What Do TSS, CTL, ATL, and TSB Actually Calculate? The Capabilities and Limitations of Training Load Quantification Models

Why Training Needs to Be Quantified Into Numbers

Open any mainstream training analysis software, and you’ll almost always see a curve fluctuating over time, marked with terms like Fitness, Fatigue, and Form. Behind these lie the CTL, ATL, and TSB that this article will discuss, along with the foundational unit driving all three: TSS. This model was originally developed to solve a very practical problem: coaches and athletes needed a way to convert vague training experiences—like “rode for two hours today, heart rate stayed in a certain zone, power fluctuated up and down”—into numbers that could be compared, accumulated, and used to plan training cycles. Before quantification tools existed, assessing training load relied heavily on the athlete’s subjective feelings and the coach’s experiential judgment. Both are valuable, but both are also easily influenced by factors such as mood that day, sleep, and external weather conditions, making long-term trend tracking difficult.

This model is not meant to replace subjective feelings, but rather to provide a relatively objective, long-term accumulative, and cross-cycle comparable supplementary metric. Understanding what this model calculates—and what it cannot calculate—can help athletes use their training software more intelligently, rather than blindly trusting a score that pops up on the screen.

TSS: Compressing a Single Workout Into One Number

TSS (Training Stress Score) is the foundational unit of this entire model, representing the overall load a single workout places on the body. The concept of TSS originated in the world of power-based training, with the core logic being to consider both the “intensity” and “duration” of a workout, combining the two into a single score using a formula: the higher the intensity and the longer the duration, the greater the TSS; conversely, low intensity or short duration results in a low TSS. A frequently cited reference point is that if an athlete rides for exactly one hour at their Functional Threshold Power (FTP), the TSS for that workout will be approximately 100. This benchmark gives TSS an intuitive anchor: 100 roughly corresponds to a load equivalent to “sustaining threshold intensity at full effort for one hour,” with actual values fluctuating based on the athlete’s true riding intensity and duration.

TSS is useful because it compresses scattered training records (heart rate, power, time, route) into a single number that can be directly summed and compared. An easy one-hour ride might yield only 30 or 40 TSS, while a two-hour workout packed with threshold intervals could accumulate over 150 TSS. A long-distance endurance race, even at less-than-maximal intensity, can accumulate a very substantial TSS simply because of the extended duration. This additive property is the prerequisite that allows CTL and ATL to function.

It’s worth noting specifically that TSS was originally designed based on power meter data. For athletes without a power meter, training software typically offers an alternative algorithm based on heart rate (sometimes called hrTSS or similar). The logic is similar, but accuracy is limited by the lag in heart rate response (heart rate rises half a beat behind power output, making load assessment for short, high-intensity intervals less accurate). Running and swimming training software also have their own corresponding load quantification methods, based on the same principles but using different input data. Regardless of which estimation method is used, readers should understand that TSS is a “relative” rather than “absolute” precise metric. It helps us compare the relative load between different workouts, but it should not be treated as a physiological measurement with medical-grade precision.

CTL: The Rolling Average of Long-Term Fitness

With the TSS from each workout in hand, the next step is to look at the “long-term trend.” This is where CTL (Chronic Training Load, often labeled directly as Fitness in training software) comes in. The calculation logic of CTL is to take the daily TSS over a longer past period (the window commonly used by training software is roughly four to six weeks) and compute a weighted exponential moving average. Workouts closer to today receive higher weight; workouts further in the past receive lower weight, but they don’t completely disappear—they fade out gradually like an echo.

The intuition behind this design is that fitness accumulation is a slow process that requires time to build. A single workout won’t cause fitness to skyrocket overnight, but long-term, consistent training load will gradually allow the body to adapt to higher intensity and greater training volume. This cumulative effect is exactly what CTL aims to capture. When the CTL curve is steadily rising, it means the athlete’s overall training load over the past few weeks is higher than before, and the body is enduring and attempting to adapt to greater training stimuli. When the CTL curve is flat or declining, it means training load is stable or decreasing.

The practical use of CTL is mainly for planning long-term periodized training. Coaches and athletes can observe the rate of CTL curve increase to determine whether the rise in training volume is too aggressive. A generally accepted principle is that if CTL rises too quickly, it means training load has increased substantially in a short period, and the body hasn’t had time to adapt gradually, significantly raising the risk of injury and overtraining. If it rises too slowly or plateaus for a long time, it may mean the training stimulus is insufficient to drive further fitness gains. There is no universally applicable “safe rate of increase” number, because individual recovery capacity, training background, age, and life stress all affect how quickly the body can handle increasing load. This is also why, even when two different athletes see the same CTL number on the same training software, what that number means for each of them can be completely different.

