Training Stress Balance: how to use it
This is the advanced companion to Training load explained. TSB is a model estimate of whether recent training load is above or below what you're used to. Here's exactly how it's calculated, why apps disagree, how to handle the first weeks of data, and why it should never be read on its own.
The formula in plain English
TSB = CTL − ATL. CTL ("fitness") is a roughly 42-day exponentially weighted average of your daily Training Stress Score, and ATL ("fatigue") is a roughly 7-day one [1]. Despite the "fitness" and "fatigue" labels, both are calculations from recorded load, not measurements of your body. Because ATL reacts faster than CTL, a hard week pushes TSB down and a few easy days push it back up. This is the fitness-fatigue idea behind the Banister impulse-response model [1].
How CTL and ATL update each day
A common way to write the daily update (with a time constant k of 42 for CTL and 7 for ATL) is:
CTLtoday = CTLyesterday + (TSStoday − CTLyesterday) ÷ 42
ATLtoday = ATLyesterday + (TSStoday − ATLyesterday) ÷ 7
Same-day or previous-day?
Some platforms calculate today's TSB from yesterday's CTL and ATL (so it reflects how you start the day), and others include today's workout. Others again use exact exponential factors such as e−1/42 instead of dividing by 42. The differences are small but real, so don't compare TSB values across apps. In this guide's example TSB includes the same day's TSS.
Starting values and new users
The recursion needs a starting CTL and ATL. Common choices are zero, your first day's TSS, or an estimate from earlier training. Because CTL has a 42-day time constant, the starting value influences it for weeks, so CTL and TSB can be misleading in the first weeks of data. How long depends on how far the starting value is from your real recent load, and it's worst for anyone who was training before they started recording.
Rest day or missing workout?
The model can't tell the difference. A day with no recorded workout counts as zero load, whether you rested or your watch didn't sync, or you did a session with no heart rate. Missing data therefore lowers CTL and ATL artificially and pushes TSB up, making you look "fresher" in the model than your training justifies. Check that every session was captured before reading the numbers.
Worked example (illustrative)
Invented numbers, rounded. Start: CTL 50, ATL 50, TSB 0.
| Day | TSS | CTL | ATL | TSB (CTL − ATL) |
|---|---|---|---|---|
| 1 | 100 | 51.2 | 57.1 | −6.0 |
| 2 | 120 | 52.8 | 66.1 | −13.3 |
| 3 | 0 | 51.6 | 56.7 | −5.1 |
| 4 | 150 | 53.9 | 70.0 | −16.1 |
| 5 | 40 | 53.6 | 65.7 | −12.1 |
| 6 | 0 | 52.3 | 56.3 | −4.0 |
| 7 | 0 | 51.1 | 48.3 | +2.8 |
Two zero-load days at the end turned TSB positive while CTL stayed above where it started. That's the arithmetic behind a taper in miniature. It only holds if those days really were rest days and not missing data.
How coaches typically read it
There are no universal TSB zones. The descriptions below are coaching heuristics, not validated thresholds, and the numbers behind them vary between athletes, sports and coaches [1]. Your own history (what TSB you raced well at, and how deep you can go before sleep and HRV suffer) is the better guide:
- Clearly positive: recent load well below longer-term load. Many coaches aim for this before races and tests.
- Around zero to slightly negative: recent load close to longer-term load, typical of steady training.
- Moderately negative: recent load above longer-term load, often seen in building phases. Some athletes feel heavy legs here; others don't.
- Very negative for many days: sustained high recent load. The model can't tell you whether you're coping, so check sleep, HRV and how you feel, and consider planning recovery.
Common mistakes
- Racing deep in negative TSB when the goal is performance.
- Tapering too long. TSB keeps rising, but CTL falls too, so some fitness is lost.
- Reading TSB without context. It ignores sleep, HRV, illness and life stress. Monitoring reviews recommend combining load measures with how the athlete is responding [2].
- Outdated FTP or threshold heart rate, which skews TSS and therefore everything downstream.
- Sudden load spikes. Big jumps relative to what you're used to have been associated with higher injury risk, mainly in elite team-sport athletes [3].
How RadiantHealth combines it
In the app, RadiantHealth calculates TSS for Strava workouts (beta; training load from Garmin workouts is not yet calculated and is being fixed) (true power-based TSS only when a Strava power stream is available; otherwise an estimate from the heart-rate stream, average heart rate or, as a last resort, duration), then calculates CTL and ATL as 42- and 7-day exponential averages and TSB including today's load. The briefing reads TSB alongside HRV and sleep and notes when load values are estimated.
The free TSB calculator is separate: paste at least 7 daily TSS values (42 or more recommended), oldest first. It uses the divide-by-42 and divide-by-7 recursion shown above, starting from your first value, so its numbers can differ slightly from the app's and from other platforms.
FAQ
What TSB should I race at?
A positive TSB on race day is a common coaching convention, not a proven target, and the right value depends on the athlete and event. Use past races to find what works for you.
Is negative TSB bad?
No. Moderately negative TSB is normal during building phases. Very negative TSB over long periods means sustained high recorded load. If you also see poor sleep or falling HRV, that's a reason to consider recovery.
Why does my TSB differ between apps?
Different TSS methods, starting values, exponential factors and same-day versus previous-day conventions. Compare trends within one app, not values across apps.
What is the difference between TSB and form?
They're the same thing. Some platforms call TSB "form".
Sources
- Allen H, Coggan AR, McGregor S. Training and Racing with a Power Meter, 3rd ed. VeloPress, 2019.
- Halson SL. Monitoring training load to understand fatigue in athletes. Sports Med 2014;44(Suppl 2):S139-S147. doi:10.1007/s40279-014-0253-z
- Gabbett TJ. The training-injury prevention paradox. Br J Sports Med 2016;50(5):273-280. doi:10.1136/bjsports-2015-095788
Try the free TSB calculator
Paste your daily TSS values to see CTL, ATL and TSB. No signup needed.
Try it →