Every TrainingForever metric: what each one measures and how to read it

By JuanPa, creator of TrainingForever · Reviewed by JuanPa · Updated on 23/10/2026

Short answer: TrainingForever metrics convert each training into a Total Load (CT) and filter it into four curves: Chronic Adaptation (AC), Acute Fatigue (FA), Load Differential (AC) and Performance Potential (PR). They are signs to ask better questions, not verdicts: the coach interprets them in the context of the athlete.

A metric does not replace a conversation, a well-read session, or the feel of the athlete. Yes, he avoids blind training. TrainingForever gathers signals of load, intensity, recent response and observed performance so that coach and athlete can ask themselves better questions: does the week look like what had been planned?, is fatigue rising faster than the base?, is that stable shooting still stable?

This guide explains all the metrics defined today and separates two models that coexist in the product:

• The current daily model is part of the activities carried out and it is the one that feeds Evolution: daily CT, AC, FA, DC, PR, Pace of progression and projection of planned sessions.
• The historical weekly model still exists in some summaries, notices and in the inherited generator. Use hours and unevenness with another formula. Their names look alike, but their figures should not be compared one by one with those of the daily graph.

The distinction matters: two numbers called “chronic adaptation” can answer different questions if they come from different layers. In case of doubt, look at the view and period, not just the acronym.

About the data in this guide: all sample figures come from five example athletes (Sergio Ibáñez, Álvaro Soto, Unai Ferrer, Marta Quintana and Pablo Mena). They are simulated test data, with anchor date 2026-10-19, not real people. These are not clinical recommendations or universal ranges.

How do you read the metric panel in a minute?

To read the TrainingForever metrics panel in one minute, follow these five steps:

1. Start with CT: what load was made every day or week and from what source it comes out (power, pace, HR, RPE or estimate).
2. Compare curves: AC moves slowly; FA, fast. His distance is D.C. PR is another transformation of those two curves, not a brand prediction.
3. Read the direction and context, not an isolated color or cut. Negative DC can be normal under construction; positive can be a taper... or an unwanted pause.
4. Separated from projected. From tomorrow the CT can be an estimate of the plan, not a training that has already occurred.
5. Check the origin and quality. An estimated CT, short history or alternation between HR and RPE should be interpreted more cautiously.

In Sergio (example athlete), the bars add up the daily CT of each seven-day block: it rose from 305 to 435 CT between 1–7 and 8–14 September. The lines are values of the daily model at the end of each block. On September 14 AC 44 and FA 56 are shown; the DC of −14 is not 44−56, but the differential calculated before the load of that day, with the previous values AC 44 and FA 58. The following week the CT dropped to 395, but the FA was still above the AC. In the last week shown the DC reached 0. There is no diagnosis in that sequence: there is a concrete conversation to have about the visible objective, recovery and sensations.

What metrics does the current daily model calculate?

The current daily model of TrainingForever calculates the load of each activity and, from it, the intensity, efficiency, drift and best efforts. These are your metrics, one by one.

What is the Total Load (CT) and how is it calculated?

What it's for and what it stands for. CT is the common unit with which TrainingForever summarizes the load of an activity. One hour at the threshold equals 100 CT per scale construction. Several activities of the same day are added; a day without activity contributes 0. It is modeled load, not calories, muscle damage or a “good training” note. To see how these concepts relate to the chronic, acute and balance load used by other platforms, see the Training load guide (CTL, ATL and TSB).

How it's calculated. The general rule is:

h is the time valid in hours, IR the relative intensity against the existing threshold; and k Depends on the source. The application chooses the first available source: cycling power, running/trail pace, CSS swimming, HR, RPE and, finally, estimation. In race and trail joins 0,1·D+ only if the IR did not incorporate the slope.

How to interpret it. There is no universal “good” CT. Compare it with similar sessions and the history of the same athlete. A change of origin also changes the scale: a session measured by RPE is not identical to the same session measured by HR.

