Where Acute:Chronic Ratio Quietly Lies to You

You open Runna or Garmin Coach, or paste last month’s Strava log into ChatGPT, and the plan comes back with a tidy acute:chronic workload ratio (ACWR) of 1.1. Green light. Safe zone. Off you go.

Maybe. ACWR is a useful smoke alarm, but it has blind spots, and they sit in exactly the situations where recreational runners get hurt. This post covers three of them, with numbers, and then what to check next to the ratio so you aren’t relying on one figure.

A quick definition, so we’re using the same numbers

ACWR divides your recent load (the acute window, usually 7 days) by your longer-term load (the chronic window, usually 28 days, averaged to a weekly figure). The popular guidance, drawn from Tim Gabbett’s work, treats roughly 0.8 to 1.3 as the “sweet spot” and anything above about 1.5 as a spike.

ACWR = (load this week) / (average weekly load over the last 4 weeks)

Note the word “load” there. It can be kilometres, minutes, or something weighted by intensity such as Garmin’s training load or TRIMP. What goes in changes what comes out, which is where the trouble starts.

For the wider picture on how load relates to injury, see our pillar guide on training load and injury risk. The rest of this post is about where the ratio misleads.

Failure 1: Low chronic load after time off

This one catches people returning from illness, travel, a niggle, or a lazy August.

Say you were running 40 km a week through spring. Then a chest infection takes you out for two weeks. Your last four weeks look like this:

Chronic average: 20 km a week. Now your plan asks for 24 km this week. ACWR: 24 / 20 = 1.2. Comfortable, according to the number.

But your tissues don’t care about a four-week average. Tendons and bone adapt slowly, and detraining erodes some of that adaptation over weeks, not days. The more important point is that a 1.2 here means you’re jumping from zero to 24 km in a single week, after two weeks with nothing. The ratio is normal because the denominator was dragged down by the break. It reports “not a big increase on a small number”, and the small number is the problem.

Now the flip side: the ratio can also flag you falsely. If you were off for a week and the plan restarts you at normal volume, the ratio may look fine while the first long run is the biggest stress in a month.

What to check instead:

• Compare against your last healthy stable block (say, the 8 weeks before the break), not only the last 28 days.
• After 10+ days off, restart at around 50 to 60 percent of your pre-break volume and rebuild over 2 to 3 weeks, whatever the ratio says.
• Look at how long it has been since your longest run. If your last 16 km was six weeks ago, a 16 km long run is not “maintenance”.

Apps differ here. Runna generally re-plans after missed sessions if you tell it, but it won’t always notice you’ve had an illness you never logged. A ChatGPT-built plan has no idea at all unless you tell it the dates.

Failure 2: Distance-only inputs

Paste a CSV of weekly kilometres into a chatbot and ask whether your build is safe, and you’ll get an ACWR computed on distance alone. Distance treats all kilometres as equal. They aren’t.

Compare two weeks, both 35 km:

• Week A: 5 easy runs at 7:00/km, average heart rate 140.
• Week B: 3 easy runs, one 8 x 800 m interval session, one hill-heavy tempo, same 35 km total.

On distance, the ratio is identical. On load, they are far apart. Garmin’s training load (an EPOC-based figure) might read around 380 for Week A and 560 for Week B. If your previous four weeks averaged 400 on that scale, your distance ACWR is 1.0 for both weeks, but your Garmin-load ACWR is 0.95 for A and 1.4 for B.

Distance also misses terrain, heat, and surface. A 20 km trail run with 900 m of climbing is not a 20 km road run. Flat-pace kilometres undercount it badly.

What to check instead:

• Use a load metric that includes intensity: Garmin training load, Coros training load, Strava’s Relative Effort, or TrainingPeaks TSS/hrTSS. None is perfect, but all beat raw distance.
• Count hard sessions separately. A useful rule of thumb: don’t add a new intensity type (first track session, first hills) in the same week you also add volume.
• If you’re in a spreadsheet, add a column for elevation gain and for “hard minutes” (time above threshold heart rate or at tempo pace or faster).

Failure 3: A fine ratio on an absolute load you’ve never handled

This is the quietest lie. ACWR is relative, so it works at any scale, including scales you are nowhere near ready for.

Consider a beginner preparing for a first marathon:

Every row is inside the safe band. Yet in two months the runner has doubled their weekly volume. Each step is “only” 10 percent or so, because the ratio compares to a rolling average that is itself climbing. Steady compounding growth gives you a permanently pleasant ACWR and a body that’s absorbing twice the load it was in week 1.

Absolute thresholds matter too. A week with a 32 km long run is the first time your skeleton meets that duration, regardless of what the denominator says. Most injury risk in novices, tibial stress reactions and Achilles trouble especially, comes from cumulative exposure to new magnitudes, not from sudden ratio spikes.

What to check instead:

• Eight-week change. Compare this week to the same week eight weeks ago. If it’s more than about 50 percent higher, treat the plan as aggressive no matter the ratio.
• Longest run ever. Don’t jump the long run by more than about 2 to 3 km (or 10 to 15 minutes) over your previous longest in the last 6 weeks.
• Long run share. If your long run is over roughly 40 percent of the week’s distance, the weekly total hides a big single-session stress. A 20 km long run in a 40 km week is 50 percent.
• Cutback weeks. A plan that climbs for 8 weeks with no reduction is fighting the physiology. Look for a lighter week every 3rd or 4th week, cutting around 20 to 30 percent.

A quick audit you can do in five minutes

Pull your Strava or Garmin Connect weekly totals and answer these before trusting the next plan block:

1. Did I have any break of 7+ days in the last 8 weeks?
2. Is the plan’s load input distance only, or does it weigh intensity?
3. Is this week more than 50 percent above eight weeks ago?
4. Is the coming long run more than 3 km beyond my longest in six weeks?
5. Is there a cutback week in the next three?
6. Have I added a new type of session this week?

Two or more “yes” answers means the ratio is probably flattering you. Dial back the next week by 10 to 20 percent, or hold it flat, and see how your legs, resting heart rate and HRV (Garmin’s HRV Status or Whoop’s recovery score, if you wear one) respond.

Giving the plan the context it lacks

If you use a chatbot, give it more than a column of kilometres. Something like:

Last 10 weeks (km): 38, 40, 42, 0, 0, 12, 20, 26, 30, 34
Break in weeks 4-5: flu.
Longest run in last 6 weeks: 14 km.
Current weekly Garmin training load: ~420 (was ~500 pre-illness).
Goal: half marathon in 7 weeks.
Please flag anything in the plan that exceeds these.

That prompt exposes the exact failures above: a break, a recent longest run, an intensity-weighted load. A model can then reason about them instead of returning a clean 1.1.

Treat the ratio as one gauge on a dashboard. When it’s red, listen. When it’s green, check the other dials before you believe it.