Training Load Metrics Decoded: TSS, TRIMP, CTL/ATL/TSB

Training is fundamentally a dose-response phenomenon: apply the right amount of physiological stress and the body adapts — becoming fitter, faster, and more resilient. Apply too little and you stagnate. Apply too much, or too much too quickly, and you break down into injury, illness, or overtraining syndrome. The challenge is that the line between 'productive overload' and 'destructive overload' is invisible without measurement. Subjective feel — 'I feel tired' versus 'I feel good' — is notoriously unreliable as a sole guide because perceived effort is modulated by sleep, nutrition, stress, motivation, weather, and dozens of other confounders that have nothing to do with actual training load on the musculoskeletal and cardiovascular systems.

The foundational framework for understanding training load comes from Eric Banister's impulse-response model, published in 1975. Banister proposed that every training session produces two simultaneous but opposing effects: a positive 'fitness' impulse that decays slowly over approximately 42 days, and a negative 'fatigue' impulse that decays more rapidly over approximately 7 days. Performance at any given moment is the algebraic difference between accumulated fitness and accumulated fatigue. This elegant model explains why a taper works (fatigue dissipates faster than fitness, leaving a net performance gain), why overreaching can be productive (short-term fatigue masks underlying fitness that emerges after recovery), and why chronic overtraining is catastrophic (fatigue accumulation eventually overwhelms fitness gains).

Modern training load metrics are all descendants of Banister's model, each attempting to quantify the 'dose' of training stress more precisely. The proliferation of metrics — TRIMP, TSS, rTSS, CTL, ATL, TSB, EPOC, Training Effect, Relative Effort, Training Load, Exercise Load — can overwhelm runners, but the underlying logic is consistent: measure how hard and how long you trained, accumulate that load over time, and use the ratio of recent load to long-term load to assess your current state. The goal of this article is to decode each metric, explain its calculation, identify its strengths and limitations, and help you build a practical monitoring system that works for your training level and available tools.

Why can't you just track weekly mileage? Volume alone misses intensity entirely — 50 miles of easy running produces vastly different physiological stress than 50 miles including two hard interval sessions and a tempo run. Conversely, purely intensity-based metrics miss the cumulative mechanical load of high mileage. The ideal training load metric captures both dimensions: the volume (how long) and the intensity (how hard) of each session, weighted appropriately to reflect the actual physiological cost. Getting this right is the difference between a training plan that builds you up and one that tears you down.

TRIMP (Training Impulse) was the first formal method for quantifying internal training load, developed by Banister and colleagues in the 1970s and refined through the 1980s. The original Banister's TRIMP calculates load as the product of exercise duration (in minutes), mean heart rate intensity (expressed as a fraction of heart rate reserve), and an exponential weighting factor that accounts for the disproportionate stress of higher intensities. The formula is: TRIMP = duration (min) x HRratio x 0.64e^(1.92 x HRratio) for males, where HRratio = (exercise HR - resting HR) / (max HR - resting HR). The exponential weighting means that time spent at high intensities contributes exponentially more load than equivalent time at low intensities — reflecting the reality that 10 minutes at lactate threshold is far more stressful than 10 minutes of easy jogging.

Edwards' TRIMP (1993) simplified the approach by dividing heart rate into five zones and assigning integer weights: Zone 1 x 1, Zone 2 x 2, Zone 3 x 3, Zone 4 x 4, Zone 5 x 5. Total load equals the sum of minutes in each zone multiplied by its weight. This method is computationally simple, requires only zone boundaries, and correlates well with perceived exertion and blood lactate responses in recreational athletes. Its weakness is the arbitrary, linear zone weighting — the physiological cost difference between Zone 4 and Zone 5 is likely much greater than between Zone 1 and Zone 2, but Edwards' TRIMP treats each zone increment equally.