Case. The long run of October 18 by Sergio Ibáñez, an example athlete, accumulated 126 CT with origin ritmo; it was 2:25:07 running. Its figure is better understood along with pace, HR, unevenness and subsequent response, not as an isolated score.

What is relative intensity (IR)?

What it's for and what it stands for. IR places the effort against the personal threshold of the day. It is the bridge between a physical signal and the CT: weighted power/bike FTP, equivalent speed/threat speed in race, speed/CSS in swimming or, in the back of HR, medium/HR threshold.

How to interpret it. IR = 1 means to work at the threshold of the chosen signal; less than 1, below; greater than 1, above. It's not a moral evaluation or a comparison between people. The engine limits improbable values: minimum 0.30 and configurable maximum values of 1.50 in power and 1.30 in pace/CSS, for example.

Case. Álvaro Soto trains cycling with power. Your historical FTP changes apply from your date; so the same power can change from IR when a threshold is updated. In validation with sample data, the 36 post-FTP 251 W activities of 2026-07-07 used that value.

What are weighted power (PP) and Variability (VI)?

What they serve and what they represent. Weighted power (PP) estimates the cost of a cycling output with power changes better than a simple average. Variability (VI) compares PP to average power: describes how much the effort changed, not how hard it was.

The series is shown at 1 Hz, uses a moving average of 30 s and then the fourth power. You need a potentiometer, at least 20 minutes and sufficiently continuous data. Without these conditions there is no data, never zero.

How to interpret them. VI = 1 describes a perfectly constant effort. As a technical orientation, up to 1.05 is typical of very constant efforts; a group exit or with changes can be much greater. PP and VI are only compared between similar efforts and the same cyclist.

Case. Álvaro is the example of a cyclist with the power of the lot of athletes of example. The validation contrasted 427 comparable calculations of cycling weighted power and the calculables matched the independent contrast within tolerance. This verification says that the data was calculated as expected; it does not claim that Álvaro has improved.

How is the difference taken into account when measuring the pace?

What they serve and what they represent. In running and trail, the gross pace doesn't tell the whole story: climbing slowly can cost a lot. Therefore, when there is speed and slope stream, TrainingForever transforms the speed to an equivalent flat speed before obtaining IR. The External Load of the unevenness is not added again in that case.

The polynomial cost of the slope comes from Minetti et al. (2002). If those streams are missing, average speed is used and, in race/trail, CT incorporates 0,1·D+. It's a product convention, not a perfect physiological equivalence.

How to interpret it. For trail, it compares mostly similar terrain sessions and duration. The slope explains metabolic cost, but does not fully capture eccentric damage from the descents or technical difficulty.

Case. Unai Ferrer illustrates why this difference matters: his trail activity of 2026-08-30 was 203.31 CT with the current calculation. The validation found that the previous model would have given 331,61 CT when treating worse slope and speed; they are not two different workouts, but two formulas for the same record.

How is CT calculated in swimming with CSS?

What it's for and what it stands for. Critical Swim Speed (CSS) acts as a speed threshold in swimming. CT compares swimming speed with CSS; rest on the wall does not add up as time swimming.

How to interpret it. IR 1 is equivalent to swimming to CSS. Exponent 3 is a convention consistent with the growth of drag, not an individual truth. The concept of critical speed as a performance index in swimming is described in Wakayoshi et al. (1992). If there is no CSS, swimming does not reuse a race threshold HR: it passes to RPE or to estimate.

Case. Marta Quintana is the example triathlete and combines swimming, cycling and racing. His 172 swimming activities by CSS were satisfied with the local contrast: it is a good basis for reading the CT of his sessions, without turning it into a direct comparison with his bike CT.

How is CT calculated with heart rate (TRIMP)?

What it's for and what it stands for. When there are a series of FCs and references measures of rest, maximum and threshold, TrainingForever uses integrated TRIMP. It gives more weight to high minutes than to soft ones; two sessions with the same average HR may have different loads if one contains peaks.

