HYROX race time predictor

What this tool assumes

• Your road-race fitness translates to HYROX running the way it did for the athletes behind the ratio curve — a median for your division at your own road pace, applied to a fresh station test the same way it would a race result. The percentile matching absorbs some of that on average, but your mileage will literally vary.
• The reference data is 93,311 results for the finish-time matching — 39,247 solo and 54,064 doubles teams, matched only within their own category — and 1,980 road-race pairings across 9 divisions, quantile-binned by road pace, for the run-pace lookup. 11 venues so far — Bangkok Aug 2026, Cape Town 2026, Cardiff 2026, Gdansk 2025, Glasgow 2026, Istanbul 2026, Jakarta 2026, New York 2026, Perth 2026, Sydney 2026 and São Paulo 2025 — pooled rather than kept apart. Course quirks (a short opening run leg, lane traffic) ride along in the data, and so does venue-to-venue field strength, which varies by 13–20 minutes at the median.
• The uncertainty band isn't decoration — it's the measured leave-one-out error range on the 93,311-result finish-time dataset. Predictions clamp at the edges of the road-pace range we measured for your division (roughly the 5th–95th percentile).
• Adding station times cuts both ways, and the two race types differ. For all four solo divisions, more inputs measurably improve the prediction. For all five doubles divisions the opposite held in validation: the band got wider, not narrower. Doubles station times are a noisier predictor than run pace — how a team divides reps varies far more than one athlete's pacing — so for doubles, extra inputs move the estimate toward your team's profile without making it more certain.

Got a prediction? The pace calculator turns it into segment-by-segment splits. The station guides cover the movements behind each input, and the race calendar tells you how long you have. When you want the whole plan built for you, that's what customized plans will be for.

FAQ

How does the Roxmetric race predictor work?

You pick a division and enter a recent 10K or Half Marathon time, plus any fresh station test times. Your road time becomes a road pace, and we look up the median Pace Ratio (HYROX run pace ÷ road pace) for athletes who race at that same road pace, in your division. Apply that ratio to your pace to get a predicted run pace, then blend it with any stations you provided against the distribution of real HYROX finish times for your division (percentile matching) to predict your overall finish time.

Why match by road pace instead of by finish time?

We tried finish time first — bucket athletes by their own HYROX result, look up the typical ratio for that bucket. It has a real flaw: predicting a finish time from a ratio, then using that finish time to look up the ratio, is circular, and our finish-time dataset has very few slow/recreational finishers, so the percentiles nearly collapse at the slow end and distort predictions there. Road pace has neither problem — it's something you already know before you predict anything, and we have plenty of data across the full pace range.

Why a 10K or Half Marathon time instead of a "1 km repeat pace"?

Because almost nobody can reliably test a repeatable 8 × 1 km pace on their own, while a recent 10K or HM result is something most athletes already have. We derived the ratio curves from 1,980 real paired results — the same athletes' road-race and HYROX results, after excluding ratios below 0.6 or above 1.4 (a sign something unusual happened — injury, cramping, racing doubles with a much slower or faster partner — not just a fast or slow athlete) — so the input you're asked for is also the one we can predict from accurately.

What if I only have a 10K time, not a Half Marathon?

Use it — we convert it to a Half-Marathon-equivalent pace with the Riegel formula (a standard running-science conversion) before matching, since our curves are fitted on Half Marathon pairings (1,868 of them, vs 112 for 10K — too thin to bin by pace on its own).

What data is the prediction based on?

