How Titan measures training load

Train at a steady 150 load points a day for two months, then double it to 300 for one week, and Titan's load ratio climbs from 1.00 to 1.41. Short-term load rises to about 245. Long-term load rises only to about 173. Take the same week off instead and the ratio falls to 0.43. Those numbers come from the training load help article, and every step that produces them traces back to a systems model Eric Banister and his colleagues published in 1975.

What follows traces that line to the load chart in Titan, then computes a full week by hand, including sessions with no heart rate. The performance physiology guide covers the oxygen delivery and threshold physiology that these load numbers stand in for.

01What training load is measuring

Internal load and external load

Impellizzeri, Marcora and Coutts split training load into two measurable parts in 2003 and restated the framework in 2019. External load is the work you do, measured in kilometers, watts, sets and repetitions. Internal load is your body's response to that work, measured with heart rate, blood lactate or perceived exertion. A 10 kilometer run at the same pace is the same external load on a cool morning and on a hot afternoon after a short night. Internal load differs, and internal load drives adaptation.

The gap between the two is informative. When internal load rises at the same external load, something has changed, and fatigue, illness, heat or detraining are the usual causes. When internal load falls at the same pace or power, fitness is improving. Heart rate TRIMP measures internal load, power-based scores blend the two, and weekly distance measures external load only.

When critics argued that "load" belongs to mechanics, Impellizzeri and colleagues answered in 2022 that training load is a broad concept whose components can be measured. On a wearable, that construct gets measured through one signal, usually heart rate, and the number is only as good as that signal and the weights applied to it.

Why duration alone hides how hard a session was

Sixty minutes of easy spinning and sixty minutes of hill repeats are the same duration. Every heart rate load method fixes this by weighting each minute by how hard the heart worked. The methods differ in how steeply they weight hard minutes against easy ones.

02The Banister lineage behind every load number on a wrist

The 1975 and 1976 systems models

Banister, Calvert, Savage and Bach described a systems model of training for athletic performance in the Australian Journal of Sports Medicine in 1975. Calvert and the same group published the indexed version in IEEE Transactions on Systems, Man, and Cybernetics in 1976. The idea came from control engineering. Each session is an input impulse to the athlete, and performance is the output. The input was the training impulse, TRIMP.

The exponential formula and its male and female constants

TRIMP multiplies session duration by the fraction of heart rate reserve used, then applies a weighting factor that grows exponentially with intensity.

ΔHR = (exercise HR − resting HR) / (max HR − resting HR)

TRIMP = duration in minutes × ΔHR × a × e^(b × ΔHR)

The constants that set the curve come from Banister's 1991 chapter on modeling elite athletic performance. For men, a is 0.64 and b is 1.92. For women, a is 0.86 and b is 1.67. Some published calculators apply 0.64 to both sexes, which understates female load.

A man with a resting rate of 50 and a max of 190 who runs 60 minutes at 155 bpm scores about 122. A woman with the same numbers scores about 135, since her weighting at that intensity is about 3.01 against his 2.70. The same run at 130 bpm scores about 66 for the man. Twenty-five extra beats nearly double the load, because the weight on each minute rises faster than heart rate itself.

Training stress score does the same job with power

Andrew Coggan's training stress score, TSS, applies the same idea to a power meter. Normalized power adjusts for variability in output, and the intensity factor is normalized power divided by functional threshold power, FTP. One hour at FTP scores 100 by definition. A 90-minute ride at a normalized 220 W for a rider with a 260 W FTP has an intensity factor of 0.846 and scores about 107. TSS reflects mechanical work relative to threshold, and TRIMP reflects cardiovascular strain. For cyclists with both, TSS is less affected by drift, caffeine and heat. For runners and swimmers without power, heart rate load is the practical choice.

Morton's 1990 equation turned the model into a forecast

Morton, Fitz-Clarke and Banister formalized the model in the Journal of Applied Physiology in 1990. Each impulse produces two responses, fitness and fatigue, and both decay exponentially without further training, fatigue faster than fitness. Predicted performance is a baseline plus a scaled fitness term minus a scaled fatigue term. The authors fitted the model to athletes and non-athletes through training and tapering and found significant correlation between predicted and measured performance.

A hard week raises both responses. Fatigue clears faster, so a lighter week lets performance rise while fitness holds. That is the mechanism of a taper, and the model behind the two lines on most load charts.

Where the model breaks down

The fitness and fatigue model describes groups well and predicts individuals poorly. Hellard and colleagues tested its parameters in 2006 and found them unstable. The fitted fitness and fatigue decay constants correlated at 0.99 with each other, and the two magnitude factors at 0.91. Coefficients of variation exceeded 30 percent, and the bootstrap range for time to peak performance after training stopped ran from 25 to 61 days. The authors concluded that using the fitted parameters to steer training was hazardous, even though the model's performance predictions looked stable.