How to interpret it. Standardisation makes 60 minutes to the threshold HR worth 100 CT. Banister's formula published b = 1,92 for men and b = 1,67 for women. The product uses a configurable coefficient of 1.92 (a single value for all people); it is not a universal physiological constant. Using the sex profile to change it is pending, because it would be a new use of a personal data. Hydration, stress, and HR delay change the result: it is not a pure measure of external work. If there is no stream or measured references, the backup uses medium HR and should be read with caution.

Case. Pablo Mena has days with pulse alone and others with RPE. In its activities that had both sources, the mean bias RPE versus HR was +4,9 %; alternating both sources moved AC up to 1,91 points and FA up to 9,33. That oscillation can be instrumentation, not physical.

How is CT calculated with the RPE or when data is missing?

What they serve and what they represent. The session RPE (Foster et al., 2001) allows you to load sessions without sensor — including strength — from the perceived effort of the entire session. If there is no EPR, an estimated TC, labelled as such, is retained.

The valid EPR is limited to 1–10; 7.5 is the configurable reference from “one hour to the threshold”. The duration-estimated CT does use an implicit and neutral IR of 0.70 (about 49 CT per hour) for the calculation, but the product does not store or display a measured IR: it leaves it empty to point out that a better measurement is missing.

How to interpret them. A RPE 6 for 45 minutes provides 60 CT. Do not convert the figure into an exact equivalent with power or TRIMP; its value is to sustain a consistent history when there is no device.

Case. In the lot of example athletes, Unai had 90 manual executions and Paul 39. They are the practical reminder that a session without GPS is not a session without load, but it must retain its origin so that coach and athlete read the panel with the right uncertainty.

What is time in the area?

What it's for and what it stands for. Count seconds in HR zones, power or pace. It serves to describe the intensity distribution: where the time went, not how much “valued” a session.

Each sample brings the time to the next; large gaps and time without movement do not add up. The last sample inherits the previous interval. A value in the gap between adjacent areas goes to the nearest area.

How to interpret it. Look for session patterns, not magical percentages; if you need to define the zones, you can use the heart rate zone calculator. The intensity distribution is studied from Seiler (2010). HR arrives late and can place short intervals in low areas. During Sergio's run, the product recorded 1:03:39 (43.9%) HR in Z4 and simultaneously 1:54:16 (78.7%) pace in Z1: two signals describing different dimensions of the same output.

What is aerobic efficiency?

What it's for and what it stands for. The product and documentation call it Aerobic Efficiency. Related external production to HR: weighted power per bike pulse, or equivalent speed in plain per pulse in race and trail.

How to interpret it. The higher, the more production by beat in that session, but it is only comparable for the same athlete, discipline, terrain, duration and conditions. Dream, caffeine, and hydration move it; don't use it to compare two people.

Case. In Sergio's long run, aerobic efficiency was 1,1161 (m/min)/lpm. That figure alone does not say “good” or “bad”; it allows to compare their stable circulations of similar characteristics and feed the cardiac drift.

What is cardiac drift and how is it interpreted?

What it's for and what it stands for. It measures how much efficiency falls between the two halves of a stable session (the phenomenon is described in Coyle and González-Alonso, 2001). You can point out that, at the same approximate production, the pulse went up; it is a useful reading to talk about aerobic basis, heat, hydration or initial pace.

Here, E₁ y E₂ are the aerobic efficiency of the first and second half respectively.

How it's calculated. It only exists if the session lasts more than an hour and passes a stability test. By bike VI ≤ 1.05 is required; in running/trail the stability of the equivalent speed is checked after cutting the first ten minutes. Swimming does not calculate Deriva.

How to interpret it. Practical reference: less than 5 % usually indicates good aerobic endurance to this intensity; 5–10 % calls for context; more than 10 % suggests that the effort may have been above the aerobic threshold. It is not a diagnosis: a short session, intervaled or with poor data appears worthless, not with 0%.

Case. Sergio's long run was 3.49 %, below that 5 % indicative. The graph normalizes the first half to 100 and shows the second in 96.51; it does not invent two absolute efficiencies that the record does not publish.

What are the best efforts and the best-effort curve?

