Load01Muscles02Reaction03Formulas04Research05Your data06Pricing07FAQ08Get the beta, free

Every number, as its equation.

Nothing in Kovura is a black box. Pick any score the app shows, move its inputs, and watch the result work itself out with the same constants the app uses. Hover a highlighted value to see where it goes.

12 equationsAll on your phoneSource files named
Your score
Load
Muscles
Logging

Readiness

Whole-body recovery and fresh muscles, blended by how you train. Load is never part of it, and fresh muscles can't lift the score more than 15 points above a run-down body.

ReadinessEngine.swift · PersonaModel.swiftYour score
readiness = 0.5 × 58 + (1 − 0.5) × 78 = 68.0
cap = 58 + 15 = 73 ← under the cap
bands = 70+ ready · 40–69 train light · under 40 recover
58
78
Persona weight w
68TRAIN LIGHT
What it learns

w moves up to 0.15 toward whichever half has predicted how your workouts felt, after 30 rated sessions.

The honest limit

Muscles you haven't trained recently don't count toward the muscle half.

Recovery

Each signal becomes 0–100 where 50 is your usual, scored against your own last 14 days. Signals blend by weight, then your check-in answers add or take away a few points.

RecoveryEngine.swiftYour score
HRV = 50 + 20 × -0.6 = 38 weight 0.45
RHR = 50 − 20 × 0.3 = 44 weight 0.30
sleep = ½ guideline + ½ your 7.2 h normal, 6.5 h = 62 weight 0.30
blend = Σ score × weight ÷ Σ weight = 46.6
check-in: (3 − 3) × 5 = +0
-0.6
0.3
6.5
3
47BELOW YOUR USUAL
What it learns

Every baseline is yours: HRV needs 7 earlier readings, resting HR and reaction time need 3.

The honest limit

Yesterday's score carries over at half strength, and a sick or injured day is capped at 35. Neither is shown here.

Acute:chronic ratio

A fast exponential average of daily load against a slow one. λ = 2 ÷ (N + 1) with N = 7 and 28, the EWMA version Murray and Gabbett showed is more sensitive than rolling averages.

ACWR.swiftLoad
λ acute = 2 ÷ (7 + 1) = 0.25 λ chronic = 2 ÷ (28 + 1) = 0.069
floor = 0.3 × 60 = 18.0
ratio = 52 ÷ max(44, 18.0) = 1.18
zones = under 0.8 light · 0.8–1.3 low · 1.3–1.5 moderate · 1.5+ high
52
44
60
What it learns

Both averages are yours. A ratio appears after 14 days of history and 4 training days in the last 28.

The honest limit

The zone cutoffs come from earlier rolling-average research and are approximate.

Session load

Once you've finished a workout, each session counts as your single 0–10 session rating times its minutes (Foster 2001). Before that, or if you skip the rating, it's the sum of each set's RPE.

ACWR.swift · StrainMethodLoad
day load = 7 × 60 = 420
a gym session ends at Finish or 10 min after its last set, whichever is first
7
60
420ARBITRARY UNITS
What it learns

Nothing. It's what you did, counted the way sports science has for two decades.

The honest limit

Per-muscle load always uses set effort, so it won't add up exactly to the session total.

Variety

Seven identical days are their own warning sign. Monotony is the 7-day mean over its standard deviation; variety turns it the right way up.

ACWR.swiftLoad
monotony = 40 ÷ max(24, 40 ÷ 10) = 1.67
variety = 100 ÷ 1.67 = 60 50 at Foster's 2.0 line
40
24
60VARIETY OK
What it learns

Nothing: the 2.0 line is Foster's (1998).

The honest limit

Needs at least 2 days with load in the last 7.

Muscle readiness

A fitness-fatigue model for each of 17 regions. Fatigue fades over about a day, fitness over about ten. A rested muscle sits at 80.

