Acute:Chronic Workload Ratio, Explained Without the Jargon

Your Garmin says “Unproductive.” Runna just dropped a 26-mile week on you after three weeks in the high teens. Somewhere in a Reddit thread you saw someone say their ACWR hit 1.6 and they got a stress fracture two weeks later. So now you’re staring at a number you half understand, wondering whether to trust the plan.

Here’s the short version of what acute:chronic workload ratio actually is. It compares what you did this week against what you’ve been averaging recently. That’s it. This week versus your recent normal, expressed as a single number.

If your recent normal is 30 miles a week and this week you ran 36, your ratio is 1.2. You did 20% more than usual. If you ran 24, it’s 0.8, and you did 20% less. The number is a fraction, not a diagnosis.

The actual sum

The standard version uses seven days for “acute” and 28 days for “chronic.”

Acute load: the total for the last 7 days. Chronic load: the daily average over the last 28 days, multiplied by 7 to make it comparable. ACWR: acute divided by chronic.

Say your last four weeks looked like this:

Week -4:  24 miles
Week -3:  27 miles
Week -2:  30 miles
Week -1 (this week):  35 miles

Chronic (28-day total / 4):  (24+27+30+35) / 4 = 29 miles
Acute (this week):  35 miles
ACWR:  35 / 29 = 1.21

That’s a 21% bump over your rolling normal. Perfectly ordinary for a build week.

Note what just happened: this week is inside the chronic average as well as being the acute number. That’s the “coupled” version, and it’s what most consumer tools use. It drags the ratio toward 1.0 and makes big spikes look smaller than they are. The “uncoupled” version excludes the current week from the denominator, so the same data gives 27 chronic and a ratio of 1.30. Neither is wrong. They’re just different rulers, which is the first hint that you should not be reading this number to two decimal places.

Mileage is the crude version. Use something better.

Miles ignore intensity. A 10-mile easy run and a 10-mile session with 6 x 1K at 5K pace are not the same stress on tendon and bone, but they log identically.

Two better inputs, both of which you already have:

Session RPE. Rate each run 1 to 10 for effort, multiply by duration in minutes. A 50-minute easy run at RPE 3 is 150 “load units.” A 50-minute session with hard intervals at RPE 8 is 400. Add up the week. This takes fifteen seconds a run in a notes app and it captures intensity better than anything your watch calculates, because you were there and it wasn’t.

Your watch’s own load metric. Garmin’s Training Load is a 7-day rolling sum of EPOC-derived numbers per session. Its Acute Load and Chronic Load fields (on newer Forerunners, Fenix and Epix, under Training Status) are precisely the two halves of this ratio, and the “Load Ratio” gauge Garmin shows you is ACWR under a friendlier name. Polar calls its version Cardio Load Status and shows the same relationship as a “Strain vs Tolerance” ratio. Coros has Training Load with a 7-day and 42-day comparison. Strava’s Fitness and Freshness chart (Impulse and Form, if you have a power meter or heart rate data) uses 42-day and 7-day exponentially weighted averages, which is the same idea with a longer tail and a smoother decay.

If you’re on Runna or a ChatGPT-built plan and want one number without new gadgets, session RPE is the honest choice. It costs nothing and it doesn’t pretend to precision it lacks.

The 1.5 thing

You’ve probably seen the claim: stay between 0.8 and 1.3, and above 1.5 is the danger zone. This came out of Tim Gabbett’s work on Australian rules football and rugby league in 2016, and the original graph showed injury risk climbing sharply past about 1.5 in those athletes.

The idea spread fast, partly because it’s tidy and partly because it fits neatly into a dashboard gauge. It has taken a real beating since. A 2020 critique by Impellizzeri and colleagues pointed out serious methodological problems, including that dividing one number by another that contains it creates a mathematical artefact (this is the coupling issue above, and it can generate an apparent relationship where none exists). A 2021 consensus statement from the International Olympic Committee working group was notably cooler on ACWR than the 2016 enthusiasm. Several replication attempts in running populations found weak or no association between the ratio and subsequent injury.

Which is not the same as saying the concept is useless. Here’s the distinction that matters for you: ACWR is a poor predictor and a decent descriptor. It won’t tell you that you’re going to get injured. It will tell you, accurately and instantly, that the week your plan has given you is 45% bigger than anything you’ve done in a month. That second thing is genuinely worth knowing, and it’s the only thing you should use it for.

