Overstriding: What It Really Looks Like on Video

Your watch does not know whether you overstride. Neither does Runna, nor Garmin Coach, nor the ChatGPT plan you built in March. What they have is a cadence number, maybe a ground contact time, maybe a stride length figure, and a threshold someone chose. From those they print a verdict. That verdict is a guess dressed as a measurement, and a lot of runners are now doing cadence drills to fix a problem they do not have while ignoring one they do.

The fix is not another app. It is ninety seconds of slow-motion footage and about four minutes of scrubbing through it. This post is about how to tell if you overstride by looking at where your foot lands relative to your knee and your hips, in frames, with actual numbers you measure yourself.

What the apps are actually measuring

Start with what your data can and cannot say. Garmin’s running dynamics (wrist-based on a Forerunner 265/965 or Fenix 7 Pro, or via an HRM-Pro strap) give you cadence, ground contact time, vertical oscillation, vertical ratio and a figure labelled “stride length”. That last one is worth understanding, because it drives most false positives.

Take a 5:30/km easy run at 158 spm. Speed is 1000 ÷ 330 = 3.03 m/s. Steps per second is 158 ÷ 60 = 2.63. Divide: 1.15 m. Open Garmin Connect and you will see roughly 1.15 m in the stride length field. So despite the name, it is distance per step, and it is pure arithmetic from pace and cadence. It contains zero information about where your foot was when it hit the ground.

Here is why that matters. Most of a 1.15 m step happens behind you: hip extension, push-off, flight. The bit in front, the reach, might be 18 cm of it. Two runners can post identical 1.15 m steps at identical cadence, and one of them lands with a near-straight leg 34 cm ahead of the hip while the other lands 16 cm ahead with a bent knee. No wearable on the market separates those two. A video does it in one frame.

The three things you are actually looking for

Overstriding is a position, not a length. Three checkpoints, in order of how much they tell you:

1. Foot relative to knee. At the instant of first contact, is the ankle in front of the knee? Draw a line down the shin. If the shin tilts backwards (toes further forward than the kneecap), you are reaching. If the shin is vertical or tipped slightly forwards, you are not, regardless of how long the step is.

2. Knee flexion at contact. A near-locked knee at touchdown has nothing left to absorb with. Recreational runners typically land with something in the region of 12° to 20° of knee bend at easy pace. Under about 10° and the leg is behaving like a pole.

3. Time from contact to the foot being under the hip. This is the braking phase, and you can count it in frames without any geometry at all. More on that below, because it is the single most useful thing in this post.

Filming footage that can answer the question

Phone on a tripod or wedged in a shoe against a kerb, lens perpendicular to your direction of travel, at roughly hip height, 3 to 4 metres back. Height matters: film from waist level and a flat foot-plant looks like a heel strike from above. Shoot at 240 fps if your phone offers it (iPhone 8 and later at 1080p, most recent Pixel and Galaxy models at 240 in “slow motion” mode). At 240 fps you get a frame every 4.17 ms. Running at 3.03 m/s, that is 1.26 cm of travel per frame. At 120 fps you get 2.5 cm per frame, which is workable but coarse.

Two practical details that make the footage measurable rather than decorative. Put a strip of electrical tape on your shorts over the greater trochanter, the bony bump on the outside of your hip. That is your vertical reference, and eyeballing it through a baggy waistband is how people get 8 cm errors. Then film yourself standing still in the same spot holding a metre stick or standing beside a 1 m strip of tape on the ground, so you can calibrate distances later.

Run through the frame three or four times at genuine easy pace, then twice at your 10K effort. Do not run at “being filmed” pace. If you fix your form for the camera you have measured nothing.

For the actual scrubbing, Kinovea is free on Windows and has line, angle and calibration tools, which is the only thing here that gives you real centimetres. On a phone, OnForm and Hudl Technique both let you step one frame at a time and draw over the image. Ochy will produce automated numbers from phone video and is genuinely useful, though check its skeleton keypoints against the frame before you believe its ankle position: they drift on loose shorts and dark shoes.

The frame-count method

This is the version that needs no protractor. Find the first frame where any part of the shoe touches the ground, call it frame 0. Step forward until the tape on your hip sits directly above your ankle. Count.

Ground contact time at easy pace for most recreational runners runs about 240 to 290 ms. At 240 fps, 250 ms of contact is 60 frames total. Getting to midstance should take somewhere near 40 to 45 percent of that, so roughly 24 to 27 frames.

Easy run, 5:28/km, 240 fps, left foot, three strides

            contact → midstance     total contact
stride 1         38 frames             79 frames
stride 2         36 frames             77 frames
stride 3         39 frames             81 frames

38 frames @ 4.17ms = 158 ms spent braking
total contact 79 frames = 329 ms

Thirty-eight frames to get the hip over the ankle is a long time to spend decelerating. That runner’s Garmin was reporting 329 ms GCT, and Garmin’s own colour bands put that in the “purple” slowest zone, but the watch had no way to tell them why. The video did: the foot was landing a long way out front and the body had to travel over it.

Two runners, both flagged

Runner A, 1.84 m, marathon block on a Runna plan, cadence 156, told by a form-analysis app that their stride was “too long”. Measurements from Kinovea, calibrated against a 1 m floor tape:

Runner A is not overstriding. They are tall, they have long legs, they run with a low cadence because they cover 1.28 m per step, and the foot is under a bent knee when it lands. The app saw 156 spm and 1.28 m and fired. Nothing in the footage supports it.

Runner B’s app said nothing at all. Cadence of 174 sails past every threshold, and a 1.02 m step looks conservative. But the ankle is 31 cm in front of the hip with a nearly straight knee and the shin tipped backwards, and the body needs 154 ms to get over the top of it. That is a genuine reach, hidden behind a cadence number that looks textbook. Short, fast, reaching steps are entirely possible, and they are the case automated flags miss most often.

If the video says yes

Cadence still helps, but use it as a lever rather than a target. Heiderscheit’s 2011 work found that increasing cadence 5 percent measurably shortened stride and reduced loading, with 10 percent doing more. Applied to Runner B: 174 spm at 3.03 m/s is 1.02 m per step. Add 5 percent to get 183 spm and the step becomes 0.99 m. Add 10 percent for 191 spm and it is 0.95 m. Seven centimetres off the step length, most of which comes off the front.

Run the same filming protocol four weeks later before you decide it worked. The number to watch is not cadence, which you already know changed because you made it change. Watch the ankle-to-hip gap and the frame count to midstance. If Runner B’s 37 frames comes down toward 28 and the gap drops under 25 cm, the change reached the ground. If cadence rose and the frame count did not move, all that happened is a faster shuffle, and the reach is still there.

There is also the question of what else in your footage deserves a look while you have it open, since foot position rarely misbehaves alone. Hip drop, cross-over and trunk lean all show up in the same clips, and the broader walkthrough of AI gait and form analysis from phone video covers how to read them from the same three passes.

One last thing about the plan you are following. An AI-built block that ramps volume 15 percent a week is a load problem, and it will not be solved by a form cue. But if your video shows a 30 cm reach with a straight knee, every one of those extra kilometres is being paid for with a braking impulse you could have measured on your phone in an afternoon, and no red-flag badge was ever going to tell you which one you were dealing with.

So: charge the phone tonight, put the tape on your shorts, and get three strides at 240 fps before the next easy run turns into another guess.