ATL: The Immediate Reflection of Short-Term Fatigue

ATL (Acute Training Load, often labeled as Fatigue in training software) uses the same calculation logic as CTL—both are exponential moving averages of TSS—but the difference is that ATL uses a much shorter time window, typically around one week. This makes ATL highly sensitive to training changes in the most recent days, quickly reflecting whether an athlete has recently experienced a dense block of training load.

ATL exists to compensate for CTL’s slow response. Because CTL’s window is long, even if an athlete suddenly schedules a very intense training camp over a few days, CTL will only rise slowly and slightly, failing to provide an immediate warning that “the load over these past few days is already quite demanding.” ATL, on the other hand, rapidly reflects this short-term load accumulation, becoming an important reference for judging whether the athlete is currently in a fatigued state. When ATL spikes quickly and is clearly higher than CTL, it means the athlete’s recent training load far exceeds their long-term accumulated fitness level, and the body is in a relatively fatigued state where recovery hasn’t caught up. Continuing to pile on high-intensity training in this state can easily trigger overtraining or sports injuries.

TSB: The Gap Between Fitness and Fatigue—In Other Words, Form

TSB (Training Stress Balance, often labeled as Form in training software) is the value obtained by subtracting CTL and ATL from each other. The formula logic is roughly long-term fitness minus short-term fatigue. This number aims to capture the athlete’s current “readiness”: if TSB is negative, it means short-term fatigue (ATL) is higher than long-term fitness (CTL), and the body is in a state of accumulated training load without full recovery. In this state, it’s appropriate to continue accumulating training stimulus, but not suitable for pursuing peak performance. If TSB is positive, it means short-term fatigue has subsided to below the long-term fitness level, and the body is in a relatively fresh, fully recovered state. This is exactly the effect a taper aims to achieve, and it’s the curve direction athletes hope to see as they prepare for an important race.

TSB is regarded by many athletes and coaches as the most practically valuable metric in this model, because it converts the abstract question of “how good do I feel?” into a trackable numerical trend, providing a reference for taper planning: by gradually reducing training volume (thereby lowering ATL) while trying to maintain the accumulated CTL without dropping too much, TSB can gradually turn positive before race day, theoretically allowing the athlete to arrive at the start line in a fresher physical state.

What This Model Can Help You Decide

The primary value of this TSS/CTL/ATL/TSB model lies in providing a long-term, trackable, cross-cycle comparable view of training load, filling the gap left by relying purely on subjective feelings. Specifically, this model can help athletes make the following types of decisions:

First, planning the overall direction of periodized training—by observing the long-term trend of the CTL curve, you can confirm whether training volume is steadily increasing according to the periodization plan, or whether enough fitness base has accumulated after the base phase to transition into the specific and peak phases. Second, detecting early whether short-term load is excessive—through the gap between ATL and CTL, you can judge whether recent training has been scheduled too densely, allowing for early intervention and adjustment before fatigue accumulates to a point that is difficult to recover from. Third, planning the pace of the taper—by observing the rate of change in TSB, you can determine how long the taper should last and how much training volume should be reduced, so the body reaches an ideal state of readiness on race day. Fourth, enabling longitudinal comparisons across cycles and seasons, allowing athletes to objectively compare overall load across different years and training phases, serving as a reference for adjusting the next cycle’s training plan.

What This Model Cannot Do: Limitations and Common Misuses

Although this model provides a useful quantitative perspective, it has several structural limitations. If athletes over-rely on the numbers while ignoring the assumptions behind the model, they can easily make flawed judgments.

First, this model quantifies “training load,” not “training effect” or “fitness improvement.” A rising CTL only indicates that the total training stimulus an athlete is absorbing is increasing; it does not guarantee that the body has actually produced the expected adaptation. If an athlete is chronically undernourished, sleep-deprived, or under excessive life stress, actual fitness and race performance may stagnate or even regress even as CTL continues to climb. In such cases, a clear gap emerges between the numbers and true physical condition—this is also where the model is most commonly misused: equating “doing more load” directly with “improving more.”

Second, this model completely ignores sources of life stress outside of training load. Work stress, sleep quality, emotional state, illness, injury, and even the extra physiological burden caused by weather all materially affect the body’s ability to recover and respond to training, yet none of these factors are reflected in the calculation of TSS, CTL, ATL, or TSB. An athlete whose TSB has nominally turned positive and who should theoretically be in good form may, if simultaneously experiencing severe sleep deprivation or high-pressure life events, be in far worse actual condition than the numbers suggest. Therefore, this model should be interpreted alongside other signals such as subjective fatigue ratings, sleep quality, and resting heart rate, rather than relying on a single score in isolation for decision-making.