What they serve and what they represent. A better effort is the best average that appears in a window of 5 s, 1, 5, 20 or 60 min. The best-effort curve takes the best of each duration over a period. There are curves of power, pace and HR.

For power and HR it is maximized; for pace the shortest time per kilometer is chosen. A window with invalid gaps doesn't count.

How to interpret them. These are detected opportunities, not guaranteed maximum tests. A high HR is not performance. In pace and HR, the 5 s have little practical validity by GPS and cardiac delay; from 1 minute reading is usually more useful. It compares equivalent periods and checks whether there really was a sustained effort. If you need a controlled effort, consult the 6 minute test for VAM.

Case. In its long run, Sergio recorded his best paces of 4:30/km (5 s), 4:35/km (1 min), 4:50/km (5 min), 5:27/km (20 min) and 5:31/km (60 min). This describes this activity; it does not allow to conclude that it has improved its marathon performance.

What is the rate of progression?

What it's for and what it stands for. It summarizes how the AC changed during the last week. It's a loading ramp signal, not a predictor of injury.

How to interpret it. Positive means that today's AC exceeds that of seven days earlier; negative, that it went down. It does not have a universal safe cut: acute/chronic ratios and injury inferences from them have mathematical coupling problems. Compare it to the athlete's usual progression, plan and response.

Case. The validation with sample data found that, during a construction of Álvaro, his AC went from 38.24 to 58.07. This results in an increase in the load model, not a proven improvement in its power of 20 minutes.

What is the expected load of a planned session?

What it's for and what it stands for. It's the expected CT from a plan session. It allows to draw the future stretch of Evolution and contrast later planned and realized.

Use the corresponding branch of the actual CT, zones and thresholds. The midpoint of each zone produces the IR; the recoveries use IR 0.30; by bike, the steps are added with average order 4 to approximate the weighted power. Force uses RPE × minutes. Without threshold, numerical zone or sufficient structure, it is marked estimated.

How to interpret it. It is a cargo forecast, not a compromise or a result. For example athletes, with a duration of between 95 and 105 % of the planned duration, the average absolute deviation was 3,9 % in swimming, 5,1 % in running, 6,8 % in trail and 3,9 % in cycling with recoveries after correction of the formula.

Case. Marta, with three disciplines, is the best reminder to read it as a hypothesis: the total of one day can reach 368 CT (17 May in the example series), but the plan of the next day remains an estimate until it is executed.

How do you read AC, FA, DC and PR in Evolution?

AC, FA, DC and PR are the four curves of the Evolution panel of the daily model. They are based on Banister and Calvert's impulse-response model (1980).

These four curves are the most visual part of the daily panel. There are four sensors: AC and FA are filters over CT; DC and PR are derived from them. The series works for UTC days and is heated with previous history so as not to artificially start at zero. All examples in this section — including Sergio's chart — use the daily scale: the graph bars add seven CTs per day, and AC, FA, DC, and PR are the last day daily values of each block. The different inherited weekly scale is explained below.

What is Chronic Adaptation (AC)?

What it's for and what it stands for. AC approximates the burden that the athlete has been tolerating: a slow memory of the daily load. It does not directly measure physical condition, VO2max or specific preparation.

How to interpret it. It rises slowly in a sustained construction, stabilizes with constant load and gradually lowers in a stop. The 42 days are a common configuration, not an individual setting. Compare it with your own trajectory, not a population table.

Case. In Sergio's four weeks of the chart, the AC went from 41 to 46 although the weekly CT ranged from 305 to 435. That slowness is precisely its function: not reacting like a training counter yesterday.

What is Acute Fatigue (FA)?

What it's for and what it stands for. FA approximates the recent weight of the load with a fast memory. You do not know sleep, pain, illness, work stress or HRV (for that, consult the HRV Guide for Coaches).

How to interpret it. Climb faster than AC by pressing and fall before downloading. Seven days is not a judgment on individual recovery: it is the current filter configuration.

Case. The week of 435 CT brought Sergio's FA to 56 while his AC was 44. The data invites you to review how you were responding, not automatically cut the next session.

