How Training Load Models Work: TSS, ATL, CTL, TSB

You finish a big training week and feel great. The next week you do the same volume and feel wrecked. Nothing changed except how much fatigue you were carrying into it.

Training load models exist to explain exactly this. They quantify how stress accumulates, how fitness builds, and how the balance between the two determines whether you feel sharp or shattered on any given day.

What is training load, really?

Every workout creates stress. A hard interval session creates more than an easy jog. A two-hour ride creates more than a thirty-minute spin. Training load is simply a way to put a number on that stress so you can track it over time.

The most common unit is Training Stress Score, or TSS. Originally developed for cycling using power meters, the concept has been adapted for running (rTSS), swimming (sTSS), and heart-rate-based activities (hrTSS). The details differ, but the principle is the same: harder and longer sessions produce bigger numbers.

TSS accounts for both intensity and duration. A 60-minute session at threshold might produce the same TSS as a 90-minute session at moderate effort. This makes it possible to compare fundamentally different workouts on a single scale.

The formula itself is straightforward for power-based training: TSS equals the duration in seconds, multiplied by the normalized power squared, divided by your FTP squared, divided by 3600, and then multiplied by 100. What that boils down to is a ratio of how hard you went relative to your threshold, scaled by how long you sustained it. An hour at exactly your FTP produces a TSS of 100. That’s the reference point everything else is calibrated against.

The fitness-fatigue model

Here’s where it gets interesting. Sports scientists in the 1970s proposed that every training session produces two competing effects: a fitness gain and a fatigue cost. Both decay over time, but at different rates.

Fitness builds slowly and fades slowly. Fatigue builds quickly and fades quickly. Your actual performance readiness at any point is the difference between accumulated fitness and accumulated fatigue.

This is called the impulse-response model, or the fitness-fatigue model. It was formalized by Banister and colleagues in 1975, and it remains the foundation of most training load tracking systems today. The elegance of the model is that it reduces an enormously complex biological process into something you can calculate with basic exponential decay math.

The original research studied swimmers and weight lifters, tracking performance predictions against actual results. What they found was that a two-component model (one positive, one negative) predicted performance remarkably well, even with relatively simple inputs. Decades of refinement have followed, but the core insight hasn’t changed.

CTL: your fitness trend

Chronic Training Load (CTL) represents your fitness. It’s an exponentially weighted moving average of your daily TSS, typically calculated over 42 days. Think of it as the answer to “how much training has this person been absorbing over the past six weeks?”

A higher CTL means you’ve been consistently training more. It rises slowly when you train hard and falls slowly when you rest. This matches how fitness actually works in the body. You don’t get fit from one big week, and you don’t lose fitness from one easy week.

The 42-day window isn’t arbitrary. It roughly corresponds to the time course of physiological adaptations like mitochondrial biogenesis, capillary development, and neuromuscular efficiency gains.

Mathematically, today’s CTL equals yesterday’s CTL plus the difference between today’s TSS and yesterday’s CTL, divided by 42. Each day, the average nudges slightly toward whatever you did today. Miss a day and it nudges toward zero. Stack consistent training and it climbs steadily.

One thing worth noting: CTL doesn’t tell you what kind of fitness you have. A CTL of 80 built on long slow runs represents different physiological adaptations than a CTL of 80 built on short intense intervals. The number captures volume and intensity together, but not the distribution. You still need to think about what your training actually develops, which is why understanding your training zones remains important alongside load tracking.

ATL: your fatigue trend

Acute Training Load (ATL) represents your fatigue. Same math as CTL, but calculated over a shorter window, typically 7 days. It answers: “how much stress has this person absorbed recently?”

ATL responds fast. One very hard day spikes it. A few rest days bring it back down. This also matches reality. You feel yesterday’s hard session in your legs today. By Thursday, Monday’s workout is mostly processed.

The relationship between ATL and CTL is where the insight lives. When ATL is much higher than CTL, you’re accumulating fatigue faster than you’re building fitness. That’s fine for a focused training block, but unsustainable for long.