Lucia's TRIMP (2003) addressed this limitation by collapsing the five-zone model into three zones based on ventilatory thresholds: Zone 1 (below VT1, weight = 1), Zone 2 (between VT1 and VT2, weight = 2), and Zone 3 (above VT2, weight = 3). This three-zone model aligns with Seiler's intensity distribution research and the physiological reality that the metabolic cost of exercise increases non-linearly across threshold boundaries. Lucia's approach is particularly popular in research contexts because it maps directly to lactate threshold and VO2 max testing data, making it practical for laboratory-based athlete monitoring.

Each TRIMP variant has its place. Banister's exponential TRIMP is the most physiologically accurate but requires continuous heart rate data and accurate max/resting HR values. Edwards' zone-based TRIMP is the simplest to calculate and is built into many consumer platforms (Strava's Relative Effort is essentially a modified Edwards' TRIMP). Lucia's three-zone TRIMP is the gold standard in research but requires threshold testing for accurate zone placement. For most recreational runners, Edwards' or a similar zone-weighted approach provides a good balance of accuracy and practicality.

TRIMP Methods Compared

Training Stress Score (TSS) was developed by Andrew Coggan and Hunter Allen for use with cycling power meters and later adapted for running. TSS quantifies the total training load of a session relative to your threshold capacity: TSS = (duration in seconds x NP x IF) / (FTP x 3600) x 100, where NP is Normalized Power (a weighted average that accounts for variability in effort), IF is Intensity Factor (NP divided by FTP), and FTP is Functional Threshold Power (the highest power you can sustain for approximately one hour). A TSS of 100 corresponds to one hour of riding or running at exactly your threshold — the maximum sustainable steady-state effort.

Intensity Factor (IF) is the key modifier that distinguishes easy sessions from hard ones. An IF of 0.55-0.75 represents easy/recovery effort, 0.75-0.85 represents tempo work, 0.85-0.95 represents threshold intervals, 0.95-1.05 represents VO2 max work, and above 1.05 represents anaerobic capacity efforts (only sustainable for short durations). IF allows direct comparison of workout intensity regardless of duration — a 30-minute tempo session at IF 0.85 is producing fundamentally different physiological stress than a 30-minute recovery jog at IF 0.65, and TSS captures this difference.

For runners without power meters, rTSS (running TSS) or hrTSS (heart rate-based TSS) provides analogous calculations using heart rate or pace as intensity proxies. TrainingPeaks calculates hrTSS from the time spent in each heart rate zone, weighted by the physiological cost of each zone (similar to an exponential TRIMP but normalized to your lactate threshold heart rate). Pace-based rTSS uses running speed relative to threshold pace, analogous to how cycling TSS uses power relative to FTP. The emergence of running power meters (Stryd, Garmin Running Power, COROS) has made true power-based TSS increasingly accessible for runners, offering the advantage of capturing terrain-independent effort (unlike pace, which varies with hills and wind).

TSS values provide a common currency for comparing the cost of different workouts and planning weekly training loads. The table below provides guidelines for interpreting TSS values in terms of recovery requirements and training impact. These are approximate — individual recovery capacity varies significantly with fitness level, age, sleep quality, and overall life stress.

TSS Guidelines for Runners

The Performance Management Chart (PMC) is the visual representation of Banister's impulse-response model applied to daily training data, and it is arguably the single most powerful tool in quantitative training analysis. It tracks three metrics over time: CTL (Chronic Training Load), ATL (Acute Training Load), and TSB (Training Stress Balance). CTL is the exponentially weighted moving average of daily TSS over approximately 42 days — it represents your 'fitness' in Banister's model, reflecting the cumulative training load your body has adapted to. CTL rises slowly with consistent training and decays slowly with detraining, with a time constant of approximately 42 days (meaning it takes about 6 weeks for a sustained change in training load to be fully reflected in CTL).

