a fraction of a hard set on the muscles the activity uses:

H = Σ (minutes / 3) × 0.4 × h(RPE / 10)        one class leaves at most 1 × T_ref

RPE is the one you logged for the rounds, else the activity's default: 3 + 5 × its systemic load, which gives about 7.5 for sparring, 6 for pads and bag, 6.5 for clinch and 4.5 for shadowboxing. Those defaults sit where MMA athletes put the same activities: sparring at RPE 7 or more, striking and grappling drills at 4 or less (Kirk et al., 2021).

Logged effort changes a lot. A kickboxing class in the catalogue's shape — rope, shadowboxing, 20 minutes of technique, five pad rounds, three bag rounds and two clinch rounds — adds about 5 hours to the front and side delts at the default effort, but with pads and bag logged at RPE 8 it adds:

Muscle	Hours added
Front delts, side delts	23.8 each
Obliques	17.8
Triceps	11.9
Calves	8.9
Hips & adductors, neck, upper back, traps, abs	1–2.5 each

Every combat exercise in the catalogue tags the front and side delts alike, and the clinch rounds tag the upper back and the traps alike, so the parts of a muscle read the same hours here.

#Worked example: a real training week

The example is my own log: a push/pull split I lift four to six days a week, logged in Sparlift with each set ticked as I did it. It covers the evening of Tuesday 29 September 2026, after a pull day I logged on Monday night and a push day on Tuesday.

#The push day, set by set

I logged RIR on only two sets. The rest were judged against my best set of each exercise in the previous 8 weeks.

Exercise	Sets → effort q → hard-set value h	Hard sets
Overhead press	50×4 q 0.37 → 0 · 55×6 RIR 0 → 1 · 55×6 q 0.83 → 0.52	1.52
Rear delt fly (cable)	3×12 q 0.24 → 0 · 3×12 → 0	0
Face pull	15×12 q 0.56 → 0.04 · 15×15 q 0.70 → 0.25	0.30
Incline press (Smith)	85×6 q 0.86 → 0.60 · 85×6 → 0.60 · 85×7 RIR 0 → 1	2.21
Single-arm preacher curl	20×5 / 20×5 → 0.12 · 20×6 / 20×5 → 0.35 (left/right pairs)	0.48 × 1.25*
Lateral raise	14×14 q 0.74 → 0.33 · 14×14 → 0.33	0.65

* The preacher curl in my log is an exercise I created myself, with a systemic load of 0.5, so the heavy-lift factor applies to it.

Two things stand out. The second overhead-press set of 55×6, identical to the first, counts 0.52 instead of 1 because it had no RIR: it was judged against a best set with an assumed rep in reserve. And the rear-delt flies count nothing, because my best on that cable is far heavier than 3 kg for 12.

Muscle	Hard sets H	Hours from this session
Shoulders	3.58	52.4
Chest	2.21	35.3
Triceps	1.87	29.9
Biceps	0.60	9.6
Forearms	0.24	2.9 (a light touch)
Back & traps	0.15	2.4 (a light touch)

The pull day the evening before — seated and single-arm cable rows, single-arm preacher curls, hammer curls — put 4.95 hard sets and 61.6 hours on the biceps, 2.96 hard sets and 47.4 hours on the back, and 28 hours on the forearms.

#The clocks on Tuesday evening

Muscle	Hours left	State	Ready in	Where the hours came from
Biceps	49.8	Worked	43 h	Pull day 35 · push day 8
Shoulders	44.6	Worked	37 h	Push day 37
Back & traps	28.3	Worked	21 h	Pull day 19 · push day 2 (light)
Chest	27.5	Worked	20 h	Push day 20
Triceps	23.5	Recovering	16 h	Push day 16
Forearms	9.4	Recovering	4 h	Pull day 3 · push day 1

Across six weeks of the same log (28 sessions, 20 August – 29 September), the back and biceps were Fresh at the start of all 12 pull days, and the chest and triceps at the start of all 14 push days. The shoulders were the exception twice (10 and 21 hours left, both "Recovering"), in a week with three push days in five days.

What an unlogged class changes. I also train kickboxing or MMA twice a week, but those classes weren't in my log. Had the kickboxing class above (pads and bag at RPE 8) been logged on Monday at 19:00, the triceps would have read 41.9 hours right after Tuesday's push day instead of 30.0, the shoulders 56.3 instead of 51.1, and the whole-body score below would have been 73 instead of 63. The model can only count what the log holds.

