Understanding Training Stress Metrics and their Origins

In modern endurance sports, training decisions are increasingly guided by data. Coaches no longer rely solely on intuition or training hours when evaluating an athlete’s workload. Instead, many use training load frameworks that help estimate how stress accumulates in the body over time and how athletes adapt to that stress.

As discussed in the previous article, modern training systems typically follow a two-step logic. First, platforms estimate the stress of individual training sessions using a workload metric. Second, this stress is used within a fitness–fatigue model to estimate how training influences performance over time.

One of the most widely recognized implementations of this approach is the Performance Management Chart (PMC), which visualizes metrics such as Chronic Training Load (CTL), Acute Training Load (ATL), and Training Stress Balance (TSB). These values represent long-term training accumulation, short-term fatigue, and the balance between the two, providing coaches with a structured way to monitor athlete readiness.

It is important to understand that many of the commonly known metric names originate from specific training platforms. Terms such as Training Stress Score (TSS®), Normalized Power (NP®), and Intensity Factor (IF®) are registered trademarks of Peaksware, the company behind TrainingPeaks. In addition, terminology such as CTL, ATL, TSB, and the Performance Management Chart is widely associated with the TrainingPeaks and WKO software ecosystem.

However, while these names and implementations are platform-specific, the scientific principles behind them are not proprietary. The theoretical foundation of most training load models comes from the impulse–response framework developed by Dr. Eric W. Banister in 1975, which describes performance as the interaction between accumulated fitness and fatigue. This model was later applied to endurance sports by researchers and coaches, including Dr. Andy Coggan in the context of cycling performance analysis.

Within this framework, training sessions are first converted into a measure of training load, sometimes referred to as a training impulse. In the TrainingPeaks ecosystem, this impulse is typically calculated using Training Stress Score (TSS), which combines workout duration and intensity relative to an athlete’s functional threshold power.

Other platforms estimate training load using different physiological signals. For example, Garmin calculates Training Load using heart-rate derived EPOC, Strava uses Relative Effort, and Whoop estimates cardiovascular Strain. Although these systems rely on different algorithms and data sources, they all attempt to quantify the physiological cost of training sessions.

Once this workload is quantified, platforms apply variations of the Banister impulse-response model to estimate how training stress influences fitness, fatigue, and readiness over time. While the terminology differs between systems, the underlying goal remains the same: understanding how training stress accumulates and how the body adapts to repeated training stimuli.

Understanding these origins is important for coaches because it clarifies that many modern metrics are different implementations of the same scientific idea. Rather than focusing on the exact terminology used by a specific platform, coaches should focus on the broader concept of how training load, recovery, and adaptation interact to influence performance.

The Scientific Basis: The Banister Model

One of the foundational models in training load analysis is the Banister impulse-response model, introduced in the 1970s.

The central idea is that training produces two simultaneous physiological responses:

1. A positive fitness adaptation

2. A negative fatigue response

Performance at any given time is the balance between these two effects.

Performance(t)=Fitness(t)−Fatigue(t)

Performance(t)=Fitness(t)−Fatigue(t)

Fitness develops relatively slowly but persists for a longer time, while fatigue accumulates quickly but dissipates faster. This dynamic explains why athletes often feel tired during heavy training blocks but perform well after tapering before competitions.

Modern training load metrics such as CTL, ATL, and TSB are simplified implementations of this theoretical model.

Training Impulse (TRIMP)

Before power meters were widely available, researchers needed a method to quantify training stress using heart rate data. This led to the concept of Training Impulse (TRIMP).

TRIMP estimates training stress using heart rate data.

TRIMP=Duration×ΔHR×y

TRIMP=Duration×ΔHR×y

Where:

• Duration = training duration

• ΔHR = normalized heart rate elevation

• y = weighting factor for intensity

The idea behind TRIMP is simple: longer sessions and higher intensity both increase physiological stress.

Practical Example

Consider two sessions:

• 60 minutes of moderate endurance training

• 30 minutes of high-intensity intervals

Despite the shorter duration, the interval session may produce a similar TRIMP value because high-intensity efforts place significantly greater stress on the cardiovascular system.

TRIMP is widely used in academic research and sports science, but many commercial platforms now rely on alternative metrics derived from power output or proprietary recovery algorithms.

Training Stress Score (TSS)

Training Stress Score is one of the most widely recognized workload metrics in endurance sports. It was developed within the TrainingPeaks ecosystem to estimate the physiological stress of a workout by combining duration and intensity into a single numerical value.