ATL is the exponentially weighted moving average of daily TSS over approximately 7 days — it represents 'fatigue' in Banister's model, capturing the acute training stress from your most recent week. ATL responds rapidly to training spikes and recovers quickly during rest days, with a time constant of 7 days. A hard training week will spike ATL well above CTL, while a recovery week will cause ATL to drop below CTL. The ratio of these two values drives the third and most actionable metric: TSB = CTL - ATL. When ATL exceeds CTL (you're training harder than your chronic average), TSB goes negative — you're accumulating fatigue. When ATL drops below CTL (you're training less than usual, as during a taper), TSB goes positive — fatigue is dissipating while fitness is largely preserved.

The practical application of TSB for race planning is well-established in coaching practice. The consensus target for race day TSB is between -10 and +25. Negative TSB (down to about -10) means you still have residual sharpness from recent training without excessive fatigue. Highly positive TSB (above +25-30) suggests you may have tapered too aggressively and lost some of the neuromuscular activation and metabolic efficiency that recent training maintains. During heavy training blocks, TSB of -20 to -40 is common and expected — this represents productive overreaching that will convert to fitness once you recover. TSB below -40 for extended periods (more than 2-3 weeks) is a warning sign for non-functional overreaching or overtraining.

The PMC's strength is its simplicity and visual clarity — at a glance, you can see whether your fitness (CTL) is trending up, whether you're carrying excessive fatigue (deeply negative TSB), and whether your taper timing is appropriate. Its limitation is that it treats all training stress as equivalent: a TSS of 100 from an easy long run is weighted identically to a TSS of 100 from threshold intervals, despite producing different physiological adaptations and fatigue signatures. Advanced coaches supplement the PMC with separate tracking of intensity distribution (polarized vs threshold), mechanical load (total distance), and subjective markers (RPE, sleep, mood) to capture the dimensions that CTL/ATL/TSB alone miss.

PMC Metrics Explained

Garmin's training metrics are powered by Firstbeat Analytics (acquired by Garmin in 2020), which uses EPOC (Excess Post-Exercise Oxygen Consumption) as its primary internal metric rather than traditional TRIMP or TSS. EPOC estimates the total oxygen debt created by a workout — a measure of the metabolic perturbation that must be resolved during recovery. Firstbeat's algorithms translate EPOC into several consumer-facing metrics: Training Effect (aerobic and anaerobic, scored 0-5), Training Load (cumulative 7-day EPOC in arbitrary units), Training Status, and Training Readiness.

Training Status is Garmin's headline metric, displaying one of several states: Productive (fitness improving, optimal load), Maintaining (fitness stable, steady load), Recovery (short-term recovery after hard training), Unproductive (training load high but fitness declining — often indicates insufficient recovery), Detraining (training load below maintenance threshold, fitness declining), Overreaching (very high short-term load relative to fitness — unsustainable if prolonged), and Peaking (reduced load allowing fitness peak). These states are derived from the relationship between 7-day load, VO2 max trends, HRV status, and recovery time estimates. The algorithm requires approximately 2-3 weeks of consistent data with an optical or chest HR monitor to establish reliable baselines.

Training Readiness (0-100 scale, available on newer Garmin devices) integrates multiple overnight and morning metrics: HRV status, sleep quality and duration, recovery time remaining, acute training load, stress level, and sleep-derived body battery status. A score above 60 suggests readiness for a moderate-to-hard workout, while scores below 30 indicate significant fatigue and suggest easy activity or rest. Garmin's Body Battery (0-100) provides a simpler version of this concept, estimating energy reserves from HRV, stress, activity, and sleep data — rising during rest and declining during activity and stress.

The key limitation of Garmin's system is its opacity — the proprietary algorithms make it impossible to verify or customize calculations. Garmin's VO2 max estimates can be inaccurate by 5-15% depending on running conditions (heat, hills, treadmill), and Training Status can give misleading readings during periods of non-running stress (illness, travel, life events) that affect HRV and resting HR without reflecting actual training load changes. The best approach is to treat Garmin metrics as one input among several, not as an absolute oracle. If Garmin says 'Unproductive' but you feel well-recovered, are sleeping well, and your resting HR is stable, the algorithm may simply be reacting to a confounding variable.