The ratio between ATL and CTL is sometimes called the acute-to-chronic workload ratio (ACWR). Research in team sports has explored whether this ratio predicts injury risk, with some studies suggesting that rapid spikes (ratios above 1.5) correlate with higher injury rates. The evidence is still debated, but the principle is sound: sudden jumps in training load relative to what you’re conditioned for increase risk. Gradual progressions are safer than dramatic ones.

TSB: the balance

Training Stress Balance (TSB) is simply CTL minus ATL. It represents your form, or readiness to perform.

A positive TSB means your fitness exceeds your recent fatigue. You’re fresh. A negative TSB means recent training stress is outpacing your accumulated fitness. You’re digging a hole.

Runners and cyclists have known this intuitively forever. You taper before a race, reducing volume while maintaining intensity, specifically to let ATL drop while CTL stays relatively stable. The result is a rising TSB. You arrive at the start line fresh but fit.

Some coaching platforms use positive TSB values as a practical freshness indicator, but optimal race-day TSB varies by athlete and should be interpreted individually. Sustained strongly negative TSB suggests recent load is high relative to your chronic fitness, though the exact threshold varies by person. These aren’t rigid thresholds. They’re guardrails that help you notice patterns.

The tricky part of TSB is that maximum freshness and maximum fitness pull in opposite directions. Complete rest maximizes your TSB, but your CTL drops. Too much training maximizes CTL but buries TSB into negative territory. The art of peaking for an event is finding the sweet spot where CTL is high and TSB has recovered enough to let you access that fitness.

The math behind the curves

If you want to understand what’s happening beneath the surface, the exponentially weighted moving average works like a filter. Recent days contribute more than older days. A session from yesterday weighs more heavily in ATL than a session from five days ago, even though both fall within the 7-day window.

The decay constant determines how quickly old data loses influence. For ATL with a 7-day time constant, each day’s contribution decays by roughly 13% per day. For CTL at 42 days, the decay is about 2.3% per day. After two time constants (14 days for ATL, 84 days for CTL), a given session’s contribution has faded to about 13% of its original weight.

This explains why a single rest day barely moves CTL but noticeably reduces ATL. And why two weeks off significantly impacts your fitness (CTL drops by roughly 28%) but nearly eliminates acute fatigue (ATL drops by roughly 86-87%). The different time scales create the dynamic tension that makes the model useful.

You don’t need to calculate any of this by hand. But understanding the mechanism helps you interpret the numbers when they seem surprising. If your CTL drops faster than expected, it’s probably because you had a big block of training 4-5 weeks ago whose influence is now fading as the exponential weighting shrinks its contribution.

A practical example

Say you’ve been running four days a week, with a load that averages out to 50 TSS per day once rest days are counted. Your CTL settles around 50. Your ATL hovers near the same. TSB sits close to zero. You’re in a steady state.

Now you add a long run on Saturday that scores 120 TSS. Your ATL jumps. Your CTL barely moves because it’s a 42-day average and one day can’t shift it much. TSB goes negative. You feel it on Sunday.

If you follow that with an easy week (averaging 30 TSS/day), ATL drops quickly. CTL dips slightly. TSB recovers to positive territory. You feel good again, and you’re marginally fitter than before because that big session contributed to long-term adaptation.

That cycle of loading and recovering is periodization in a nutshell. Training load models just make the pattern visible in numbers rather than relying entirely on feel.

Ramp rate: how fast is too fast?

One of the most practical applications of CTL is tracking your ramp rate. This is how quickly your CTL is climbing week over week. A common guideline is to increase CTL by no more than 5-7 points per week for experienced athletes, and 3-5 for newer ones.

Push the ramp rate too high and you’re adding training stress faster than your body can adapt. The workouts might feel fine for a week or two, but fatigue accumulates beneath the surface. By week three or four, performance drops, motivation disappears, and nagging pains appear.

This is distinct from a single hard week. You can spike ATL during an intentional overreaching block and recover from it. The danger is sustained aggressive ramp rates over multiple weeks with no relief. That’s the pattern that leads to overtraining, or more accurately, under-recovering.

Tracking ramp rate also helps you plan training blocks backward from a goal event. If your current CTL is 45 and you want to be at 70 for race day, that’s 25 points to gain. At 5 points per week, you need at least 5 weeks of building, plus a taper. Now you have a timeline that’s rooted in physiology rather than wishful thinking.