Third, the calculation of TSS is highly dependent on the accuracy of baseline values such as Functional Threshold Power (FTP) or corresponding heart rate thresholds. If these baselines become outdated—for example, if the athlete’s fitness has improved or declined but the baselines have not been updated—all subsequent derived TSS, CTL, ATL, and TSB values will carry systematic bias, causing the model’s trend direction to remain correct while the absolute values lose meaning. This is why it is recommended to periodically retest and update baseline values—for example, every training cycle or every few months—otherwise the accumulated load numbers will become increasingly unreliable over time.

Fourth, TSS values across different sports and different training modalities are not directly comparable. For example, a long low-intensity endurance ride and a short high-intensity interval session may yield similar TSS numbers, but the fatigue patterns they impose on the body—metabolic fatigue, neural fatigue, and degree of muscle damage—are fundamentally different. The same TSS number can hide completely different actual recovery requirements, which is why experienced coaches interpreting these numbers do not look at a single score alone but also consider the actual content and modality of the training.

Fifth, the model’s parameters—time window lengths and the decay rate of exponential weighting—are general rules derived from observations of the general population, not calibrated individually for each athlete. In reality, recovery rates vary considerably between individuals. An athlete with strong recovery capacity may tolerate a faster CTL ramp rate than the model suggests, while another with weaker recovery capacity or higher concurrent life stress may need a more conservative load progression than the model recommends. Therefore, this model provides a “reference baseline under population-level general rules,” not a personalized prescription, and should always be used alongside your own long-term accumulated understanding of how your body responds.

Why This Model Has Been Widely Adopted: The Planning Power of Accumulability

The reason this model has been so widely adopted in the endurance sports training community lies in the accumulability of TSS. This sounds simple, but it actually solves a long-standing problem in training planning. Before this model existed, coaches and athletes comparing “is this week’s training volume sufficient” or “is this month too much training” often had to rely on crude metrics like training hours. But training hours completely ignore intensity: one hour of easy recovery riding and one hour of all-out interval training impose vastly different loads on the body. If training volume is planned using hours alone, it is easy to underestimate the actual load of high-intensity sessions and equally easy to overestimate the burden of large volumes of low-intensity hours. Once TSS merges intensity and duration into a single number, coaches and athletes finally have a consistent unit to accumulate and compare different types of training. This gives periodized training planning a common yardstick and makes visualizing long-term trends—namely, the CTL curve—feasible.

This also explains why the model only truly became widespread after the rise of digital training platforms. Manually calculating and accumulating daily TSS, then computing rolling exponential moving averages, is extremely tedious for the average user. But with training software automatically linking data from power meters, heart rate monitors, and other devices, these calculations can be completed automatically in the background. Athletes open the app and see a curve changing over time, dramatically lowering the barrier to applying this model. This convenience is also one of the reasons the model is easily misused: precisely because the numbers are obtained effortlessly, athletes easily develop the habit of “letting the numbers dictate training,” while overlooking the entire set of assumptions and limitations behind those numbers. This is also why this article has devoted particular attention to explaining the model’s limitations.

The Dynamic Relationship Between CTL, ATL, and TSB: A Practical Scenario Demonstration

To make the abstract formulas easier to understand, a common training scenario can be used to illustrate how the three interact. Suppose an athlete maintains a steady but not intense training rhythm during the base period. CTL rises slowly and steadily, while ATL fluctuates in small increments around CTL, and TSB remains roughly in a range near zero or slightly negative. This represents a healthy training state where the athlete is “continuously accumulating load while roughly keeping up with recovery.”

Next, entering the specific preparation period, the athlete decides to arrange a one-to-two-week high-intensity training camp with significantly increased training volume. During this period, daily TSS greatly exceeds normal levels, and ATL rises rapidly, much faster than CTL (because ATL’s time window is short and more sensitive to recent changes). TSB therefore quickly turns clearly negative. During this phase, athletes typically subjectively feel increased fatigue, and training performance may stagnate or even decline—this is a common and normal phenomenon during such a deliberately load-stacked phase, provided that a corresponding recovery period follows this high-load block.