What is the Load Differential (DC)?

What it's for and what it stands for. DC compares the adaptation and fatigue the athlete brings before training that day. It is the evolution of the old “Balance de Frescura”; it is not shown as an additional metric.

How to interpret it. DC negative points out that the recent load weighs more than the load memory; it can be normal under construction. Close to zero indicates relative balance. Positive usually appears in download, taper or inactivity. There are no definitive clinical ranges and no positive DC is equivalent to complete recovery.

Case. On September 14, Sergio showed DC −14. It is the result of the previous values that the product showed on that day (AC 44 and FA 58), not to subtract the closing values after loading (AC 44 and FA 56). The last week of the chart reached 0; that describes the change in signal, not a “risk” label.

What is the Performance Potential (PR)?

What it's for and what it stands for. PR draws the theoretical potential to express accumulated load when FA is lower. It is deliberately a model reading, not a prediction of time, podium or power.

How to interpret it. It usually goes up if AC is preserved and FA falls, like in a paper. You can also climb a stop because the FA falls before the AC: that’s why it should never be sold as “you’re in more shape.” On screen, PR and AC refer to different temporary moments during a day with load; they do not have to visually add up with the figures of the same day.

Case. In Sergio, PR went from 30 to 45 when DC returned from −14 to 0 between the weeks of 8–14 and 22–28 September. The mathematical explanation is not enough: you should look if you kept the specific marathon stimuli.

What do the metrics of the historical weekly model mean?

The historic weekly model of TrainingForever uses hours and unevenness to calculate the load of each week.

The following formulas belong to the first weekly generator and remain alive on certain inherited surfaces. They are not the daily CT of activities described above. You should know them so you don't mix scales.

What are Internal Load (CI) and External Load (CE)?

CI: What it's for, calculation and reading. Weekly CI represents training time:

His case is Álvaro: a week of many hours bike increases IC even if part is soft. The limitation is important: in this model a gentle hour and an hour of intervals weigh the same before transforming the load.

EC: what it is for, calculation and reading. CE represents mechanical stress; the first model directly used weekly positive unevenness:

It also defines equivalent distance to compare mountain and plain:

The older version used /100; the rear engine, /120. For Unai, CE and equivalent distance give context to the mountain that IC does not see. Neither capture completely lowered, technical or eccentric damage.

How is the Total Weekly Load calculated?

What it's for and what it stands for. Merge CI and CE of the weekly model to feed your notices and their summary.

How to interpret it. The coefficients 55 and 0.1 are internal calibration, not physiological constants. A week of 4 h and 500 m D+ would give 270 CT weekly in this model. The case of Unai makes it clear why you have to look at both entries: the slope can raise the CT even if the hours do not change.

What is Total Base Load?

What it's for and what it stands for. It is the reference with which the weekly generator avoids starting from scratch when creating a plan.

V_med is the weekly average of hours of history and D_med, the weekly mean of positive unevenness. It replicates as the initial four historic weeks. Interpretation: describes the starting point of the plan, not a form measurement. In Paul, whose continuity is irregular, the base must be read together with his real history: an average can hide weeks of pause. There is no universal “correct” base range.

How are AC, FA, DC and PR calculated in the weekly model?

The inherited model uses the same words with other formulas and another scale: here CT, AC, FA, DC and PR are weekly. Therefore, Martha's example in this section is weekly and does not compare numerically with the daily examples of Sergio, Álvaro or Pablo above.

What are they for? AC summarizes four weeks; FA is the current week; DC expresses its distance and PR its transformation. How to interpret them. Negative DC can fit into construction and positive on discharge, as in the daily model, but the answer is much more abrupt because FA is directly the current week. Case. In a construction phase of Marta, a week with more volume can immediately alter the weekly FA although the weekly AC takes four weeks to reflect the change. That is why their values are not compared to those of Evolution day by day.

What is the recent load ratio?

What it's for and what it stands for. It measures this week's load against the average of the previous four. In the prototype it was called "load-pending", although it is a ratio.

