CTL, ATL and TSB: Fitness, Fatigue and Form Explained

Short answer: let your fitness number rise slowly, expect your form number to be negative in hard weeks, and arrive at a key race with form back above zero. Fitness, fatigue and form are two running averages of your training load, and the gap between them. Fitness is the slow one, weighted towards roughly the last six weeks; fatigue is the fast one, roughly the last week; form is fitness minus fatigue. (Apps often label them CTL, ATL and TSB.) They are calculations from your training data, not measurements of your body, so read them alongside how you feel (1, 2).

Here is how to use them in a normal training week, what they mean, and where they stop being useful.

How should you use fitness, fatigue and form?

Watch the trends, not the daily values.

• Fitness should rise slowly through base and build training, with small dips in easier weeks.
• Form will be negative in most hard weeks. That's expected: you're tired because you're training. As a rough guide from coaching practice, form between about −10 and −30 is normal in a hard block. If it sits below about −30 for more than a few days, ease off. These bands assume a mid-range fitness (roughly 40 to 80). Form scales with fitness, so if yours is low, a smaller negative number is already a hard block; some coaches think of it as a share of fitness instead, with about −10 to −30% normal in hard training.
• Every third or fourth week, most amateurs take an easier week, cutting load by roughly a third to a half (again a coaching rule of thumb). Coaches often recommend every third week for beginners and older athletes.
• Before a key race, reduce your training so form rises well above zero while fitness dips only a little. Many coaches aim for roughly +10 to +25 on race morning, though no single value suits everyone. Much higher can mean the taper ran long enough to cost some fitness.

If you're new to training or to these numbers: look at fitness once a week and keep its rise gradual. Ignore the daily swings in form.

If you've used them for a while: compare form with how sessions actually go. Say you used to complete threshold intervals comfortably at a form of −15 and now fail them at −15. Then the numbers are missing something: sleep, illness or work stress, fatigue carried over from earlier weeks that the 7-day setting underweights, or a threshold that's out of date. Listen to your body over the number.

Whatever your level, the numbers are only as good as the data going in. Results also depend on where you start: from a low base your body often improves faster for the same training, but the fitness number only rises as fast as you add load, so beginners should still keep its rise slow.

What do the three numbers mean?

All three start from TSS (Training Stress Score): the training load of one session, from its duration and how hard it was relative to your threshold, roughly the hardest effort you could hold for about an hour. One hour at threshold scores about 100. Apps that use TSS, Tito included, use power for rides with a power meter, heart rate otherwise, and pace for runs.

Fitness (CTL, chronic training load) is an average of your daily TSS in which recent days count most. Technically it has a 42-day time constant: a day's training still counts for about half as much after a month, and fades out over the following two to three months. If you trained the same amount every day, fitness would settle at your average daily TSS. Averaging 400 TSS a week, it settles near 57.

Fatigue (ATL, acute training load) is the same kind of average over a much shorter window (a 7-day time constant): a day's training counts for about half as much after five days. It rises fast after a hard few days and falls fast when you rest.

Form (TSB, training stress balance) is fitness minus fatigue. When you've trained harder than usual recently, fatigue is above fitness and form is negative. After a few easy days, fatigue drops below fitness and form turns positive.

The idea comes from the fitness-fatigue model that Banister and colleagues developed. Each session adds a lasting fitness effect and a short-lived fatigue effect. In the model, your performance on a given day reflects the balance of the two (3, 4).

A training block in numbers

Say you've been training steadily at about 350 TSS a week (for example, seven hours of mostly easy training with a little harder work). Fitness and fatigue both sit near 50, and form near 0. Then you do one hard week and one easy week. The numbers follow from the standard settings, rounded; your app may differ slightly in how it does the maths.

Week	Fitness	Fatigue	Form
Usual (350 TSS)	50	50	0
Hard (525 TSS, half as much again)	54	66	−12
Easy (245 TSS)	51	46	+5

A week at half as much again is a deliberate overload week, done occasionally and followed by an easier week, not a weekly increase. In the hard week you'd expect to feel tired, and the last sessions of the week to feel harder than the first. In the easy week, fitness slips back a little but stays above where you started. That dip is the average catching up with a lighter week; one easy week is very unlikely to cost you real fitness.