Strava's Relative Effort is essentially a zone-weighted TRIMP calculation normalized to each user's heart rate zones. Time spent in higher heart rate zones contributes exponentially more to the Relative Effort score than time in lower zones, similar to Edwards' TRIMP but with proprietary weighting. A typical easy run might score 30-80, a tempo run 100-180, and a marathon race effort 250-400+. Strava calculates your heart rate zones automatically from your estimated max HR (derived from your hardest recorded efforts), though you can manually set custom zones in your profile for more accurate scoring.

Strava Fitness (available to Summit/subscription members) is the cumulative rolling average of Relative Effort — functionally equivalent to CTL in the TrainingPeaks PMC. Strava also calculates Fatigue (equivalent to ATL) and Freshness (equivalent to TSB). The Relative Fitness chart provides the same visual framework as the PMC, showing fitness trends, fatigue accumulation, and form over time. Because Strava uses HR-based input rather than power-based TSS, the absolute numbers differ from TrainingPeaks, but the trends and patterns should be directionally similar.

One significant advantage of Strava's system is its massive user base and social integration — you can compare your Relative Effort with training partners and see how your load compares to others doing similar workouts. However, this is also a limitation: because heart rate response varies enormously between individuals (a well-trained runner with a low max HR will generate different zone distributions than a less-fit runner at the same effort), Relative Effort scores are not directly comparable between users. They are meaningful only for tracking your own trends over time.

Strava's system is particularly unreliable for activities without heart rate data (it estimates Relative Effort from pace alone, which ignores terrain, wind, and fatigue) and for activities that begin with a warm-up already in progress (delayed HR monitor pairing creates artificially low scores for the early portion). For runners who train primarily by heart rate and use Strava as their main platform, Relative Effort provides a practical and free (for subscribers) training load tracking system. For more precise analysis, platforms like TrainingPeaks, WKO, or Intervals.icu offer greater customization and additional metrics.

COROS uses its proprietary EvoLab platform to calculate training load and fitness metrics. COROS Training Load is a 7-day cumulative score based on duration and intensity, categorized as Low, Optimal, High, or Overload. The unique contribution of COROS is its Running Fitness metric — a threshold pace estimate derived from workout data that serves as a performance proxy similar to a running FTP. COROS also tracks Base Fitness (long-term training load trend, analogous to CTL) and provides Training Status labels (Optimal, Overreaching, Detraining, etc.) based on the relationship between short-term and long-term load.

COROS's advantage is its integration with the Stryd running power meter ecosystem and its native running power estimation (available on COROS PACE 3, VERTIX 2S, and later models). Running power enables more accurate load calculations than heart rate alone because power responds instantaneously to effort changes (no cardiac lag), is unaffected by cardiac drift or temperature-induced HR elevation, and captures the actual mechanical work of running including hill and wind resistance. COROS's Marathon Level metric uses accumulated training data to predict marathon finishing time, providing a goal-oriented benchmark that contextualizes training load in terms of race readiness.

Apple Watch introduced Exercise Load with watchOS 11, dividing training stress into separate cardio and strength load channels. The cardio load metric tracks running and cycling stress similarly to TRIMP-based systems, showing 28-day trends categorized as Well Below, Below, Steady, or Above your recent average. Apple's approach is uniquely positioned for general health users because it captures load from all activity types (not just structured running) and integrates with the broader Apple Health ecosystem including sleep, HRV, and environmental data.