#The same evening, with shoulders and back in parts

The app now shows shoulders and back in parts. I replayed the same logged sets, up to 21:20 that Tuesday, with the muscle tags it uses today. The muscles that are not split (biceps, chest, triceps, forearms) and the pooled back come out exactly as in the table above; the parts read:

Part	Hard sets	Hours added	From
Front delts	3.73	53.5	Push day: overhead press, incline press
Side delts	2.18	34.8	Push day: overhead press, lateral raise
Rear delts	0.30 + 1.48	4.8 (a light touch) + 23.7	Push day: face pulls · pull day: rows (helper)
Lats	2.96	47.4	Pull day: rows
Upper back	2.96 + 0.15	47.4 + 2.4 (a light touch)	Pull day: rows · push day: face pulls (helper)
Traps, lower back	0	0

Part	Hours left	State	Ready in	Where the hours came from
Front delts	45.7	Worked	39 h	Push day 39
Upper back	28.3	Worked	21 h	Pull day 19 · push day 2 (light)
Side delts	27.0	Worked	20 h	Push day 20
Lats	25.9	Worked	19 h	Pull day 19
Rear delts	7.0	Fresh	—	Pull day · push day (light)
Traps, lower back	0	Fresh	—

The pooled Shoulders read 44.6 hours left, 37 to Fresh. The front delts, which the presses load most, read 39: the same ballpark, which is what I feel after a push day. But the pooled number hid the spread: the side delts had 20 hours to go, and the rear delts, which only the face pulls and the rows touched, were already Fresh, with 7.0 hours left, just under the 7.2-hour line (0.15 × 48 h). Today's tags also count incline presses as a main mover for the front delts and rows as a helper for the rear delts, which the first version of this example did not. The back's 21 hours are the upper back's; the lats had 19, and the traps and lower back hadn't been worked (the pull day had no deadlift). Which part an exercise trains is an estimate (see above); the clocks are only as good as those tags.

#Nervous system: whole-body training load

Since October 2026 the main reading of this view is training stress in %, built over weeks from your working sets: the last 7 days set where it's heading, and it climbs there slowly and falls fast — see how Sparlift counts training stress and rest days. This section covers session load and the comparison with your usual that stays as the trend line "Load vs your last 3 weeks".

#Why "load", not "CNS fatigue"

After heavy squats (10 × 5 at 80% of 1RM), strength stayed reduced for up to 72 hours while voluntary activation — the measure of central fatigue — was barely affected; the lasting deficit was peripheral, in the muscle (Thomas et al., 2018; Carroll et al., 2017). So Sparlift doesn't claim to measure a neural state. "Nervous system" is whole-body training load compared with your own usual, the way sports science monitors athletes.

#Session load

Session load follows Foster's session-RPE method — effort × minutes, in arbitrary units (AU) — which is valid and reliable across many sports, combat sports included (Haddad et al., 2017).

Load = RPE_session × minutes × (0.8 + 0.4 × systemic load)
RPE of a lifting set = 3 + 7 × h          (half effort or less reads 3, a set to failure 10)
RPE of a round       = logged RPE, else 3 + 5 × systemic load
RPE_session          = mean over sets (rounds weighted by minutes)
minutes              = the logged duration (5–240 min), else 2.5 per set + 4/3 × round minutes + 5

The systemic factor runs from 0.8 (curls) to 1.2 (deadlifts): heavy compound work costs more than the same minutes of isolation. The session's effort is shown in words: light under 5, moderate under 7.5, hard above.

#Your usual, and the score

acute   = Σ Load × e^(−age / 48 h)   scaled to AU per day
chronic = Σ Load × e^(−age / 21 d)   scaled to AU per day, at least 250
ratio   = acute / chronic
score   = 100 × (1 − 2^(−ratio))

Both averages are divided by the share of their full weight that the history covers (a bias correction, as in the Adam optimiser), so a short history doesn't read low. This is the fitness–fatigue family of models (Banister; the performance-manager chart) in its exponentially weighted form. The acute:chronic ratio as an injury predictor is heavily criticised (Impellizzeri et al., 2020), and Sparlift never uses it that way: the labels only say how today compares with your usual. A ratio also can't see a high plateau — train hard every day for three weeks and that becomes your usual — which is why it's no longer the main reading.

Ratio	Label	What it means
1.5× and up	Well above your usual	Time to ease off: a light day or two, or rest
1.15–1.5×	Above your usual	Fine for a few days; don't stack more hard ones
0.85–1.15×	Your usual	Train as planned
0.4–0.85×	Below your usual	Room for a hard session
under 0.4×	Rested	Fully recharged

Back to your usual. When the ratio is 1.15 or more, the engine steps both averages forward an hour at a time with nothing more logged, and reports the first hour the ratio drops under 1.15. The acute average falls faster than the chronic one, so rest always brings it back.