Taper. A meta-analysis of tapering studies in competitive athletes found the best results from about two weeks in which training volume was cut progressively by roughly 41 to 60%, with intensity and training frequency kept the same (5). For example, a 9-hour week might become about 6 hours, then about 4 in race week, not counting the race, keeping a few short efforts at race pace. As a rough guide, shorter races need shorter tapers. Fitness drifts down a little, fatigue drops a lot, and form rises. Don't add training to stop the fitness number falling: in the taper studies performance improved despite the cut in volume (5), so the drop mostly reflects the calculation; whatever fitness you lose over two weeks is outweighed by the fatigue you shed.

When form and feel disagree before a race. In the last few days, how you feel is a weak guide too: taper legs often feel flat, and nerves can make you feel great or awful. Don't add a hard session to "sharpen up", and don't rest extra days because the number looks low. Race to your pacing plan; don't change it because of the number or pre-race nerves. Adjust for the conditions on the day, such as heat, and pull out for real signs of illness, such as a fever or chest symptoms.

When a session goes badly. If form is well below zero and a key session falls apart, take an easy day rather than repeating the session. As a rough guide from coaching practice, two failed key sessions in a week call for an easier few days whatever the numbers say.

How fast should fitness rise?

Slowly enough that you can keep going. The change in fitness per week is often called the ramp rate. As a rule of thumb, and this is coaching practice rather than a research finding, many coaches treat a rise of about 5 points a week as sustainable for amateurs and more than about 8 as steep, to be held for a week or two at most. These are points, not percentages: if your fitness is low (say under 40), stay at the lower end, around 2 to 4 points a week. Beginners and anyone returning from injury should stay lower too. The hard week above added about 4 points.

The evidence behind any precise limit is weak. A critical review of a related measure, the ratio of recent to longer-term load, found no evidence that managing it reduces injuries (6), although Gabbett's review argued that rapid, excessive increases in load are likely behind many soft-tissue injuries in team sports (7). That supports building gradually; it doesn't validate a particular number.

What the numbers can't tell you

They assume everyone responds alike. The 42- and 7-day settings are general defaults. When researchers fitted the model to nine elite swimmers, the values varied widely from person to person: the fitness figure ranged from about 2 to 9 weeks, and the fatigue figure was usually well beyond 7 days, though each estimate was itself uncertain (8). Either way, the defaults are unlikely to fit you exactly: your own recovery may be faster or slower than the default, so treat form as approximate. A 2022 commentary advised against general constants for that reason, and recommended using the model to be "data-informed" rather than "data-driven" (2).

They miss anything the data misses. Sleep, illness, travel and work stress add fatigue that TSS doesn't record; heat is caught partly by heart-rate scores and not at all by power. Scores from heart rate and from power also aren't fully equivalent, so switching between them can move the numbers (2). Strength sessions and other activities without heart rate, power or pace may add little or nothing. A review of training-load monitoring found that very few markers have strong evidence behind them, that no single definitive marker has emerged, and that monitoring should be individual (1).

They depend on your threshold being right. If your threshold is set too low (your FTP, the power you can hold for about an hour; your threshold heart rate; or a threshold running pace that is too slow), every session scores too high, and fitness and fatigue are inflated; set too high, they're too low. As a rough guide, retest or update your threshold every six to eight weeks or at the end of each training block, with a field test or a recent race, and check that your app's value matches. Expect fitness and fatigue to shift when you do: that's the scoring changing, not your body.

Form isn't a prediction of your race. In the research version of the model, fitness and fatigue each get their own multiplier and time constants, fitted to the athlete's own performances (3, 4). Apps subtract one from the other, usually with the same default settings for everyone, which is simpler and not the same thing.

What if…

…you do several sports? Most apps add swim, bike and run TSS into one fitness number. That's useful for overall load, but a high combined fitness doesn't mean your running legs are ready for a big jump in running. Watch your weekly running time on its own, and raise it gradually even when the combined number says you can handle more.

…you've just connected your training history? Fitness builds from the first day of data. If only a few weeks are loaded, your fitness will read low for several weeks: a steady load takes about six weeks to reach about two-thirds of its final value.