MuscleReadinessEngine.swiftMuscles
fatigue = 16 × e^(−24 ÷ 31) = 7.4
fitness = 16 × e^(−24 ÷ 240) = 14.5
readiness = 80 + 20 × min(1, 14.5 ÷ 90) − 100 × min(1, 7.4 ÷ 35) = 62
map = fatigue ÷ 35 = 0.21 under 0.2 green · 0.2–0.5 a bit sore · 0.5+ very sore
16
24
Fatigue τ (hours)
What it learns

Each muscle's τ is tuned by your soreness check-ins, between 12 and 220 hours.

The honest limit

A model of recent load, not a measurement. Your check-in wins for the day.

How a set reaches muscles

Every movement pattern splits a set's effort across the muscles it works. Effort is the RPE you logged, else 10 minus reps in reserve, else 7.

MovementProfile.swift · SetEntry.swiftMuscles
bench press split = chest 50% · front delts 25% · triceps 25%
chest = 4 × 8 × 0.50 = 16.0
front delts = 4 × 8 × 0.25 = 8.0
triceps = 4 × 8 × 0.25 = 8.0
8
4
16.0CHEST LOAD
What it learns

Nothing: the splits are fixed per movement pattern.

The honest limit

Splits come from strength-and-conditioning references, not lab measurements for each exercise.

Sleep-adjusted recovery

Last night's sleep against your 14-day average nudges each muscle's recovery clock. Small on purpose, and capped.

SleepRecoveryEngine.swiftMuscles
ratio = 5.5 ÷ 7.5 = 0.73 held to 0.6–1.2
speed = 1 + 0.50 × (0.73 − 1) = 0.87 held to 0.8–1.1
5.5
7.5
0.5
87%RECOVERY CLOCK SPEED
What it learns

k, from 20+ soreness check-ins, down to zero if sleep makes no difference for you.

The honest limit

Total sleep only, not stages.

Learning your recovery speed

Each check-in compares how sore Kovura predicted a muscle would be with how sore you say it is, and nudges that muscle's τ.

MuscleSorenessEngine.swiftMuscles
error = 3 × 20 − 30 = 30
step = 30 ÷ 12 = 2.5, held to ±8 h
τ = 31 + 2.5 = 33.5 h
30
3
31
33.5hRECOVERS SLOWER
What it learns

Starts counting at check-in 3, fully yours at 14. You can also state it: τ = days × 24 ÷ ln 5.

The honest limit

Skipped within 12 hours of training that muscle, before soreness shows.

Estimated 1RM

From your own recent working sets: the average of Epley and Brzycki. Reps plus reps in reserve over 10 lower the confidence; over 15, the set isn't used.

OneRepMaxEstimator.swiftLogging
Epley = 225 × (1 + 5 ÷ 30) = 262.5
Brzycki = 225 × 36 ÷ (37 − 5) = 253.1
1RM = (Epley + Brzycki) ÷ 2 = 257.8 lb
225
5
258FROM YOUR SETS
What it learns

Lifts you've never done get a guess carried over from similar lifts, always labelled as a guess.

The honest limit

A formula, not a tested max.

Weight progression

Next to each exercise: add weight, hold, or go lighter, from how recovered the muscles it works are right now.

ProgressionEngine.swiftLogging
76 ≥ 70 → add 5 lb (2.5 kg), or 2.5 lb (1.25 kg) under 40 lb
40–69 → hold · under 40 → go lighter
next = 140 lb
76
135
+5ADD WEIGHT
What it learns

It follows your learned recovery speeds.

The honest limit

It doesn't see rep quality or form. Your call wins.

CNS reaction score

The median of five taps, scored against your own last 14 days. Slower than usual scores lower.

ReactionTimeCheckView.swift · RecoveryEngine.swiftLogging
z = (268 − 252) ÷ max(14, 6% of 252) = 1.06
score = 50 − 20 × 1.06 = 29 weight 0.25 in Recovery
268
252
14
29SLOWER THAN USUALOpen the full simulator
What it learns

Your own usual reaction time, after 3 readings.

The honest limit

A shortened psychomotor vigilance test, not validated on its own. One input, not a diagnosis.

Get the beta, free.

Kovura is in open beta on TestFlight for iPhone. It's free to download. Beta testers keep Premium free for good. After launch, Premium is $4.99 a month or $34.99 a year, with a 7-day free trial on the annual plan.