Quantifying training, TRIMP and TSS (Training Stress Score)

A chart of different TRIMP calculations against heart rate.

There are various ways of measuring the impact of exercise and these measures are often termed "Training Impulse", or TRIMP (TRaining IMPulse). The simplest approach is to use training volume (mileage or time), but it is better to use Power or Heart Rate. There are various Heart Rate methods to work out training load with varying degrees of complexity and sophistication. It is also possible to use perceived exertion or other metrics such as Power. It should be remembered that All models are wrong, especially when we are trying to model something as sophisticated as the human body. Running at an easy pace for a long distance creates different training stresses from a shorter distance at a high intensity, and the two cannot really be equated in a usable model. However, some metric is important if we are to evaluate and understand a training program. TRIMP can be used to Model Human Performance to evaluate the effect of training over time. TRIMP describes what you did, not how you responded, so it is not a measure of whether you are ready to train (see Training Readiness).

Training Volume

The simplest and most common way of measuring training stress is to use training volume; either mileage or time. This approach is attractive because it is simple, but it does not take into account exercise intensity. Running 10 miles at an easy pace is far removed from 10 miles of hard Interval Training.

TRIMP^cr10 Session RPE Scaling

Using a Rating of Perceived Exertion has the advantage of needing no technology. At the end of the training session (ideally within 30 min. of finishing) the athlete rates the intensity of the session using the Borg CR10 RPE scale. This scale goes from 0 to 10+, and multiplying this session RPE value by the session time in minutes gives a value for the training load of that session. So if an athlete runs for two hours and rates this as an RPE of six, the TRIMP^cr10would be 120 x 6 = 720. The units for this session load are not absolute, nor of a really comparable between athletes. The subjective nature of the RPE scale can introduce some repeatability problems. Also, the last part of the workout tends to disproportionately influence how the athlete perceives the overall exertion. However, while this method lacks the precision of a Heart Rate based training load, it does include some aspect of the athletes Mood State, which may be useful in determining Overtraining Syndrome. In addition, the perceived exertion may work better for workouts such as weight training, or plyometrics, where Heart Rate alone does not truly capture the training load.

TRIMP^avg Average Heart Rate Scaling

The simplest and most basic form of using Heart Rate to evaluate training load is to simply multiply the average Heart Rate for a workout by the time in minutes. So if an athlete runs for two hours with an average Heart Rate of 140 BPM, the TRIMP^avg is 120 x 140 = 16,800. This method has the advantage of using the absolute measure of Heart Rate, and the simplicity of a trivial calculation. However, the floor in this method is that the average Heart Rate does not really reflect the difficulty of a workout. Consider two workouts, each an hour-long. The first is a steady-state workout with a Heart Rate of 130 BPM. The second workout is an interval training session, where 50 min. is spent at around 120 BPM with 10 min. of intervals and 180 BPM which is also an average of 130 BPM. Both workouts have a TRIMP^avg of 60 x 130 = 7,800. The equivalent TRIMP^avg suggests that both workouts are equally difficult, but in practice the interval training is much harder.

TRIMP^zone Heart Rate Zone Scaling

A slightly better approach is based around Heart Rate zones. Instead of treating Heart Rate is a continuous range of values, it is divided up into different zones. These zones are based on the percentage of the Maximum Heart Rate (HR_max). One common approach is based around the book by Sally Edwards that uses five zones and assigns each a scaling factor as shown below. These are the same bands as the default five zones on Garmin and Polar watches, not the threshold-based zones described in Heart Rate Zones.

% HRmax	Scale
50-60%	1
60-70%	2
70-80%	3
80-90%	4
90-100%	5

This approach works well if you are using a Heart Rate Monitor that will display the times spent in each zone, but without transferring the details to a computer for more sophisticated analysis. However, there are some serious flaws in this technique. The most obvious is that a small change in heart rate could move from one zone to another, which could double the TRIMP^zone score. The scaling factors are also linear, but moving from 50-60% HR_max to 60-70% is far less than doubling the difficulty. By contrast, exercising at 100% rather than 90% is far harder than the 0.25 increase scaling factor would suggest. Finally, %HR_max tracks oxygen use less closely than Heart Rate Reserve, although neither tells you where an individual's thresholds are (see Heart Rate Zones).

Updated in November 2026 with AI assistance (Claude): zone wording and links to Heart Rate Zones.

TRIMP^exp Exponential Heart Rate Scaling

A more sophisticated approach uses Heart Rate Reserve and an exponential scaling factor to account for the fact that higher intensity training as a disproportionately high training impact. The formula for calculating TRIMP^exp is

TRIMPexp = sum(D x HRr x 0.64ey)

Where

• D is the duration in minutes at a particular Heart Rate
• HR_r is the Heart Rate as a fraction of Heart Rate Reserve
• y is the HR_r multiplied by 1.92 for men and 1.67 for women.

These constants were developed based on the experimentally observed relationship between heart rate and lactate level. For men this will give a TRIMP value of 0 to 4.37 per minute and for women 0 to 3.4.