Two datasets: (1) 93,311 complete performances — 39,247 solo and 54,064 doubles teams (19,788 Open Men, 14,678 Open Women, 3,017 Pro Men, 1,764 Pro Women, 14,029 Men Doubles, 18,983 Women Doubles, 17,638 Mixed Doubles, 2,044 Pro Men Doubles, 1,370 Pro Women Doubles) — for the percentile matching, and (2) 1,980 road-race-to-HYROX pairings across 9 divisions (Open, Pro, Doubles; Men, Women, Mixed), quantile-binned by each athlete's own road pace, for the road-race-to-run-pace lookup — after excluding the ratio anomalies described above. Doubles rows are team results and are only ever matched against the same doubles category. 11 venues so far (Bangkok Aug 2026, Cape Town 2026, Cardiff 2026, Gdansk 2025, Glasgow 2026, Istanbul 2026, Jakarta 2026, New York 2026, Perth 2026, Sydney 2026 and São Paulo 2025), pooled deliberately: their fields finish 13–20 min apart at the median, so pooling widens the bands 9–24% for doubles versus a single-venue fit. A one-venue model validated on its own field reads as sharper than it is for the races we hold no data for.

What happens if my division has too little data for its own curve?

Pro Women is the one case right now: only 10 athletes, too few for a stable curve of their own, so Pro Women currently borrows the Open Women curve — the tool tells you this directly when you select it. Every other division (Open Men, Open Women, Pro Men, and all three Doubles categories) has enough data for its own fitted curve.

How accurate is the prediction?

We measured both stages instead of guessing. Finish time: re-predicting every athlete in the 93311-athlete dataset from their own numbers (leave-one-out), 80% of Open-division athletes land between about 9.6 min fast and 5.7 min slow with typical inputs, and between 10.7 min fast and 4.6 min slow with a full input set. Note the skew: the band is wider on the fast side, meaning the model more often predicts a time you beat than one you miss. Pro fields are smaller and spread wider still. Run pace: comparing individual athletes against their own road-pace bin's median, run pace deviates by roughly -65 to +41 seconds per km 80% of the time — that's how much even people who race at the same road pace vary in how their HYROX running holds up. Anyone quoting your finish to the second is guessing — the range is the honest answer.

Why does the road race count more than a station time?

Because running is 8 of the 16 segments and roughly half of total race time in the data. The prediction weights your predicted run pace accordingly (8× a single station).

I’m racing HYROX Doubles — is this for me too?

Yes. Switch to Doubles mode for a predicted run pace, plus a predicted total finish time once you fill in your team's station and Roxzone times. Doubles athletes rest while their partner works a station, so at the same personal road-race fitness they hold a measurably faster run pace than solo athletes do, and the gap widens the slower you are (more of your race is stations, so the rest matters more). We confirmed this directly in the data rather than assuming it: at matched road pace, men's doubles ratio runs 1–6% lower than solo across most of the range, growing to roughly 6% at recreational paces.

Does Doubles work the same way as Solo now?

Yes. Doubles used to require all 9 station and Roxzone fields because we had no doubles reference data and the total was plain addition. We now have 54,064 complete doubles team results (14,029 Men Doubles, 18,983 Women Doubles, 17,638 Mixed Doubles, 2,044 Pro Men Doubles, 1,370 Pro Women Doubles), so Doubles gets the same percentile matching as Solo: a prediction appears from your road-race time alone and tightens as you add station times, and you get a field position against real teams in your category. Two caveats. Doubles numbers are the team's, matched only against the same doubles category — never against solo. And the predictor's Doubles categories stop at Men, Women, and Mixed, because the road-pace-to-run-pace curve is only fitted for those three; Pro Doubles has station and finish data but no run-pace curve yet, so it appears in the pace calculator rather than here.

Why enter my team’s station time instead of one partner’s fresh test time?

Because a fresh solo test time only represents what one person can do untired — HYROX Doubles teams often split a station's work between partners, or have one partner rest fully while the other goes all-out fresh. Either way, the number that matters for your total is what your team actually produces at that station, not an individual benchmark. If one partner is doing the whole station solo while the other rests, their fresh test time is a reasonable stand-in — just enter it as the team's number.

What data is the Doubles run-pace estimate based on?

1,289 doubles road-race-to-HYROX pairings across the three team categories (Men, Women, Mixed Doubles), quantile-binned by each athlete's own road pace — the same method and the same underlying dataset as the Solo curves, just filtered to doubles divisions.