A simplified representation of the concept is:

TSS=Duration×NP×IFFTP×3600×100

TSS=FTP×3600Duration×NP×IF×100

Where:

• Duration = workout duration in seconds

• Normalized Power (NP) = adjusted power value accounting for variable intensity

• Intensity Factor (IF) = ratio between NP and FTP

• FTP = Functional Threshold Power

The formula essentially measures how demanding a session is relative to the athlete’s threshold capacity.

Interpretation

A value of 100 TSS roughly corresponds to one hour of training at Functional Threshold Power. Because of this scaling, the score is intuitive for coaches.

Typical interpretations may look like:

Easy endurance ride → 40–60 TSS

Moderate training session → 70–120 TSS

Long race simulation → 150–250 TSS

Coaches use these values to compare sessions of different structures and to track accumulated workload over time.

For example, a two-hour endurance ride at moderate intensity might produce around 120 TSS, while a shorter but highly intense interval workout may produce 80–100 TSS. Despite the shorter duration, the higher intensity increases the physiological stress of the session.

Trademark Context

The terms:

• Training Stress Score (TSS®)

• Normalized Power (NP®)

• Intensity Factor (IF®)

are registered trademarks owned by Peaksware, the company behind TrainingPeaks and WKO.

Because of this, most training platforms avoid using these exact terms directly in their products. Instead, they develop their own workload metrics that follow similar principles but use different algorithms and terminology.

How Other Platforms Estimate Training Stress

Although many companies aim to quantify training stress, they often use different physiological signals and modeling approaches.

Garmin — Training Load

Garmin devices estimate training load using Excess Post-Exercise Oxygen Consumption (EPOC), a concept derived from exercise physiology research.

Instead of relying on power data alone, Garmin primarily analyzes:

• Heart rate data

• Training intensity

• Session duration

• Individual physiological characteristics

EPOC reflects how much oxygen the body consumes during recovery after exercise. Higher EPOC values indicate greater physiological stress. Garmin aggregates EPOC values across sessions to produce a 7-day training load estimate.

Although the goal is similar to TSS—quantifying workout stress—the calculation method is fundamentally different because it is based on physiological recovery cost rather than power output.

Strava — Relative Effort

Strava’s Relative Effort metric is based primarily on heart rate data.

It calculates training stress by analyzing:

• Time spent in different heart rate zones

• Duration of the workout

• Individual heart rate thresholds

Higher intensity zones are weighted more heavily, meaning that time spent near maximal effort contributes disproportionately to the score.

This approach is conceptually similar to the TRIMP model, which was originally developed using heart rate data. Unlike TSS, Relative Effort does not require power data and therefore works across many sports.

Whoop — Strain

Whoop calculates a metric called Strain, which estimates cardiovascular load across the entire day.

Instead of focusing solely on workouts, Whoop measures:

• Continuous heart rate data

• Time spent in heart rate zones

• Overall cardiovascular load throughout the day

The Strain score is calculated on a logarithmic scale from 0 to 21, reflecting increasing levels of cardiovascular stress.

Because Whoop continuously tracks physiological data, its model captures not only training sessions but also other stressors such as daily activity and lifestyle factors.

Intervals.icu — eTSS

Intervals.icu uses eTSS (estimated Training Stress Score), which attempts to replicate the concept of TSS using available data.

Depending on the sport and available sensors, eTSS may be calculated using:

• Power data

• Heart rate data

• Pace-based models

This flexibility allows the platform to estimate training stress even when power meters are not available.

Do These Platforms Use the Same Calculation?

Although these metrics aim to measure the same general concept—training stress—they do not use identical calculations.

The main differences include:

Despite these differences, the underlying goal remains consistent: estimating how much physiological stress a training session places on the athlete.

Why This Matters for Coaches

Understanding how these metrics are calculated is important because different platforms may produce different workload values for the same workout.

For example:

• A high-intensity interval session may produce a high TSS due to elevated power output.

• The same session may produce an even higher Strava Relative Effort if heart rate remains elevated.

• Whoop may calculate additional strain if recovery between intervals is incomplete.

For coaches working with athletes using multiple platforms, it is therefore important to focus less on the absolute value of a single score and more on long-term trends within a consistent system.

Ultimately, workload metrics are most useful when they help coaches answer a simple question:

Is the athlete adapting positively to the training stimulus over time?