Busso showed in 2003 that the relationship is also nonlinear. Letting fatigue grow with the amount of training fit six subjects over a 15 week program better than the fixed models and produced an inverted U between daily training and performance. Use the model to describe recent training. Do not trust it to predict race day to the week.

03Zone weighted load is TRIMP's simpler cousin

How Titan weights a minute

Titan uses a simpler weighting than Banister's curve. It sorts every heart rate sample recorded inside a workout into one of five heart rate zones, then multiplies each minute by its zone number. A Zone 1 minute counts 1, a Zone 2 minute counts 2, and a Zone 5 minute counts 5. Thirty minutes in Zone 2 plus ten in Zone 4 is a daily load of 100. The heart rate zones article sets the default zones as percentages of max heart rate, with Zone 1 from 53 to 73 percent, Zone 2 from 74 to 80, Zone 3 from 81 to 86, Zone 4 from 87 to 92, and Zone 5 from 93 to 100. Only heart rate inside a workout counts. A racing pulse in a stressful meeting adds nothing.

How the integer weights compare with the curve

How far a linear 1 to 5 scale drifts from Banister's curve depends on resting heart rate, because Banister uses heart rate reserve and Titan's zones use percent of max. The table below computes the male Banister weight per minute at the midpoint of each Titan zone for a max heart rate of 190, then expresses it relative to Zone 1.

For an athlete with a resting heart rate of 50, Titan weights hard minutes slightly more steeply than Banister's curve, 4 against 3.44 in Zone 4. For a resting heart rate of 65 it weights them slightly less steeply. For resting rates from 40 to 65, every Titan weight falls inside the spread of Banister's. A Titan Zone 4 minute counts exactly twice a Zone 2 minute. Banister's formula puts that ratio between 1.66 and 1.86 across the three resting rates above.

What a five-zone weight gives up

The integer scale buys clarity. You can check any day's load with a pencil, and the same number feeds the Exertion score. It loses distinctions between effort levels in three places.

First, every heart rate within a zone gets the same weight. With a max of 190, a minute at 164 bpm and a minute at 175 bpm both count 4. Second, the boundaries are steps. A steady run at 152 bpm sits at the top of Zone 2 and scores 2 a minute. At 153 bpm it crosses into Zone 3 and scores 3, a 50 percent jump for one beat. Third, everything below 74 percent of max lands in Zone 1, so a 95 bpm walking warm-up inside a workout counts the same as 135 bpm easy running. At a resting heart rate of 50, Banister's curve would weight that warm-up minute at less than half of a Zone 1 midpoint minute.

Across a week these errors mostly average out. They matter when two sessions straddle a zone boundary, and they are the reason to keep your max heart rate accurate, since a wrong max shifts every boundary.

04From one workout to a fitness and fatigue trend

Short-term and long-term load

Titan turns daily load into two exponentially weighted averages. Short-term load uses a 7 day time constant, so each day it moves about 13 percent of the way from yesterday's value toward today's load. Long-term load uses a 42 day time constant and moves about 2.4 percent a day. Short-term load stands for fatigue and long-term load for fitness, Morton's two responses, on the 7 and 42 day constants most coaching software uses.

Coaching software calls the same two averages acute training load, ATL, and chronic training load, CTL. Their difference, often called training stress balance, goes negative during hard blocks and positive during tapers. An athlete who has trained at a steady 80 units a day has both averages near 80. Five days at 120 lift acute load to about 100 and chronic load to about 84, a balance of −16. Ten days at 40 after that drop acute load to about 54 while chronic load falls only to about 75, a balance of +21. The athlete has shed most of the fatigue and kept most of the fitness, which is the arithmetic of a taper.

The daily update is one line.

new load = old load + α × (today's load − old load)

For short-term load, α is 1 − e^(−1/7), about 0.133. For long-term load, α is 1 − e^(−1/42), about 0.0235. Days with no data count as rest days with a load of 0, so a week of forgetting your watch reads as a week off. Titan starts the calculation 84 days before the first day on the chart so the chart's starting values already reflect earlier training.

Why an exponentially weighted average beats a flat one

A 7 day rolling sum treats a session from six days ago the same as one from this morning, then drops it entirely on day eight. Real fatigue fades gradually. Williams and colleagues proposed in 2017 that load ratios use exponentially weighted averages for this reason, so that recent sessions carry more weight and old ones fade out without a cliff. Murray and colleagues tested both methods the same year in 59 elite Australian football players over two seasons. The two methods produced significantly different ratios in the moderate, high and very high ranges. Both linked very high ratios above 2.0 to injury, and the exponentially weighted version explained more of the variance in injury.

