The Four Numbers Worth Checking Every Sunday Night

Your watch gave you forty numbers this week. Training Effect, aerobic and anaerobic, scored to one decimal place. A VO2max estimate that moved from 48 to 49 and will move back on Tuesday. Ground contact time balance of 49.2% / 50.8%. Body Battery, HRV status, Training Readiness, a Load Focus pie chart telling you that you are “lacking high aerobic.” Runna has its own weekly load figure. Strava has Fitness and Freshness if you pay. Somebody on Reddit told you to watch your acute:chronic workload ratio.

Almost none of it will tell you what you actually need to know on a Sunday night, which is this: is the plan I am following moving me forward at a rate my body can absorb?

Four numbers answer that. Everything else on the list above is detail you can look at when you are curious, or when one of the four has already told you something is off. If you are following a Runna block, a Garmin Coach plan, or a plan ChatGPT wrote you in forty seconds, these are the weekly running metrics to track, and the whole review takes about five minutes.

1. Weekly volume

Total running for the week. Kilometres or miles if your plan is distance-based, minutes if it isn’t. One figure.

Get it from Strava’s weekly bar on your profile, or Garmin Connect under Reports, or the Runna app’s week summary. Don’t compute it, read it, and write it down somewhere the app can’t retroactively change.

The point of logging it by hand is that you capture what you did, not what was prescribed. AI plans are generated against an assumption about you that was true on the day you filled in the onboarding form. A Runna plan built in June around four runs a week does not know that you have since started playing five-a-side on Wednesdays, or that your “easy 8km” is routinely 9.4km because of where the loop ends.

2. The volume trend

This is the number almost nobody writes down, and it is the most useful of the four. Divide this week’s volume by the average of the previous four weeks.

This week ÷ mean(previous 4 weeks) = trend ratio

Sit between roughly 0.8 and 1.3 and you are progressing at a rate the research on training load broadly supports. Below 0.8 for several weeks running and you’re detraining (fine if it’s a deliberate down week, less fine if it’s three in a row). Above 1.3 and you’re in the zone where injury risk starts climbing steeply, and the risk compounds if you stay there for consecutive weeks.

Here’s a real shape of problem, from a marathon plan generated by a chat model with no memory of the previous conversation:

Look at week 8. Nothing in that plan looks reckless line by line: every week is roughly a 10% bump, which is the rule everyone half-remembers. But the 10% rule compounds, and the plan never inserted a second down week, so by week 8 the runner is loading 38% above their recent baseline while also doing their longest run ever. That is where calf strains and posterior tibial grief come from, and the runner never sees it, because Garmin’s Training Status probably said “Productive” the whole way.

Fixing it is unglamorous. Week 8 becomes 48km instead of 66km, week 9 picks up at 58, and the long run that was scheduled for week 8 slides to week 10. You lose nothing except a number on a chart. If you want the physiology behind why the trend matters more than the absolute figure, our training load and injury risk guide goes through the mechanism properly.

3. Hard sessions

Count them. A hard session is anything where you were working above comfortable conversational effort for a meaningful chunk of the run: intervals, a tempo or threshold block, hill reps, a parkrun you actually raced, a long run with race-pace segments, a hard group run you got dragged along on.

Two is the working number for most recreational runners. Three if you’re experienced, consistently above 60km a week, and sleeping properly. Four is where things break.

This is where AI plans quietly overreach, and where the count catches it. A Runna 10K block might give you this in a single week:

• Tuesday: 6 × 800m at 10K pace
• Thursday: 20 minutes at threshold
• Saturday: easy 6km
• Sunday: 16km long run with the final 5km at half marathon pace

Three hard sessions, and the plan knows it. Add the parkrun you did on Saturday instead of the easy 6km because a friend was visiting, and you’re at four. Nothing in the app registers the substitution as a problem. Strava logged it as a 5km run. Your legs logged it as a race.

The count also catches the opposite failure, which is real and more common than people think: a plan that has you doing six runs a week, all of them at 5:30/km, none of them genuinely easy and none genuinely fast. Zero hard sessions and 55km of grey middle. You’ll feel tired all the time and get slightly worse.

4. A subjective fatigue note

One line. Written by you, in words, before you look at any of your data.

Not a number your watch generated. HRV and Body Battery are attempting to measure exactly this thing and are noticeably worse at it than you are, because they can’t see that you drank three pints on Friday, moved house on Saturday, or that your right calf has felt like a guitar string since Wednesday’s reps.

Mine looks like this:

Week ending 28 Sep
volume   52 km (planned 52)
trend    52 / 44.8 = 1.16
hard     3  (Tue 6x800, Thu 8km tempo, Sun 32km w/ last 10 at MP)
note     3/5. Thursday tempo felt like 10K effort, not threshold.
         R calf tight Friday morning, gone by Saturday. Sleep poor Wed/Thu.

Score out of five if you like, but the sentence is the part that works. “Thursday tempo felt like 10K effort” is the observation that matters: the same session, at the same prescribed pace, cost more than it did a fortnight ago. That’s the earliest honest signal of accumulated fatigue you will get, and it arrives about a week before your paces actually fall off.

The decision rule is simple. Any complaint that appears in two consecutive weekly notes gets acted on. Once is noise. Twice is a pattern, and a pattern means you cut the next week’s volume by 20% and drop one hard session, regardless of what the plan says or how “Productive” your watch claims your training status is.

Reading the four together

The four numbers are only worth much in combination, because the combinations are diagnostic.

Volume flat, hard count at 2, note consistently 4/5 and cheerful? You have room. Add 10% next week, or a third quality session, not both.

Trend above 1.3 with a good fatigue note is the trap, because you feel fine and the plan is working and you have no reason to intervene. Intervene anyway. Tissue tolerance lags fitness by weeks, and the week you feel invincible is a well-documented setup for the week you can’t run.

Trend below 1.0 with a poor note and four hard sessions is the one to catch fastest. You’re doing less work and it’s costing you more. That’s usually the plan drifting into all-intensity-no-volume, and the fix is to make your easy runs genuinely easy for ten days.

And when the numbers say one thing and the app says another, the numbers win. Runna does not know your calf is tight. Garmin Coach does not know you raced parkrun. The plan is a hypothesis about a version of you that existed at onboarding, and a hypothesis is meant to be checked against evidence.

Open a note on your phone now and write four lines for the week that just finished. You’ll have the volume and the hard count in ninety seconds. The trend needs your last four weeks, which Strava will show you on one screen. Then write the sentence about how it felt, and next Sunday you’ll have something to compare it to.