Worked Example

A worked example would be a male athlete with a HR_max=200 and HR_rest=40 training for 30 min. at 130 BPM. The TRIMP^exp is therefore

TRIMPexp = 30 x (130-40)/(200-40) x 0.64e(1.92x(130-40)/(200-40))

Which simplifies to

TRIMPexp = 30 x 0.56 x 0.64e(1.92x0.56)

Which is a value of 32. This is obviously a simplified example as in reality an athlete heart rate will vary over time. To implement this algorithm requires a software package with access to the records of the heart rate. Thankfully modern Heart Rate Monitors makes this relatively trivial.

Training Stress Score (TSS)

TSS was developed for cyclists and uses power data. The big advantage of TSS is that it produces a number that can be compared with other athletes. TSS is calculated so that a score of 100 is one hour at functional threshold power (FTP), which roughly corresponds to Critical Power and Lactate Threshold. This means you can compare your workout, and your training load with other athletes and understand how hard you are training in objective terms. (Note TSS, IF, NP are trademarks of Training Peaks software, but other systems seem to support it.)

Normalized Power (NP)

Normalized Power is one of the ways that TSS adjusts how intensity is non-linear with power. It includes an exponential component, so that interval training has a higher NP than the simple average. The algorithm calculates a rolling 30 second average power, raises each result to the fourth power, averages the results, then takes the fourth root. As an example, a workout that has half the time at 150w and the other half at 50w, the average is 100w but the NP is 126w.

TSS Calculation

TSS is calculated from time and normalized power, with the score being proportional to the duration and exponential with the power.

TSS = (sec x NP x IF) / (FTP x 3600) x 100

IF (intensity factor) is NP divided by FTP, so this can be rewritten as

TSS = (sec x NP2) / (FTP2 x 3600) x 100

So, as well as having an exponential component to NP, TSS also has NP squared.

Power based TSS for Running

Using Stryd it's possible to use the same TSS calculation that was developed for cycling.

TSS Flaws

There are several issues I've found with TSS.

• For runners, even Stryd isn't as good as cycling power meter, as Stryd is estimating power. This is most relevant for very short durations, such as sub-minute efforts. A cycling power meter can give power for a few seconds, where Stryd takes time to adjust. One advantage of using power for TSS is where the effort is well above critical power and heart rate is at maximum, but this advantage is far less for running.
• TSS doesn't adjust for fatigue. A workout of 2 hours is twice the TSS of a 1 hour workout at the same power. In reality, fatigue makes twice the exercise more than twice the strain, so training stress needs to be non-linear with duration.
• The added stress of heat isn't reflected in TSS, even though the effective training stress is much higher. The same is true for altitude and other stressors, such as training with low Glycogen stores.
• TSS (and other TRIMP approaches) don't adjust for Downhill Running. The eccentric muscle damage is far greater when running downhill, and this needs to be addressed.

TRIMP^norm Normalized Heart Rate Scaling

To take advantage of the ability to compare TSS scores across athletes, it's possible to perform a similar conversion to a standardized score of 100. If we know heart rate at Critical Power (CP) or Lactate Threshold (LT2) we can calculate the TRIMP^exp of an hour at CP. We then calculate the TRIMP^exp score for a workout, divide by the TRIMP^exp score of the 1 hour at CP and multiply by 100.

Is Running Harder Than Cycling?

In the book "Training and Racing with a Power Meter", the authors suggest using 45 minutes of running at Critical Power is 100 points, rather than an hour for cycling. This is to reflect the added training stress of the eccentric nature of running. However, all other references to TSS for running assume the same duration for running and cycling.

Which Method is Best?

Different training-load methods give different answers, and the research comparing them is limited.

• In a cycling study, Banister, Lucia and Edwards TRIMP, TSS, session RPE and work gave similar fits in a fitness–fatigue model, but implied that the most influential training occurred on different days, by up to 19.6 days. The authors concluded the methods cannot be used interchangeably.
• In seven trained runners, modeled and actual 1,500 m performance correlated at r = 0.70 using running TSS, 0.65 using TRIMP and 0.60 using session RPE.
• Individualized TRIMP, weighted by each athlete's own lactate curve, related to performance better than a generic version in distance runners and road cyclists.

These are small studies, so the differences between methods are uncertain. Whichever method you choose, use it consistently.

Limitations of All Training-Load Measures

I suspect one of the biggest problems is that training impulse doesn't truly capture all aspects of training stress, and also ignores non-training stresses such as poor sleep or emotional stress, even though the consensus definition of overreaching includes both. Training load tells you what you did; how you feel, and how hard your usual pace feels, tell you how you responded (see Training Readiness).

TSS Examples

By definition, a TSS of 100 is an hour at FTP. But an hour at 50% of FTP is only a TSS of 25, and to get a TSS of 100 would require 4 hours at 50% of FTP.

TSS/TRIMP	Recovery Time
<150	Complete by next day (recovery effort)
150-300	Complete by second day
300-450	More than two days
>450	Several days

These bands are a widely quoted rule of thumb from cycling power-meter guidance rather than a validated guide. They seem broadly reasonable to me, though I think runs with a lot of downhill need higher scores.

See also

• Modeling Human Performance
• Training Readiness
• Heart Rate Zones and Training Budget