How to interpret it. R=1 is equal to the recent average; 1.30 is 30% above. Prototype limited R a 1.3 as a conservative product rule, not as a medical or injury threshold. Hypothetical example (Illustrative round figures, not Paul's data): If an athlete goes from a weekly average of 100 CT weekly to 130, the ratio is 1.30: it is up to ask for change, not to conclude that he is injured or that he is stronger.

What is the weekly fatigue budget?

What it's for and what it stands for. It is another safeguard of the prototype: it compares each week with the base and with AC.

How to interpret it. If overcome, the generator reduced the level and then hours. Multipliers 1,15 and 1,20 are internal configurations, not biological boundaries. Case. In trail, Unai can exceed the budget for D+ even if their hours do not seem extraordinary; the coach needs to decide whether it was a specific week planned or an accumulation difficult to absorb.

What figures are not yet operational metrics?

Do not confuse proposals with active functionalities. A standard form index of 0 to 100, a quantified penalty for downloads or technical and definitive ranges of DC or PR do not yet have a stable specification. TrainingForever does not present them as available figures or scientific conclusions.

It is also not appropriate to call DC, Pace of progression or weekly ratio “risk of injury”. Scientific literature warns conceptual problems and mathematical coupling in this type of quotients (Impellizzeri et al., 2020). They're signals to open a conversation, not permissions to automate decisions.

What does each acronym of the metrics mean?

This glossary summarizes each acronym of the guide in one line.

• AC: Chronic adaptation; slow memory of CT.
• EC: External load; in the weekly model, mainly D+.
• IC: Internal load; on the weekly model, hours.
• CSS: critical swimming speed used as a threshold.
• CT: Total load; internal loading unit. Daily it is derived from activity; weekly it has another formula.
• DC: Load differential; AC minus FA before training the day.
• D+: accumulated positive difference.
• Aerobic efficiency: Weighted power/HR or equivalent speed/HR; this is the product name.
• FA: Acute fatigue; fast CT memory.
• IR: Relative intensity compared to the current threshold.
• PP: Weighted power.
• PR: Performance potential; AC and FA transformation, not brand prediction.
• RPE: perceived session effort, scale 1–10 in the application.
• TRIMP: HR-based and exponentially weighted training impulse.
• VI: Variability, PP/average power.

How do metrics help the conversation between coach and athlete?

TrainingForever metrics help the conversation between coach and athlete by making the context of each session and each week visible.

The most useful reading brings together three things: the graph, the reality and the objective. For Álvaro, example athlete, power and IR allow to check if the bike load was the one sought. For Unai, slope and D+ return sense to a trail that a gross pace would deform. For Marta, it is appropriate to separate disciplines before reading the total. For Pablo, the origin HR/RPE avoids confusing a sensor difference with an adaptation. For Sergio, a 3.49 % drift is interpreted along with 2:25 race, pace and objective test visible, Marathon of the Ribera.

That is the purpose of metrics: not to decide for the person, but to make the context visible to decide better with it.

TrainingForever is a Spanish platform for endurance and strength coaches: free version up to 5 athletes and then 0.99 € per active athlete with a maximum of 49 €/month. Includes an AI Assistant who prepares drafts. Coach decides.

You can check the updated conditions at prices of TrainingForever, compare alternatives in the guide best software for coaches or see how to start with the free version for coaches.

What is the model based on and what are its limits?

The daily model is based, among other references, on Banister and Calvert (1980) for impulse-form/fatigue response, Allen and Coggan for weighted power, IR and load scales, in Minetti et al. (2002) for the cost of the slope, in Foster et al. (2001) for the RPE, in Wakayoshi et al. (1992) for CSS, Seiler (2010) for time in zones and Coyle and González-Alonso (2001) and Friel for drift. Part of the formulas are product conventions, not physiological laws, and are marked as such in each section. The main practical rule remains: an uncalculated metric is without data, never zero.

Frequently Asked Questions

What is Total Load (CT) in TrainingForever?