There is no cliff at 1.5. There is no safe zone at 1.29. A ratio of 1.4 in week nine of a marathon block after eight clean weeks is a different animal from 1.4 in your second week back from a calf strain, even though the number is identical. Treat it as a direction-of-travel check, the way you’d treat a car’s fuel gauge: useful for noticing you’re heading toward empty, useless for knowing exactly how many miles you have left.

Three worked examples from real plans

Example one: the app that front-loads.

A 16-week Runna marathon plan, runner coming off a fairly casual base.

Pre-plan 4-week average:  18 miles/wk
Plan week 1:  24 miles
Plan week 2:  28 miles
Plan week 3:  31 miles

Week 1 ACWR (uncoupled, against an 18-mile chronic): 24 / 18 = 1.33 Week 2, chronic now roughly 19.5: 28 / 19.5 = 1.44 Week 3, chronic roughly 21.75: 31 / 21.75 = 1.43

Three consecutive weeks north of 1.3. The plan isn’t being reckless exactly, it’s being generic: it doesn’t know what you were doing before you downloaded it. The right response isn’t to abandon the plan. It’s to run weeks 1 to 3 at the lower end of the prescribed ranges, hold the long run shorter than printed, and let the chronic number catch up. By week 6 you’ll be on the plan’s schedule with a ratio near 1.0 and you’ll have lost nothing.

Example two: the illness dip that looks like a spike.

Week -4:  38 miles
Week -3:  40 miles
Week -2:   6 miles (chest infection)
Week -1:  10 miles
This week:  34 miles (feeling better, back on plan)

Chronic over the last four weeks: (40+6+10+34) / 4 = 22.5. ACWR: 34 / 22.5 = 1.51.

Your watch will flag this. Your watch is right to flag it, and also the situation is more nuanced than the number. You have genuine fitness from weeks -4 and -3 that a 28-day window is aggressively forgetting. But two weeks of near-total rest does detrain tendon and bone stiffness, and the classic post-illness return-to-running injury is exactly this pattern. The sensible read: 34 is too much too fast, 25 would have been better, and the ratio caught something real even if its reasoning is crude.

Example three: taper panic.

Marathon week. You run 14 miles against a chronic of 42. ACWR: 0.33. Garmin says Detraining. Strava’s Form line is deep in the positive.

Ignore all of it. A taper is supposed to produce a low ratio. The metric has no concept of “race on Sunday,” and a low number here means the taper is working. This is the clearest case of why you shouldn’t outsource judgment to the ratio: it’s describing a real change in load that you deliberately caused for good reasons.

How to actually use it

Check it weekly, not daily. Daily ACWR is noise: your long run lands on Sunday and the ratio jumps, then decays for six days. Nothing happened. Look at it once a week, on the same day, and look at the trend across four or five weeks rather than the current value.

Ask three questions when a week comes in above roughly 1.3:

1. Is this the third week in a row above it? Repeated moderate spikes matter more than one big one.
2. Did the extra load come from volume or intensity? Adding 4 easy miles is a different proposition to swapping an easy run for a track session.
3. Is anything actually sore? A niggling Achilles alongside a 1.4 is a real signal. A 1.4 with fresh legs is information, not an emergency.

Down-weeks are where the ratio earns its keep. Most AI-generated plans include a cutback every third or fourth week, and most runners quietly skip them because they feel good. Watching your chronic load climb without a single dip for ten straight weeks is the pattern that precedes a lot of injuries, and it’s much easier to see in this framing than by staring at a wall of Strava activities. If you want the fuller picture of how load relates to injury risk, including the bits ACWR leaves out entirely (sleep, life stress, surface, shoe rotation, biomechanics), the training load and injury risk guide covers it properly.

One more thing about AI plans specifically. Runna, Garmin Coach and a well-prompted ChatGPT plan all have the same blind spot: they progress from where they think you are, not from where you actually are. They don’t know you took ten days off in August. They can’t see that your last four weeks contain a half marathon you ran too hard. Calculating your own ACWR at the start of a block, and again at any point where the plan’s numbers feel wrong, is the cheapest sanity check available for a plan that nobody is supervising.

And when the ratio and your body disagree, believe your body. The number is a summary of what you did. Your Achilles is a report on what it cost.