HELM
The method

From your sessions to a model of you.

HELM doesn’t score your run against a population. It builds a measured picture of your own running, what you can do, what kind of tired you are, and what is limiting you, and it tells you how sure it is. Here is how that works, without the parts it keeps to itself.

The pipeline

Four stages, one direction.

Every session moves through the same path. The first two stages learn what is true about you; the last two act on it. Nothing skips ahead of the evidence.

01Ingest
Read everything
Your watch, your phone, your sleep and recovery, and the terrain you ran, through one secure link. Every channel arrives with a data-quality flag, so gaps are known, not guessed.
devices · terrain · health
02Calibrate
Find your thresholds
HELM learns your thresholds from your own sessions, not a textbook: the pace you hold all day, hard, and flat out, in both pace and heart rate. They anchor everything downstream and sharpen as it sees more of you.
your own thresholds
03Measure
Update the picture
Each new session folds into a continuously updated read on your capability, your freshness, and the load on each tissue. Every value comes as a best estimate with a 90% confidence interval.
mean + 90% interval
04Guide
Name the limiter
HELM names the one system capping you this week and what kind of tired you are, then turns it into plain calls: push, hold, or protect. When it is unsure, the guidance gets more conservative, not louder.
the limiter · explainable
On uncertainty

Uncertainty, not false precision.

A single number with no error bar is a guess wearing a uniform. Every capability HELM shows carries a 90% confidence interval, the range your true value almost certainly sits inside.

Pace · min/km
Endurance
5:57
Performance
4:38
Peak
3:12
5:36
4:58
4:21
3:44
3:06
5:536:01
4:344:42
2:553:29

The orange tick is the best estimate; the sky band is the 90% range. Peak pace sits wide here because HELM hasn’t seen a hard, short effort lately. Run one and it tightens.

The band widens when evidence is thin, a missed week, a new device, a season opener, and narrows as sessions accumulate. You watch your own data earn confidence.

When the interval is wide, HELM holds back. It won’t push a hard session off a number it isn’t sure of. The guidance gets quieter and more cautious until the data backs it up.

What kind of tired

Three kinds of freshness.

Aerobic, anaerobic, and muscular fatigue are not interchangeable. Each recovers on its own clock and limits a different kind of session, so HELM tracks them apart instead of flattening them into one readiness score. That split is what tells you what you can still do today.

Aerobic

Cardiovascular endurance: the heart, lungs, and blood that deliver oxygen to working muscle, the base of all steady running. When it dips, easy efforts feel harder than they should and your breathing climbs early. It recovers the fastest of the three and underpins everything else. A low reading means hold back the volume. Consider a strength day if your muscular system is fresh.

Anaerobic

Anaerobic power: the glycolytic system behind intervals, surges, and the closing kick, the gear you reach for above threshold. When it dips, speed work falls flat. It recovers more slowly than your aerobic system. A low reading means skip the intervals today. If your muscular system is fresh, a day in the mountains loads aerobic and muscular while letting anaerobic power recover.

Muscular

Neuromuscular force: the muscle and tendon that produce force in every stride and absorb impact and the eccentric braking of descents. When it dips, your legs feel heavy and downhills punish you. It is the slowest of the three to recover. A low reading means avoid the high-impact work: fast running and steep descents. Consider some easy volume if your aerobic system is fresh.

Where load lands

Tissue by tissue.

“Training load” as one number hides where the strain actually went. HELM tracks load against capacity for each tissue and reads out a load ratio, recent load over what that tissue can absorb. Around 1.0, load and capacity are balanced; it is when the ratio climbs and stays high that you are outrunning what the tissue can absorb.

Shinhigh risk
1.56
1.411.68
rising · peak 1.62
undertrained < 0.80Capacity to spare. Room to build without much risk.
healthy 0.80 – 1.30Load and capacity in balance. The productive zone.
at risk 1.30 – 1.50Outrunning capacity. Watch the trend before adding.
high risk > 1.50Demand well past readiness. HELM steers load away from here.

Each tissue answers to different terrain: a steep descent loads the knees and IT band through eccentric braking, a long climb the calves and glutes. So a descent-heavy week can light up the knees while the Achilles stays quiet, and HELM tracks each on its own clock.

Distance lies, terrain doesn’t

Terrain-adjusted load.

A kilometre is not a kilometre. A steep descent pounds your legs far harder than the same distance on the flat, so HELM converts every session into equivalent kilometres, the flat-ground distance that would have cost your body the same.

Flat kilometre

Even ground · 1.0 km run

grade · 0%
distance · 1.0 km
1.0 eq-km

Steep descent

−14% grade · 1.0 km down

grade · −14%
distance · 1.0 km
1.9 eq-km

Same GPS distance, nearly double the load. A downhill back half pounds the knees and quads with eccentric braking, the damage that leaves legs sore for days, which is why HELM counts the descent for what it actually costs.

Plain dealing

What HELM won’t claim.

Knowing the edges of a model is part of trusting it. These are the lines HELM will not cross.

It doesn’t diagnose injuries. HELM flags load and risk on tissues. It is not a clinician and won’t tell you what is wrong with your body.
It doesn’t promise a PR. It tells you where you stand and what is limiting you. The race still depends on taper, course, and the day.
It doesn’t fake what it can’t measure. No heart-rate variability from your watch means none in the model. HELM leaves it out rather than invent it.
It doesn’t smooth the numbers for comfort. You see exactly what was measured, intervals and all, even on the days it looks ugly.
It doesn’t grade you against other people. No leaderboards, no population percentiles. The only baseline is your own history.
It doesn’t hand you one readiness score. Freshness is split by type, and any headline number comes with the limiter, never pretending to be the whole story.

Join the waitlist.

HELM is opening to trail and road runners in small batches. Drop your email and you’ll hear from it when your spot opens.