In the worked example the score is 63 — "Above your usual", a ratio of 1.42, back under 1.15 after 11 hours of rest. The acute load was 355 AU a day against a usual of 250, which is the model's floor: with only my lifting in the log, my usual never rose above it. The push day added the most (81 minutes, RPE 5.4, 402 AU).

#Common questions

#How long do biceps take to recover?

About 48 hours after three sets of curls taken to failure; less when the sets stop a few reps short. Rows and pulldowns add half their hard sets to the biceps, so a pull day with curls often leaves them 50–60 hours, as in the worked example (61.6 hours from one pull day).

#How long does the chest take to recover?

About 48 hours after three hard sets of pressing. In the worked example, three sets of incline press — the last one to failure — left the chest 35 hours.

#How long do legs take to recover?

Longer: Sparlift uses 60 hours for quads and hamstrings after three hard sets, and up to 90 hours after a high-volume leg day taken close to failure.

#Is 48 hours enough rest between workouts for the same muscle?

After a normal session of about three hard sets, usually yes. After eight or more hard sets close to failure, recovery can take up to about 72 hours; sets stopped well short of failure often recover within a day.

#Does the model know when I'm sore?

No. It estimates when a muscle performs again. Delayed-onset soreness peaks around 48 hours and can outlast "Fresh", especially in the first weeks of a new exercise (review).

#Limits

• Population values, not yours. T_ref, the hard-set curve and the combat factor are defaults from group studies. Recovery varies between people, with age, sleep and training history.
• Readiness, not soreness. The clock estimates when a muscle performs again, not when it stops feeling sore.
• Nothing outside the log. Sleep, stress, nutrition and training you didn't log aren't seen — as the unlogged kickboxing classes in the worked example show.
• Unmarked sets are estimates. Sets without RIR are judged against your best set; with no history at all, a set counts as taken to failure, so the first weeks read cautious.
• Short histories. With less than six weeks of logs, your usual is still settling: read the "vs your last 3 weeks" trend line as provisional.
• Not validated against outcomes. The model combines published findings; it hasn't been tested against measured performance in Sparlift users. It is orientation for planning, not a medical assessment.

The reasoning behind these choices, with more from the same training log: A muscle recovery tracker built from my own log.

#References

• Bartolomei S. et al. (2017). Comparison of the recovery response from high-intensity and high-volume resistance exercise in trained men. European Journal of Applied Physiology. Link
• Carroll T. J. et al. (2017). Recovery of central and peripheral neuromuscular fatigue after exercise. Journal of Applied Physiology. Link
• Falvo M. J. et al. (2008). Repeated bout effect is absent in resistance trained men: an electromyographic analysis. Journal of Electromyography and Kinesiology. PMC
• Haddad M. et al. (2017). Session-RPE method for training load monitoring: validity, ecological usefulness, and influencing factors. Frontiers in Neuroscience. PubMed
• Halperin I. et al. (2022). Accuracy in predicting repetitions to task failure in resistance exercise: a scoping review and exploratory meta-analysis. Sports Medicine 52:377–390. PubMed
• Helms E. R. et al. (2016). Application of the repetitions in reserve-based RPE scale for resistance training. Strength and Conditioning Journal. PubMed
• Impellizzeri F. M. et al. (2020). Acute:chronic workload ratio: conceptual issues and fundamental pitfalls. International Journal of Sports Physiology and Performance. Link
• Kirk C. et al. (2021). Quantification of training load distribution in mixed martial arts athletes: a lack of periodisation and load management. PLOS One 16(5). Link
• Miller B. F. et al. (2005). Coordinated collagen and muscle protein synthesis in human patella tendon and quadriceps muscle after exercise. Journal of Physiology. PubMed
• Morán-Navarro R. et al. (2017). Time course of recovery following resistance training leading or not to failure. European Journal of Applied Physiology. Link
• Pareja-Blanco F. et al. (2019). Time course of recovery following resistance exercise with different loading magnitudes and velocity loss in the set. Sports 7(3):59. doi:10.3390/sports7030059
• Soares S. et al. (2015). Comparison of muscle damage after multi-joint and single-joint exercises. Journal of Strength and Conditioning Research. PubMed
• Thomas K. et al. (2018). Neuromuscular fatigue and recovery after heavy resistance, jump, and sprint training. Medicine & Science in Sports & Exercise. PubMed
• Zourdos M. C. et al. (2016). Novel resistance training–specific rating of perceived exertion scale measuring repetitions in reserve. JSCR 30(1):267–275. PubMed