…you were ill or injured? Fitness falls slowly and fatigue quickly, so form will look high. That doesn't mean you're ready for hard training. Come back gradually and let the ramp rate, not form, guide you. If you've had a fever, chest symptoms or an injury, check with a doctor or physio before returning to hard training.

The science behind it

Optional reading, for anyone who wants the detail.

Morton, Fitz-Clarke and Banister described the model mathematically in 1990. Each session produces fitness and fatigue effects that fade over time, and performance is the fitness effect minus the fatigue effect, each with its own multiplier. Model-predicted performance correlated significantly with actual performance in athletes and non-athletes, during training and tapering (3). Clarke and Skiba's teaching review explains the model alongside the critical power model, as complementary tools (4).

Its fit is reasonable but limited. In seven trained runners over 15 weeks, modelled performance correlated moderately to strongly with weekly 1,500 m time-trial results, depending on the load measure, with rTSS highest (9). A separate study tested the model's input rather than the model itself: in 15 competitive cyclists over a 10-week preseason, total load from measures that account for the individual, including TSS, related most strongly to changes in aerobic fitness (10). Both studies were small.

Statistical analyses are less kind. In nine elite swimmers, the fitted fatigue time constant averaged 19 days, against the usual 7, and each swimmer's estimate was uncertain: its 95% confidence interval spanned on average about 6 to 32 days. Some parameters were so strongly correlated that interpreting them separately was "worthless", in the authors' word (8). A 2025 statistical analysis found the model's settings could not be pinned down reliably from the data, and adding the fatigue part did not significantly improve its predictions (11). Busso proposed a version in which fatigue grows with training. It fitted better, in six previously untrained people, and the fitted model implied a rise then a fall: past a point, more daily training would lower performance (12). The simple version that apps use can't show that.

Practical conclusion: fitness, fatigue and form are a reasonable, widely used way to summarise training load, and the trends are useful. The precise values, especially form, carry more certainty on screen than the model supports.

References

1. Halson SL. Monitoring training load to understand fatigue in athletes. Sports Med. 2014. doi:10.1007/s40279-014-0253-z
2. Vermeire K, Ghijs M, Bourgois JG, et al. The Fitness-Fatigue Model: What's in the Numbers? Int J Sports Physiol Perform. 2022. doi:10.1123/ijspp.2021-0494
3. Morton RH, Fitz-Clarke JR, Banister EW. Modeling human performance in running. J Appl Physiol (1985). 1990. doi:10.1152/jappl.1990.69.3.1171
4. Clarke DC, Skiba PF. Rationale and resources for teaching the mathematical modeling of athletic training and performance. Adv Physiol Educ. 2013. doi:10.1152/advan.00078.2011
5. Bosquet L, Montpetit J, Arvisais D, et al. Effects of tapering on performance: a meta-analysis. Med Sci Sports Exerc. 2007. doi:10.1249/mss.0b013e31806010e0
6. Impellizzeri FM, Tenan MS, Kempton T, et al. Acute:Chronic Workload Ratio: Conceptual Issues and Fundamental Pitfalls. Int J Sports Physiol Perform. 2020. doi:10.1123/ijspp.2019-0864
7. Gabbett TJ. The training-injury prevention paradox: should athletes be training smarter and harder? Br J Sports Med. 2016. doi:10.1136/bjsports-2015-095788
8. Hellard P, Avalos M, Lacoste L, et al. Assessing the limitations of the Banister model in monitoring training. J Sports Sci. 2006. doi:10.1080/02640410500244697
9. Wallace LK, Slattery KM, Coutts AJ. A comparison of methods for quantifying training load: relationships between modelled and actual training responses. Eur J Appl Physiol. 2014. doi:10.1007/s00421-013-2745-1
10. Sanders D, Abt G, Hesselink MKC, et al. Methods of Monitoring Training Load and Their Relationships to Changes in Fitness and Performance in Competitive Road Cyclists. Int J Sports Physiol Perform. 2017. doi:10.1123/ijspp.2016-0454
11. Marchal A, Benazieb O, Weldegebriel Y, et al. Statistical flaws of the fitness-fatigue sports performance prediction model. Sci Rep. 2025. doi:10.1038/s41598-025-88153-7
12. Busso T. Variable dose-response relationship between exercise training and performance. Med Sci Sports Exerc. 2003. doi:10.1249/01.mss.0000074465.13621.37