05A week of training computed end to end

Take an athlete whose short-term and long-term load both sit at 50, the result of weeks at about 350 load points a week. Max heart rate is 190. Here is one week, with daily load built from zone minutes and each day's short-term and long-term load after the update. The last column shows the Exertion workout points for that load, computed as 10 × (1 − e^(−load/100)), before steps are added.

The week totals 508, up 45 percent on the usual 350. Short-term load rose 29 percent, long-term load 7 percent, and the ratio ended at 1.21, inside Titan's Optimal zone. Thursday's rest alone pulled short-term load back by 8.5. The fatigue estimate changes quickly with each day's load, and the fitness estimate changes slowly.

For comparison, the same Saturday run scored with Banister's male formula, at zone midpoints and a resting heart rate of 50, comes to about 146 TRIMP against Titan's 180. The two scales measure the same thing in different units, so a load number from one app never carries over to another.

06Workouts with no heart rate still need a number

Strength, yoga and sauna sessions often record no heart rate. When a day has workouts and none of them has any zone time, Titan multiplies each workout's minutes by a factor for its type. The table comes from the Exertion score article.

A factor of 1.0 counts each endurance minute the same as a Zone 1 minute, so the estimate runs low for anything harder than easy effort. That is why Tuesday's 60 minute strength session above scores 54 and Friday's 40 minutes of yoga scores 26. A hard lifting session without heart rate will read as light cardio.

The estimate applies only when every workout that day lacks heart rate. If the athlete above had lifted on Monday after the run, Monday would still score 85 and the strength session would add nothing, because Monday had zone time. The week would lose 54 points. If you lift and run on the same day and want the lifting counted, record it with heart rate.

07The acute to chronic workload ratio and the fight over its safe zone

Where 0.8 to 1.3 came from

Gabbett's 2016 review in the British Journal of Sports Medicine made the acute to chronic workload ratio famous. His training-injury prevention paradox argued that athletes used to high loads get injured less than athletes on low loads, and that rapid increases in load, more than load itself, explain many soft tissue injuries. The review placed ratios of 0.8 to 1.3 in a lower-risk range and ratios above 1.5 in a danger zone. Windt and Gabbett followed in 2017 with a workload-injury aetiology model in which total load sets exposure, while both total load and changes in load, such as the ratio, drive fitness and fatigue.

Why sports scientists now argue about it

The critique arrived quickly. Lolli and colleagues showed in 2019 that the conventional ratio includes the acute week inside the chronic average, which couples numerator and denominator by sharing data between them and produces spurious correlation. Impellizzeri and colleagues went further in 2020. No study had tried to estimate a causal effect of the ratio on injury, so manipulating it to change injury rates was conjecture. They called it an inaccurate and ambiguous metric, not consistently or unidirectionally related to injury, and concluded there was no evidence supporting its use in training recommendations aimed at reducing injury risk.

The pooled evidence has not settled the question. Qin and colleagues meta-analyzed 22 cohort studies in 2025. Injury incidence was lowest in the 0.8 to 1.3 band, but the 95 percent confidence interval for that band ran from 14 to 94 percent, and heterogeneity across studies was 92.9 percent. Most of the cohorts were soccer players. The authors called the ratio associated with injury risk and asked for caution in using it.

Same week, three ratios

The athlete's week above shows why the method matters as much as the threshold. Assume the three weeks before it were 350 each.

One week gives 1.21, 1.30 or 1.45 depending on the arithmetic, which puts it comfortably in range, at the edge of Gabbett's band, or near his danger zone. A threshold published for one method does not transfer to another.

Titan's four ratio zones

Titan sorts its own ratio into four zones. They apply only to Titan's exponentially weighted ratio. Under the chart, Trends prints the latest ratio with its zone, such as "Ratio 1.12 · Optimal," with Optimal in green and the other three zones in red.

Read the zones as a prompt to look at recent training. No ratio forecasts injury. A Low ratio is expected during a deliberate taper. What the ratio catches well is the unplanned spike, such as the week you came back from illness and ran your old mileage. For the day-to-day response, the Exertion target already lowers today's range by 0.35 when your last three days averaged more than 0.5 above your goal's range, and shifts it again for Recovery.

08Reading load against your own history

A band built from your own quartiles

Behind the two lines on the training load chart, Titan shades the middle 50 percent of your daily long-term loads over the selected period, using interpolated quartiles. It needs at least 28 daily values covering at least 80 percent of the period, which on the 60 day view means 48 days. A long-term line above the band means you carry more load than for most of that period. A line sinking below it for weeks outside a planned break is the start of detraining. The band describes your own history, which is the only fair comparison when load units depend on each person's zones.