The persistent challenge across all watch ecosystems is that VO2 max estimates frequently disagree between devices — sometimes by 5-10 ml/kg/min. This occurs because each manufacturer uses different algorithms, different heart rate sensors (optical quality varies), and different assumptions about running efficiency. Garmin tends to estimate higher (optimistic) VO2 max values, COROS tracks closer to lab values for many runners, and Apple Watch tends to be conservative. The absolute number matters less than the trend: if your watch-estimated VO2 max is trending upward over months, your aerobic fitness is likely improving regardless of whether the number matches a laboratory test. Use the same device consistently and track the direction, not the decimal point.

The Acute:Chronic Workload Ratio (ACWR) was popularized by Tim Gabbett in a landmark 2016 paper in the British Journal of Sports Medicine that analyzed injury rates across multiple team sports. The concept is straightforward: divide the acute workload (typically the current week's training load) by the chronic workload (the rolling average of the previous 4 weeks' training load). This ratio expresses how much your recent training deviates from what your body is accustomed to. Gabbett's data showed a clear U-shaped relationship between ACWR and injury risk: both very low ratios (undertrained athletes exposed to sudden demands) and very high ratios (rapid training load spikes) were associated with elevated injury rates.

The 'sweet spot' identified by Gabbett falls between an ACWR of 0.8 and 1.3. Within this range, you're training at or moderately above your recent chronic load — sufficient for progressive overload but not so far above baseline as to exceed tissue tolerance. The danger zone begins above 1.5, where the acute week's load exceeds the 4-week average by 50% or more. In Gabbett's cricket and rugby data, athletes with ACWR above 1.5 had 2-4 times the injury risk compared to those in the sweet spot. Below 0.8, the athlete may be losing fitness through undertraining or returning from injury with insufficient preparation for the demands of competition.

Two methods exist for calculating ACWR: the rolling average approach (simple arithmetic mean of the last 4 weeks for chronic load) and the exponentially weighted moving average (EWMA) approach, which weights recent weeks more heavily. Williams et al. (2017) compared both methods and found that EWMA was more sensitive to rapid training load changes and showed stronger associations with injury risk, because it better captures the 'recency' of training — a concept that aligns with the physiological reality that the body responds most acutely to the most recent stress stimulus.

However, ACWR is not without critics. Impellizzeri et al. (2020) published a rigorous critique arguing that the mathematical coupling between the acute and chronic components (the acute week is included in the 4-week chronic average) creates a statistical artifact that inflates the apparent association with injury. They demonstrated that even randomly generated data can show the same U-shaped ACWR-injury pattern. This does not invalidate the concept of avoiding rapid load spikes — the physiological rationale remains sound — but it suggests that ACWR should be used as a general guideline rather than a precise predictive tool. Monitoring week-over-week load changes (the '10% rule' for volume, adjusted for intensity) may be equally effective and easier to implement.

ACWR Injury Risk Framework

For recreational runners logging 20-40 miles per week, the most important metrics are weekly volume (miles or kilometers), intensity distribution (percentage of running in each heart rate zone), and a subjective wellbeing score (1-10 for sleep, energy, motivation, and muscle soreness). These four dimensions capture the vast majority of what matters for health, improvement, and injury prevention. Adding a TRIMP-based load metric from your watch (Garmin Training Load, Strava Relative Effort, or COROS Training Load) provides an integrated single number that can flag when you're deviating from your normal training pattern — particularly useful for catching the gradual load creep that precedes many overuse injuries.

For competitive runners targeting specific race performances, the PMC framework (CTL/ATL/TSB) becomes valuable for periodization and taper planning. Track CTL over 12-16 week training cycles and observe how your fitness number correlates with race performances and workout quality. Many coaches aim for a CTL peak approximately 2-3 weeks before goal race, followed by a 10-14 day taper that brings TSB from its training nadir (typically -20 to -35) up to the target range of -10 to +15. The exact numbers are highly individual — some runners race best with TSB near zero, others perform well at -15 — so track your own patterns across multiple race cycles.

