The First Two Weeks Back: Ground Rules Before Any Plan

Your calf went pop six weeks ago. The physio has signed you off, the pain is gone when you walk, and Runna is sitting on your phone offering to rebuild your half marathon block starting Monday. You know what happens next if you tap accept: a 4km easy run on day one, a 6km on Wednesday, and a progression run at the weekend, because that’s where the plan picks up from your old fitness.

Don’t tap accept.

The single most useful thing I can tell you about when can i start running after injury is that the answer is never a date. It’s a set of tests you either pass or fail, on the day you take them. Every AI coaching tool available in 2026 — Runna, Garmin Coach, TrainingPeaks’ AI builder, a plan you talked ChatGPT into writing — shares the same blind spot. They schedule forward from a start date you give them. They have no idea whether your soleus can take 30 single-leg calf raises without complaining, and they will never ask.

So you ask. The first two weeks back are not training. They’re a qualifying round.

The tests, and what passing actually looks like

These are the gates. They aren’t clever, and they aren’t mine originally: they come from the loading criteria used in tendinopathy and bone stress rehab, translated into things you can do on a kitchen floor.

Gate 1: Walk 30 minutes, pain-free, same day and next morning. Not “manageable.” Not 2/10. Zero, and no next-day flare. If a 30-minute walk produces stiffness the following morning that wasn’t there before, your tissue is still reactive and any run will make that worse. This one is non-negotiable and most people skip it because walking feels beneath them.

Gate 2: 30 single-leg calf raises on the injured side, full range, no pain. For calf, Achilles, soleus, plantar fascia and most shin problems this is the gate. Straight knee for gastroc, then repeat with the knee bent about 30 degrees for soleus, which is the muscle that actually fails in most runners. Compare sides. If your good leg does 32 and your injured leg does 18 before the burn turns to pain, you are not two weeks from a half marathon plan.

Gate 3: 10 single-leg squats to roughly 60 degrees knee bend, no pain, knee tracking over the middle toes. This is the gate for knee problems, ITB, hip and glute-related trouble. Watch the pelvis in a mirror. If the opposite hip drops noticeably, the load is going somewhere you don’t want it.

Gate 4: 20 single-leg hops, on the spot, pain-free, then 20 more the next day with no flare. This is the impact gate and it’s the one that matters most, because running is a hopping sport. Roughly 2.5 times bodyweight goes through your leg on each foot strike. Hopping tests that at a lower dose in a way you control.

Gate 5: The 24-hour rule. Whatever you do, nothing new hurts the following morning. Not “a bit tight.” Nothing new.

Pass all five and you’ve earned a run-walk. Fail any one of them and you have a rehab session to do instead, which is not a setback, it’s information.

What the first two weeks actually contain

Here’s a real fortnight, for a runner whose gate tests all passed on a Sunday. Assume they were running 45km a week before a calf strain in July.

Twenty-five minutes of running by the end of week two. That’s it. A runner who was doing 45km a week is now doing roughly 12km total across a fortnight, and if you put that into any app as your “current volume” it will look like an error.

Note what’s missing. No pace target. No heart rate zone. No hills, no strides, no “easy run with 6 × 20s pickups.” Pace is an output in these two weeks, not an input. If your easy pace was 5:30/km and it’s now 6:40/km, that is the correct pace, because it’s the one that came out of running at an effort where you can hold a full conversation.

Note also the off days. The run days are alternating, always, for the full fortnight. Tendon and bone tissue adapts on a slower clock than your cardiovascular system, and the 24 to 48 hours after a run is when the remodelling happens. Back-to-back days in week one is the single most common way people re-injure.

Reading your own data instead of trusting the plan

You already have the tools to audit this. Use them properly.

Strava’s weekly mileage graph is where you check the ramp. Open your training log, switch to the weekly view, and compare the bar for the week you’re about to run against the previous four. The rule I use with runners coming back: no week increases total running time by more than about 30% over the previous week while you’re under 30 minutes of total running, and no more than 10% once you’re above it. Week one in the table above is 32 minutes. Week two is 63. That’s a 97% jump, which sounds reckless and isn’t, because the absolute numbers are tiny. The percentage rule only starts to bite once the base is meaningful.

