The Pain Rules: When to Keep Running and When to Stop

Your plan does not know you are hurting.

That is the whole problem in one sentence. Runna will serve you a 6 x 800m session at 5K pace on Tuesday whether or not your left Achilles felt like a rusty hinge on Monday morning. Garmin Coach will nudge your weekly volume up 8% because the calendar says week four, not because your shin agreed. And a ChatGPT-built plan, however carefully you described your history in the prompt, is working from the version of you that existed on the day you wrote the prompt. None of these systems have palpated anything. None of them know that you took the stairs sideways this morning.

So you need a decision rule that sits above the plan. Not a vague sense of “listen to your body”, which is advice that means nothing at 6:40am when you are already dressed and the session is loaded on your watch. Something specific enough to apply mid-run, with a number attached.

That rule is the traffic light, and it has two halves: what the pain does during the run, and what it does in the 24 hours after. Both halves matter. Runners who only check the first half get hurt, because plenty of injuries feel fine while you are warm and catch up with you the next morning.

Half one: the 0-10 scale, applied during the run

Rate the pain out of 10, where 0 is nothing and 10 is the worst pain you can imagine. Be honest, and rate the specific structure that is complaining, not your general fatigue. Legs burning at the end of a threshold rep is not pain in this sense. A focal ache in one tendon is.

Green: 0-2/10. Keep running. This is the background noise of a body under load. It should not increase as the run goes on, and it very often decreases once you are ten minutes in. Warming up should help.

Amber: 3-4/10. Keep running, but with conditions. The pain must stay at or below 4, it must not change your gait, and it must not be climbing. If you find yourself shortening your stride on one side or landing differently to protect something, that is no longer amber. Gait change is a hard stop regardless of the number.

Red: 5/10 and above. Stop. Walk home or call someone. There is no session worth completing at 5/10 in a single structure, and the argument that you will “just finish this rep” is how a two-week problem becomes a ten-week one.

The amber band is where all the judgment lives, and it is the band most runners get wrong in the same direction: they treat 4/10 as a licence to complete the session as written. It is not. It is a licence to keep moving while you gather information.

Half two: the 24-hour rule, which is the one that actually decides things

Here is the part that AI plans have no access to at all. How the tissue behaves the morning after tells you more than anything you felt during the run.

Three questions, asked the next day:

1. Is the pain worse than it was before yesterday’s run? Compare like for like: first ten steps out of bed, both mornings.
2. Has it lasted more than 24 hours? An ache that is gone by lunchtime is a different animal from one still present at the same time the following day.
3. Is it worse than it was at the same point after the previous run? This is the trend question, and it is the one people skip.

If the answer to all three is no, yesterday was tolerable and you can proceed with the plan. If any one of them is yes, the plan is now wrong and you reduce.

This is the rule that overrides your app. Not your motivation, not your streak, not the fact that Runna has marked the session as “key”. If Monday’s easy 5K left your plantar fascia worse on Tuesday morning than it was on Monday morning, Tuesday’s session gets cut whatever the app says.

A worked example

Rachel, week six of a Runna half marathon plan, right Achilles that has been grumbling since a shoe change.

The Wednesday breach is the interesting one. She finished the session (legitimately: 4/10, no gait change, held steady through the last rep). But the 24-hour check failed on question two, pain still present at 5pm the next day. So Thursday’s easy run came out, and Friday returned at reduced load. The plan said 34K that week. She ran 28K. The Achilles held.

Compare that with the version where she ignores the Thursday signal, runs the 6K on a tendon that has not settled, and arrives at Saturday’s long run with the pain at 5/10 by kilometre three. That is a four to six week detour, and it started with one easy run that looked harmless in the app.

What to actually change in your app

The traffic light is useless if it only lives in your head, so translate it into the tools you already use.

In Runna, you can move or skip sessions directly. Skip, do not shuffle: dragging Thursday’s run to Friday and Friday’s to Saturday preserves the weekly volume you just decided was too much. When a 24-hour check fails, delete the session. If two checks fail in a week, use the plan-adjustment flow to drop to a lower weekly mileage tier rather than fighting the schedule session by session.

In Garmin Coach, the daily suggested workout responds to your training load and HRV, which is helpful and also not the same thing as responding to your Achilles. The practical move is to log the run as an easy effort and let the load algorithm see the reduction. Watch Training Status: if it flips to Unproductive while you are managing a niggle, that is the algorithm confirming what your morning pain already told you.

On Strava, the number worth checking is not pace, it is cadence and its symmetry across the run. Open the run, scroll to the cadence graph, compare the first kilometre with the last. A drop from 172 spm to 163 spm over an easy 8K, with no hills to explain it, often means you were compensating before you consciously noticed. Pair that with your pain notes in the activity description (write them there, every time, one line: R Achilles 3/10 from 4K, no gait change) and you build the exact dataset the app cannot generate on its own.

If you built your plan with ChatGPT or Claude, do not ask it to “adjust for my sore Achilles” in a fresh message and accept whatever comes back. Give it the actual sequence: pain scores per run, next-morning scores, what you already cut. An LLM will produce a plausible-sounding deload from almost any input. It will produce a useful one only when it has the 24-hour data, because that is the signal that distinguishes a tendon that is adapting from a tendon that is failing.

Where the rules bend

Tendinopathy is the honest exception. Achilles and patellar tendon pain that sits at 2-3/10, warms up during the run and settles within 24 hours is not just tolerable, it is often part of the treatment. Loading the tendon is how it remodels. Complete rest tends to make it worse. So a mid-Achilles ache that is 3/10 for the first kilometre and 1/10 by the third, and is back to baseline the next morning, is a run you should probably do.

Bone is the opposite. Focal pain on a bony landmark, the medial tibia, a metatarsal, the femoral neck or pubic rami, that gets worse as the run progresses, gets no bend at all. It does not warm up, and pain that increases with continued loading is the classic bone stress presentation. Any suspicion here means stop and get imaging, not a reduced session. The traffic light does not apply: it is red at 2/10 if the location and the pattern fit.

Joint pain with swelling, locking or giving way also exits the framework. That is a structural question for a physio, not a load question for your plan.

For the fuller version of how to rebuild once you have stopped, including the graded return protocols and the walk-run progressions that get you back to continuous running, our guide to returning to running after injury covers the stages in sequence.

The honest accounting

An AI plan optimises for plan completion. That is the thing it can measure. It will show you a green tick, a completed week, an adherence percentage, and none of those numbers contain any information about whether your tissue is tolerating the load.

You are the only sensor for that. Two data points, taken every time: the number during the run, and the number the next morning. Write them in the Strava description so you have a record you can actually read back over three weeks.

Then next Tuesday, when the watch buzzes with 6 x 800m and your Achilles was 3/10 getting out of bed and still 3/10 at 4pm, you already know the answer. You do not have to negotiate with yourself in the kitchen at 6:40am. The rule decided it yesterday.