After the high-load training camp ends, the athlete enters a taper period with significantly reduced training volume. Because of its short time window, ATL declines noticeably faster than CTL, so the gap between CTL and ATL begins to narrow or even reverse, and TSB gradually recovers from negative territory. If the taper is well planned, ideally the athlete will see TSB turn positive before the target race—that is, reaching a favorable balance between fitness and fatigue. This is precisely the core goal of the taper period. This cycle of “accumulating load, generating fatigue, releasing fatigue through tapering, and revealing fitness” recurs throughout the entire training periodization plan. The interactive relationship among the three curves of CTL, ATL, and TSB is, in essence, a numerical depiction of the shape of this physiological adaptation cycle.

Numbers from Different Training Software May Not Be Directly Comparable

Another practical detail that is easily overlooked is that different training software—or even different versions of the same software—may use slightly different parameters when calculating TSS, CTL, ATL, and TSB (such as the time constant of the exponential moving average or the specifics of heart rate conversion formulas). This means that even with the same raw training data, the CTL or TSB figures calculated by different software may differ to some degree. This is also why it is recommended that athletes consistently use the same software and the same calculation logic over the long term to track their trends, ensuring the numbers remain comparable. If you switch training software midway, you should expect a one-time jump in the numbers, which usually reflects differences in calculation methods rather than a dramatic short-term change in fitness. There is no need to overreact or doubt your training results.

How to Practically Use This Model

For amateur enthusiasts in Taiwan, a pragmatic approach is to treat this model as a “trend dashboard” rather than a “precise diagnostic tool.” Regularly check the direction of the CTL curve to confirm that your training volume is progressing steadily according to your periodization plan, rather than obsessing over the exact TSS score of any single day. Observe the gap between ATL and CTL as one of the reference signals for deciding whether an extra recovery day is needed, but interpret it alongside your subjective fatigue levels, sleep quality, and morning resting heart rate. If the numbers indicate good condition but your body feels unusually exhausted, you should trust your body’s signals first rather than the score on the screen. When planning the taper before an important race, you can use the TSB curve’s trajectory as a basis for adjusting training volume, but the actual reduction in volume and the number of taper days should still be flexibly adjusted based on your past tapering experience, the importance of the race, and your current physical state. You should not mechanically apply a fixed number of days or a fixed TSB value as a standard answer.

It is especially worth noting that if you ever find yourself persistently in a low TSB state (long-term negative values) without arranging corresponding recovery or taper periods, accompanied by declining sleep quality, low mood, performance that fails to improve or even declines, or repeated illness or injuries, these are all warning signs that require heightened vigilance. They indicate that your body may have entered a state of recovery deficit. In such cases, you should prioritize reducing training load rather than stubbornly pushing through the training plan based on the numbers shown by the model. If necessary, consult a coach or sports medicine professional for evaluation. Ultimately, the training load quantification model is a supporting tool. It helps athletes observe long-term trends more systematically and make more evidence-based periodization decisions, but it cannot replace sensitivity to your own body’s signals or the judgment of professionals. The two should complement each other, rather than letting numbers override the body’s true responses.

Key Takeaways and Action Checklist

• TSS is a quantitative unit of a single training session’s load, considering both intensity and duration. It is the foundation for calculating all subsequent metrics, and its accuracy depends heavily on whether baseline values (such as FTP) are updated in a timely manner.
• CTL is a rolling average of training load over the long term (approximately four to six weeks), representing the long-term fitness accumulation trend. It is suitable for planning the pace of progressive training volume increases in periodization.
• ATL is a rolling average of training load over the short term (approximately one week), representing the recent accumulation of fatigue. It reflects short-term load from intensive training more quickly than CTL.
• TSB is the difference between CTL and ATL, representing your current level of recovery and readiness. It is an important reference indicator for planning the pace of the taper period.
• This model quantifies “training load” rather than “training effect.” A rising number does not guarantee that fitness is actually improving; it must be cross-validated with actual race and test performance.
• Life stress, sleep, mood, illness, and other factors are not reflected in this model. When interpreting the numbers, be sure to combine them with subjective feelings and other physiological signals.
• Regularly update baseline values such as FTP; otherwise, long-term accumulated load numbers will develop systematic bias and lose their reference value.
• Use this model as a long-term trend dashboard rather than a precise diagnostic tool. Any persistent abnormal fatigue signals should be trusted as a priority response to your body’s reactions, and seek help from a professional coach or sports medicine professional when necessary.

Related Reading

• Training Stress Score and Performance Management: How to Read CTL, ATL, and TSB Without Being Misled
• Training Load and Fatigue Management: Applications of CTL, ATL, and TSB
• TSS, CTL, ATL, TSB Complete Guide: Calculation Principles and Applications of Training Stress Index
• TSS, CTL, ATL: A Complete Analysis of the Three Major Training Load Indicators