Garmin’s Training Status and Acute Load will scream at you during these two weeks. Ignore both. Garmin’s acute load model is built around a 7-day exponentially weighted average against a 28-day chronic load, and after six weeks off your chronic load is near zero, so a 20-minute jog will show as “Overreaching” or push your Training Status to “Unproductive.” That’s a modelling artefact, not a verdict. What you should read on Garmin is HRV Status and resting heart rate. A resting HR that’s 5+ bpm above your baseline for two consecutive mornings, or an HRV that drops out of your personal range, means take the extra day.

Cadence, in the lap data. Pull up a run in Strava or Garmin Connect and look at cadence per lap rather than the run average. What you’re looking for is drift: if you start at 168 spm and you’re at 156 by the last rep, you’re fatiguing into a longer, harder-braking stride. That’s the point where injuries happen. Stop the session there, and make the next one shorter.

Here’s what a passing session looks like when you pull the lap splits:

Lap   Type   Time    Avg Pace   Avg HR   Cadence
1     Run    2:00    6:38/km    138      170
2     Walk   1:00    9:50/km    118       -
3     Run    2:00    6:41/km    142      171
4     Walk   1:00    9:44/km    121       -
5     Run    2:00    6:35/km    145      169
6     Walk   1:00    9:52/km    123       -
7     Run    2:00    6:44/km    147      170
8     Walk   1:00    9:41/km    124       -

Pace stable across reps, HR drifting up by a normal 9 bpm, cadence holding within 2 spm. That’s a green light for the next session. If lap 7 had come in at 7:10/km with a cadence of 158, that’s a session you should have cut at lap 5.

Making the AI plan behave

You don’t have to abandon the app. You have to lie to it, slightly, and then override it.

For Runna, don’t use the “returning from injury” onboarding and then accept the suggested race date. Set your race date far enough out that the plan’s first week lands at or below your current volume, then manually swap the first fortnight’s sessions for run-walks. Runna will let you mark sessions complete without doing them as written; your plan adherence score will look bad and that number means nothing.

For Garmin Coach, the daily suggested workouts adapt off your recent load, which sounds ideal and isn’t, because it adapts upward fast when you’re starting from zero. Turn off daily suggested workouts entirely for the first two weeks and build the run-walk sessions manually in Garmin Connect as interval workouts. It takes ten minutes once.

For a ChatGPT-generated plan, the failure mode is different: it will confidently produce a twelve-week structure with named workout types and no mechanism at all for reading how you responded. If you’re using one, give it a rule it can’t wriggle out of. Something like: “Do not schedule consecutive running days. Cap week one at 35 minutes total running time. Every session is run-walk until I tell you I’ve done three consecutive sessions with stable pace and cadence.” Then re-prompt weekly with your actual completed sessions rather than letting it run ahead of you.

The deeper structure of how you get from 25 minutes of run-walk back to a genuine training block is covered in our return to running after injury guide, which walks through the full progression past week two.

The bit nobody wants to hear

If you fail a gate, the honest response is to go back to loading. A runner who can only manage 18 single-leg calf raises needs three to four weeks of heavy, slow calf work: three sets of 8 to 12 reps, with enough weight that the last two are genuinely hard, three times a week. That builds tendon capacity in a way jogging never will. You’ll come back faster having done it, which is an annoying thing to be told and also true.

And if you pass all five gates, run your eight minutes on Monday, and wake up Tuesday with the same calf niggle you had in July? You’ve learned that the gates weren’t sensitive enough for you specifically. Drop back, add a week of loading, retest. That’s not failure. That’s the system working: you found out with eight minutes of running instead of finding out in week five of a plan, at 38km a week, with a race entry you can’t defer.

Most reinjuries I see aren’t caused by people doing something obviously stupid. They’re caused by a plan that was reasonable for the runner they were in June being applied to the runner they are in September, by software that has no way of knowing the difference. Two weeks of tests costs you almost nothing. Getting it wrong costs another six.