The most reliable warning signs of impending overtraining are often visible in the gap between objective and subjective metrics. When your training load metrics say you should be fresh (positive TSB, low ATL) but your subjective feel is poor (elevated morning HR, disrupted sleep, low motivation, heavy legs), something external is adding stress that the training metrics don't capture — work pressure, relationship stress, inadequate nutrition, or incipient illness. Conversely, when your metrics show high fatigue (deeply negative TSB) but you feel sharp and eager to train, you may be riding the productive overreaching wave that often precedes a breakthrough — provided you have a recovery period planned.

The single biggest mistake runners make with training load data is treating the numbers as absolute truth rather than as one input among many. A Garmin Training Status of 'Unproductive' is a prompt to investigate — not an automatic command to add rest days. Check your sleep, nutrition, life stress, and recent illness history before making training decisions based solely on algorithm output. The best monitoring systems combine objective data (HR-based load, weekly volume, HRV trends) with subjective markers (RPE, sleep rating, motivation, muscle soreness) and physiological signals (resting heart rate trend, morning HRV, body weight stability) into a holistic picture that no single metric can provide.

Building an effective training load monitoring system doesn't require expensive gadgets or complex spreadsheets — it requires consistency. The minimum viable tracking system consists of three elements: session RPE (how hard did that run feel on a 1-10 scale, recorded within 30 minutes of finishing), weekly mileage (total distance), and a daily sleep quality rating (1-5). These three data points, recorded consistently in a simple spreadsheet or training log, will catch the vast majority of load management errors: sudden volume spikes, accumulated fatigue from high-intensity blocks, and the sleep disruption that often precedes overtraining. Foster (1998) demonstrated that session RPE-based training load (duration x RPE) correlates strongly (r = 0.75-0.90) with heart rate-based TRIMP methods, validating subjective load tracking as a legitimate alternative when HR data is unavailable.

At the intermediate level, add HR-based training load from your watch platform (Garmin, Strava, or COROS), track weekly CTL/ATL trends (available in TrainingPeaks, Strava Fitness, or the free Intervals.icu platform), and monitor morning resting heart rate (either manually or via overnight wearable measurement). The addition of HR-based load provides intensity-weighted data that catches what mileage alone misses: two 50-mile weeks may produce vastly different training loads depending on how much of that mileage was above threshold. CTL tracking gives you a fitness trendline that helps with taper planning and detraining monitoring. Morning resting heart rate is one of the earliest and most reliable indicators of accumulated fatigue — a sustained elevation of 5+ bpm above baseline warrants attention.

Advanced monitoring adds running power-based TSS (via Stryd or COROS/Garmin running power), daily HRV measurements (WHOOP, Garmin, Oura, or HRV4Training app), and a readiness score that integrates multiple inputs. Power-based TSS is the most precise load metric available for running because it captures terrain and environmental conditions that heart rate cannot. HRV provides the most sensitive window into autonomic nervous system recovery — a trending decline in resting HRV over 1-2 weeks, particularly when combined with elevated resting HR, is one of the earliest detectable signs of overreaching. The cost of this level of monitoring is time (5-10 minutes daily for data review) and device investment, but for runners training 50+ miles per week or targeting competitive performances, the injury prevention and performance optimization ROI is substantial.

Regardless of your monitoring level, the fundamental principle is the same: track consistently, watch for trends, and act early when the data suggests a mismatch between load and recovery. A simple traffic light system works well: Green (metrics stable or improving, subjective feel good, sleep quality high) = continue as planned. Amber (one or two metrics trending unfavorably, slight fatigue, or minor sleep disruption) = reduce intensity for 2-3 days, prioritize sleep, and reassess. Red (multiple metrics declining, subjective feel poor, disrupted sleep, elevated RHR) = take 2-3 full rest days before resuming easy running. This framework works equally well whether your 'metrics' are a sophisticated HRV-guided CTL/ATL dashboard or a simple notebook with RPE scores and weekly mileage.

Training Load Monitoring by Level