Total Load (CT) is the common unit with which TrainingForever summarizes how much load an activity involves: one hour at the threshold equals 100 CT. Several activities of the same day are added and a day without activity contributes 0. It is modeled load, not calories or muscle damage, and it should be read together with its origin: power, pace, HR, RPE or estimation.

What is the difference between AC, FA, DC and PR?

AC (chronic adaptation) is a slow memory of CT, with a constant of 42 days; FA (acute fatigue) is a fast memory, of 7 days. DC (Load Differential) is the AC minus the previous day's FA, and PR (Performance Potential) is 2·AC − FA of the previous day. AC and FA are calculated from CT; DC and PR are derived from them.

Does a negative Load Differential mean that the athlete is overtrained?

No. A negative DC indicates that the recent load weighs more than the load memory, something that may be normal in a construction phase. A positive DC usually appears in a download, taper, or pause. There are no definitive clinical ranges and it is not appropriate to call it "risk of injury": it is a sign to open the conversation with the athlete.

Does the Performance Potential predict a brand?

No. PR is a transformation of AC and FA (2·AC − FA of the previous day), a model reading, not a prediction of time, podium or power. You can go up on a taper, but also on a stop, because the FA falls before the AC; therefore it should never be interpreted as "you're in more shape."

When is the cardiac drift calculated and when is it not?

The cardiac drift is only calculated if the session lasts more than an hour and passes a stability test: by bike, a Variability (VI) of 1.05 or less; in race and trail, an equivalent stable speed after cutting the first ten minutes. Swimming doesn't calculate drift. If no condition is met, the metric appears without data, not with 0 %.

Why does the AC of the weekly model not match that of Evolution?

Because they're two different models with the same acronyms. In Evolution, daily AC uses a 42-day memory of the CT of each activity. In the historical weekly model, AC is the average of the last four weeks of a CT with another formula (55 per hour plus 0.1 per unevenness). Their figures should not be compared one by one.

What if a metric cannot be calculated?

It appears as "no data", never as zero. It occurs, for example, without potentiometer or less than 20 minutes for weighted power, or without CSS in swimming. The CT then uses the next available source (e.g., the RPE) or, if there is none, an estimate labeled as such, and it should be read more carefully.

Sources

1. Banister EW, Calvert TW. Planning for future performance: implications for long term training. Can J Appl Sport Sci. 1980;5(3):170-6. PubMed 6778623
2. Minetti AE, Moia C, Roi GS, et al. Energy cost of walking and running at extreme uphill and downhill slopes. J Appl Physiol. 2002;93(3):1039-46. doi:10.1152/japplphysiol.01177.2001. PubMed 12183501
3. Foster C, Florhaug JA, Franklin J, et al. A new approach to monitoring exercise training. J Strength Cond Res. 2001;15(1):109-15. PubMed 11708692
4. Wakayoshi K, Ikuta K, Yoshida T, et al. Determination and validity of critical speed as an index of swimming performance in the competitive swimmer. Eur J Appl Physiol. 1992;64(2):153-7. doi:10.1007/BF00717953. PubMed 1555562
5. Coyle EF, González-Alonso J. Cardiovascular drift during prolonged exercise: new perspectives. Exerc Sport Sci Rev. 2001;29(2):88-92. doi:10.1097/00003677-200104000-00009. PubMed 11337829
6. Seiler S. What is best practice for training intensity and duration distribution in duration athletes? Int J Sports Physiol Perform. 2010;5(3):276.91. doi:10.1123/ijspp.5.3.276. PubMed 20861519
7. Impellizzeri FM, Tenan MS, Kempton T, et al. Accute:chronic workload ratio: conceptual issues and fundamental pitfalls. Int J Sports Physiol Perform. 2020;15(6):907-13. doi:10.1123/ijspp.2019-0864. PubMed 32502973

References verified in PubMed (E-utilities) on 23/10/2026. Allen and Coggan (weighted power, IR) and Friel (derivative) are training books without verifiable link; the sample figures come from internal simulated data, not from these studies.