A weekly check

A cyclist with a 270 W FTP training ten hours a week might run this check. On the rest day, review last week's total load, both averages and the ratio. A slightly negative balance and a ratio near 1.1 to 1.2 is normal in a build week. Mid-week, compare the key session's load with the plan. Late in the week, decide whether the weekend long ride stays at full length or gets trimmed based on how easy efforts felt, sleep and morning Recovery. Once a month, look at long-term load over eight weeks. It should rise steadily in build phases and dip in recovery weeks. A flat line despite consistent training points to too little stimulus or too little recovery. A steep climb points to a loading rate the body may not absorb.

Splitting load by intensity

Below the chart, Training load focus splits zone-weighted load for the period into Low aerobic (Zones 1 and 2), High aerobic (Zone 3) and Anaerobic (Zones 4 and 5). The athlete's week above puts 305 points in Low aerobic, 60 in High aerobic and 63 in Anaerobic, so Low aerobic carries most of the load. The 80 points from strength and yoga stay out of the split, because they have no zone time. This view shows whether rising load came from easy volume or training intensity, which the ratio cannot. Volume and intensity rising together for weeks is the pattern to watch for overreaching.

09Training load in Trends

The Trends tile on Today shows your latest load ratio and a sparkline of the last 7 days of daily load. Tap it to open Trends, which starts on the Year period. The period menu at the top also offers 30 Days, 60 Days and 6 Months. The Training load chart comes first. Tap or hold it to read long-term load, short-term load and the load ratio for any day, and read the latest ratio and its zone under the chart. The training load article documents every Titan number in this post. Your zones drive all of it, so check your max heart rate in heart rate zones first.

10References

1. Banister EW et al. (1975). A systems model of training for athletic performance. Australian Journal of Sports Medicine 7. No DOI or PubMed record. The same group's indexed paper is Calvert TW et al. (1976), below.
2. Calvert TW et al. (1976). A systems model of the effects of training on physical performance. IEEE Transactions on Systems, Man, and Cybernetics SMC-6(2):94-102. https://doi.org/10.1109/TSMC.1976.5409179
3. Banister EW (1991). Modeling elite athletic performance. In MacDougall JD, Wenger HA, Green HJ (eds), Physiological Testing of the High-Performance Athlete, 2nd ed. Human Kinetics, 403-424. Book chapter, no DOI or PubMed record.
4. Morton RH et al. (1990). Modeling human performance in running. Journal of Applied Physiology 69(3):1171-1177. https://doi.org/10.1152/jappl.1990.69.3.1171
5. Busso T (2003). Variable dose-response relationship between exercise training and performance. Medicine & Science in Sports & Exercise 35(7):1188-1195. https://doi.org/10.1249/01.MSS.0000074465.13621.37
6. Hellard P et al. (2006). Assessing the limitations of the Banister model in monitoring training. Journal of Sports Sciences 24(5):509-520. https://pmc.ncbi.nlm.nih.gov/articles/PMC1974899/
7. Impellizzeri FM et al. (2019). Internal and external training load, 15 years on. International Journal of Sports Physiology and Performance 14(2):270-273. https://doi.org/10.1123/ijspp.2018-0935
8. Impellizzeri FM et al. (2022). The "training load" construct. Why it is appropriate and scientific. Journal of Science and Medicine in Sport 25(5):445-448. https://doi.org/10.1016/j.jsams.2021.10.013
9. Gabbett TJ (2016). The training-injury prevention paradox. Should athletes be training smarter and harder? British Journal of Sports Medicine 50(5):273-280. https://doi.org/10.1136/bjsports-2015-095788
10. Windt J et al. (2017). How do training and competition workloads relate to injury? The workload-injury aetiology model. British Journal of Sports Medicine 51(5):428-435. https://doi.org/10.1136/bjsports-2016-096040
11. Williams S et al. (2017). Better way to determine the acute:chronic workload ratio? British Journal of Sports Medicine 51(3):209-210. https://doi.org/10.1136/bjsports-2016-096589
12. Murray NB et al. (2017). Calculating acute:chronic workload ratios using exponentially weighted moving averages provides a more sensitive indicator of injury likelihood than rolling averages. British Journal of Sports Medicine 51:749-754. https://doi.org/10.1136/bjsports-2016-097152
13. Lolli L et al. (2019). Mathematical coupling causes spurious correlation within the conventional acute-to-chronic workload ratio calculations. British Journal of Sports Medicine 53:921-922. https://doi.org/10.1136/bjsports-2017-098110
14. Impellizzeri FM et al. (2020). Acute:chronic workload ratio, conceptual issues and fundamental pitfalls. International Journal of Sports Physiology and Performance 15(6):907-913. https://doi.org/10.1123/ijspp.2019-0864
15. Qin W et al. (2025). Acute to chronic workload ratio (ACWR) for predicting sports injury risk, a systematic review and meta-analysis. BMC Sports Science, Medicine and Rehabilitation 17(1):285. https://doi.org/10.1186/s13102-025-01332-x