Where these models are honest and where they’re limited

The fitness-fatigue model is a simplification. It treats all stress as equivalent (a tempo run and a hill sprint session might produce similar TSS but stress different systems). It doesn’t know you slept poorly, skipped meals, or are fighting a head cold.

It also assumes the 7-day and 42-day time constants work for everyone. In practice, older athletes may accumulate fatigue differently than younger ones. Someone returning from a long break responds differently than someone with years of consistent training. The model gives you a useful lens, not a perfect mirror.

This is why training load works best alongside other signals. How your nutrition supports recovery matters enormously. So does sleep quality, resting heart rate trends, and subjective feel. The number tells you what happened. You still need context to decide what to do about it.

Another limitation: TSS treats a score of 100 from a two-hour easy ride identically to 100 from a one-hour threshold session. Your body does not. The muscular damage, glycogen depletion, and neural fatigue from those two sessions are quite different. Some newer models attempt to address this by tracking load in different intensity zones separately, but the single-number approach remains dominant because of its simplicity.

Heart-rate-based TSS

Not everyone trains with a power meter. Heart rate provides a reasonable proxy for internal training stress, and hrTSS uses your heart rate data, threshold heart rate, and session duration to estimate load.

It’s less precise than power-based TSS because heart rate responds to heat, caffeine, stress, and fatigue. But for most people doing a mix of activities, heart-rate-based load tracking is more than good enough to spot trends, flag overreaching, and guide recovery timing.

The calculation uses a TRIMP-style approach (Training Impulse), weighting each minute of exercise by how high your heart rate is relative to your threshold. Time spent at high intensities contributes disproportionately more load than time at low intensities, which aligns with the exponential nature of physiological stress.

For runners without a power meter, pace-based estimates (rTSS) offer another alternative. These use your functional threshold pace and the pace of each run to estimate stress. They work well on flat terrain but can underestimate load on hilly routes where effort exceeds what pace alone would suggest.

If you’re interested in what the various training metrics mean in practice, our training metrics overview covers the broader landscape beyond just load models.

Multi-sport load tracking

Things get more complex when you train across multiple sports. A triathlete might swim in the morning, bike at lunch, and run in the evening. Each session produces its own TSS variant. The question is: can you sum them?

Generally, yes. The fitness-fatigue model operates on total stress regardless of source. Your body doesn’t reset between activities. A 50 TSS swim followed by a 70 TSS run produces 120 TSS of total daily stress, and your fatigue accumulates accordingly.

But there’s a nuance. Sport-specific fitness doesn’t transfer perfectly. A high CTL built primarily from cycling doesn’t make you a fit runner. Your cardiovascular system benefits from all aerobic training, but muscular and structural adaptations are specific to the movement pattern. A combined CTL is useful for understanding total systemic stress, but sport-specific CTL tells you more about your readiness in each discipline.

This is one reason multi-sport athletes often track separate load curves for each activity alongside a combined total. The combined number helps prevent overall overreaching. The sport-specific numbers help identify whether you’re actually progressing in each discipline or just maintaining.

What you can do with this today

You don’t need software to apply these concepts, though tracking makes it far easier to spot patterns over weeks and months.

Start by paying attention to accumulated load, not just individual sessions. A 90 TSS workout means something different at the start of a recovery week than at the end of a build week. Context matters. The same session can be productive or destructive depending on what preceded it.

Notice how you feel when you’ve had several big days in a row versus after a lighter stretch. That subjective experience correlates with TSB more than most people expect. Building a training program that respects this cycle of stress and recovery is what separates progress from burnout.

The real value of training load models isn’t in the daily numbers. It’s in the trend across weeks. A gradually rising CTL with periodic recovery weeks is the signature of sustainable training. A flat CTL with wild ATL swings suggests inconsistency. A relentlessly climbing ATL with no recovery is a red flag, regardless of how motivated you feel in the moment.

At VegaLoop, we calculate these metrics from your logged activities so the pattern is visible without requiring a spreadsheet. But the principle holds whether you track it formally or not: fitness is built through consistent, progressively dosed stress with adequate recovery. Training load models are just a way to keep yourself honest about whether you’re actually doing that.