Managing Training Load: TSS, CTL, ATL and the Numbers That Actually Matter

Every endurance athlete eventually meets the same two failure modes: doing too little to improve, or doing too much and breaking down. Managing training load is the discipline of staying in the narrow band between them — training hard enough to adapt, recovering enough to absorb it. This guide walks through the metrics that make load visible, leaning heavily on the monitoring framework laid out by Mike Posthumus in Science to Sport's Monitoring Cyclist Training Load, and ends with the one rule that matters more than any number.

External load vs internal load

There are two ways to measure what a session cost you.

• External load is the work you did, independent of your body: watts, kilometres, metres climbed. A power meter is the classic external-load tool.
• Internal load is how your body responded to that work: heart rate, perceived effort, how wrecked you feel the next morning.

It is tempting to assume the precise, objective external number is the better one. Posthumus pushes back hard on that. As the article notes, there is "currently no peer-reviewed evidence demonstrating that power-based longitudinal load monitoring is more effective than methods using heart rate (HR), rating of perceived exertion (RPE), or subjective feeling." In other words: the fancy power metrics below are useful, but a runner with a heart-rate strap or even just an honest RPE score is not flying blind.

The session metrics: NP, IF and TSS

Modern cycling software builds its load model on three numbers, popularised by Hunter Allen and Andrew Coggan in Training and Racing with a Power Meter.

Normalised Power (NP) is a weighted average that better reflects the physiological cost of a ride than plain average power. As the article puts it, if your NP was 300 watts, the session was "equivalent to maintaining the same constant power (300 watts) for the same duration" — even if the actual ride was full of surges and coasting.

Intensity Factor (IF) is simply NP divided by your FTP — your functional threshold power, the effort you can hold for a prolonged period (roughly 30–90 minutes). An IF of 1.0 means you rode the whole session at threshold.

Training Stress Score (TSS) combines intensity and duration into a single number:

TSS = (NP / FTP)² × Duration (hours) × 100

The anchor point is easy to remember: "By definition, 100 TSS is the hardest you could possibly ride for 1 hour." A two-hour endurance ride at IF 0.65 scores about 85 TSS; a brutal one-hour threshold test scores 100. That single number is what the long-term model is built on.

If you don't have a power meter, the same logic works on internal load — multiply your session RPE (1–10) by duration in minutes to get session-RPE load (sRPE), a well-validated stand-in that needs nothing but a watch and honesty.

The long-term model: ATL, CTL and TSB

A single session's TSS is noise. The signal is how those scores accumulate and decay over time. The standard model tracks three rolling quantities:

• Acute Training Load (ATL) — your short-term load, typically the last 7 days. Think of it as your current fatigue.
• Chronic Training Load (CTL) — your long-term load, typically the last 42 days. Think of it as your current fitness.
• Training Stress Balance (TSB) — simply CTL minus ATL. Think of it as your current freshness, or "form".

The intuition is clean: fitness is built slowly and lost slowly (the 42-day window), fatigue arrives and clears quickly (the 7-day window), and form is high when you have built a deep base but recently backed off — exactly the state you want on race day.

Rather than a flat rolling average, the metrics are usually computed as an exponentially weighted moving average (EWMA), which weights recent days more heavily:

ATL_today = ATL_yesterday × e^(−1/7) + TSS_today × (1 − e^(−1/7))
CTL_today = CTL_yesterday × e^(−1/42) + TSS_today × (1 − e^(−1/42))
TSB = CTL − ATL

This isn't just cosmetic. The research Posthumus cites found that an EWMA of training load "can track both performance and risk of illness and injury extremely well" — better than a simple rolling average.

The acute:chronic workload ratio

That last point is the basis of the acute:chronic workload ratio (ACWR) — your acute load divided by your chronic load. Spike your acute load far above what your chronic base has prepared you for, and injury and illness risk climbs. In the study by Murray and colleagues, calculating the ratio with an EWMA "provides a more sensitive indicator of injury likelihood than rolling averages." Practically, it is a quantitative version of the oldest coaching rule there is: don't increase your training too fast.

There is no universal "right" number

The mistake beginners make with these tools is chasing someone else's targets. The optimal values are deeply individual. The article gives realistic ranges:

• Optimal CTL runs "from 70 TSS/day for some competitive age group athletes, to as high as 140 CTL for pro tour riders." A CTL that makes a professional fly will bury an amateur.
• Optimal race-day TSB varies too: some athletes "perform better with a TSB in the range of +10 to +20, whereas some simply feel better at a TSB of −5 to +5."

You find your own numbers by pairing load with performance over months. If CTL keeps climbing but your performance stagnates or declines, that is the signal you have passed your optimal range — not a cue to add more.

Load is feedback, not a target

This is the rule that outranks every formula above, and it is the one Posthumus is most emphatic about:

Training should never be prescribed with the sole purpose of achieving load goals.

The order of operations matters. First you build a sound plan from training principles — the right intensity distribution, the right balance of easy and hard, the right zones for each session. Then you use load metrics to check that the plan is being followed, that your hard blocks are genuinely hard and your recovery weeks genuinely easy, and that fitness is trending the right way. Riding extra junk miles just to hit a CTL number is how you manufacture fatigue with no adaptation.

And no model is complete. Performance, as the article concedes, "is extremely complex and there is currently no model which may account for all possible factors contributing to human variance." Which is why the most sensitive monitoring tool is often the simplest. Posthumus notes that just "asking the athlete 'How do you feel?' can elicit valuable feedback" and may be "more sensitive for the early detection of overreaching" than any power metric. Submaximal field checks — like the Lamberts Submaximal Cycle Test, where heart rate, power and recovery at a fixed effort flag fatigue early — formalise the same idea.

Putting it together

If you want a workable monitoring routine without overthinking it:

1. Score every session — by TSS if you ride with power, by session-RPE (effort × minutes) otherwise.
2. Watch the trend, not the day. Let CTL rise gradually; keep big week-to-week jumps in check.
3. Peak by raising your form. Build chronic load, then taper so TSB swings positive into your key event.
4. Anchor it to feel and performance. When the numbers and your body disagree, your body usually wins.
5. Never train to hit a number. Train the right way, and let the numbers report back.

For how the underlying training itself should be structured, see our guides to training philosophies and heart-rate zones, or set your benchmarks with the FTP and LTHR zones calculators.

Sources

This article draws primarily on the Science to Sport monitoring framework, and on the references it cites:

• Posthumus, M. Monitoring Cyclist Training Load (Part 1): External Load & Modern Cycling Metrics. Science to Sport, 20 October 2025.
• Allen, H. & Coggan, A. Training and Racing with a Power Meter. 2nd ed. VeloPress, 2010.
• Murray, N.B., Gabbett, T.J., Townshend, A.D. & Blanch, P. Calculating acute:chronic workload ratios using exponentially weighted moving averages provides a more sensitive indicator of injury likelihood than rolling averages. British Journal of Sports Medicine, 2017; 51(9): 749–754.
• Lamberts, R.P., Swart, J., Noakes, T.D. & Lambert, M.I. A novel submaximal cycle test to monitor fatigue and predict cycling performance. British Journal of Sports Medicine, 2011; 45(10): 797–804.