Skip to content

Open-source gym & body-weight tracker, now in the App Store (iPhone + Apple Watch) and as an Android APK. Plan routines, log workouts, see which muscles are trained, import from Strong/Hevy/FitNotes. Self-host it or use it on the phone alone.

fix(stats): effort always estimates unrated sets instead of averaging rated ones

Сергей Горбатенкоcbe3eebe4402
diff --git a/frontend/src/views/Stats.jsx b/frontend/src/views/Stats.jsx
index 28fa2910d..5c2047775 100644
--- a/frontend/src/views/Stats.jsx
+++ b/frontend/src/views/Stats.jsx
@@ -21,7 +21,7 @@ import { maxRepsSeries } from '../lib/pyramid.js'
import { perSetSessions, perSetLines, dropOffSet } from '../lib/per-set.js'
import {
hasEffort, hasEstimableEffort, displayScale, scaleName, toScale, avgRir, effortSummary, effortWeeks,
- effortHistogram, isHardSet, HARD_RIR, MIN_RATED
+ effortHistogram, isHardSet, HARD_RIR
} from '../lib/effort.js'
import { anchorsByWorkout, resolveSetRir } from '../lib/recovery.js'
import { Button, Segmented, SelectRow } from '../components/ui.jsx'
@@ -315,23 +315,24 @@ function EffortCard({ S }) {
const [win, setWin] = useState(90)
const kind = displayScale(S)
const hd = scaleName(kind)
- const logged = effortSummary(S, win)
- // Thin or missing ratings fall back to estimates (same precedence fatigue scoring
- // uses), so imported histories get an effort card too. Well-rated windows keep
- // their logged numbers untouched - estimates never dilute a real average.
- const blend = logged.rated < MIN_RATED
+ // Every set without a logged rating is estimated (the same precedence fatigue scoring
+ // uses: logged effort first, Epley-inverse estimate against the 90-day anchor second),
+ // so the average always speaks for the whole training, not just the rated subset.
+ // Averaging rated sets only reads as RIR 0 for a history whose only ratings are
+ // imported sets taken to failure. Estimates never override a real rating, so a fully
+ // rated window still reads exactly its logged numbers.
// Same options as the fatigue map (unit + profile bodyweight), so a bodyweight set is
// estimated against the same body-mass-inclusive anchor in both places.
const opts = useMemo(() => ({ unit: S.unit, bodyweightKg: profileBodyweightKg(S) }), [S.unit, S.bodyweight])
const anchors = useMemo(
- () => (blend ? anchorsByWorkout(S.workouts, opts) : new Map()),
- [blend, S.workouts, opts],
+ () => anchorsByWorkout(S.workouts, opts),
+ [S.workouts, opts],
)
const resolve = (s, w, e) => resolveSetRir(s, s, e, w, anchors.get(w), opts)
- const sum = blend ? effortSummary(S, win, resolve) : logged
- const weeks = blend ? effortWeeks(S, win, resolve) : effortWeeks(S, win)
- const hist = blend ? effortHistogram(S, win, resolve) : effortHistogram(S, win)
- const estimated = blend && sum.est > 0
+ const sum = effortSummary(S, win, resolve)
+ const weeks = effortWeeks(S, win, resolve)
+ const hist = effortHistogram(S, win, resolve)
+ const estimated = sum.est > 0
const maxBin = Math.max(1, ...hist.map(b => b.n))
// The week's set count rides along in the tooltip, because the pair is the reading:
// volume up with effort up is fatigue piling up, volume up with effort flat is adaptation.