How a HYROX time predictor works

A HYROX® time predictor estimates your finish by adding up the race’s parts, eight 1 km runs, eight stations and the roxzone between them, from numbers you can measure today, and an honest one gives the range your finish could land in rather than a single time. The runs weigh most: in the 2025/26 season they were 47% of the average Open man’s finish and 48% of the average Open woman’s, against 44% and 43% for the eight stations and about 9% for the roxzone.

Below: how a predictor builds a finish step by step, which of your numbers move it most, why the answer should be a range, and what Rivo’s free Number shows and doesn’t do. The race numbers come from 702,095 public HYROX results from 98 venues, each with all 16 splits.

How does a HYROX time predictor work?

It builds the race part by part and adds the parts up. Each part comes from a number you measured or, where you have none, from athletes in past results who match you, and the parts you didn’t measure are the least certain.

1. A pace for the eight runs. It starts from a running test, usually a recent 5 km time, and slows it for racing 1 km at a time with a station after each. The runs don’t come out even: Open men who finished in 1:25:00–1:29:59 averaged 5:10 per km (8:20 per mile) on runs 2 to 7, then 5:53 on run 8, after the sandbag lunges.
2. A time for every station. Your own station times where you have them, made slower for doing them mid-race on tired legs; for the rest, the times of athletes who run as fast as you. The median Open man spent 38:24 on the eight stations and the median Open woman 41:11.
3. The roxzone. Every move between a run and a station is on the clock: the median Open man spent 7:27 in the roxzone and the median Open woman 8:05.
4. The sum. Eight runs, eight stations and the roxzone make the finish.
5. A range around it. Each part could come in faster or slower on the day. The range adds that doubt up, and it is widest when most of the parts were estimated rather than measured.

Is a HYROX finish time calculator the same as a predictor?

No. A finish time calculator splits a goal you choose into targets for every run and station; a predictor works forward from what you can do today and tells you which goal is realistic. Use the predictor to set the goal, then pace it with the splits real finishers of that time ran, in the splits guide or how to pace HYROX.

Which numbers matter most in a HYROX prediction?

The ones where athletes differ most. Between the top 10% of Open men and the slowest 10%, the eight runs were 22:01 apart and the eight stations 22:08; among single stations the wall balls spread them most, by 7:00, and the SkiErg least, by 0:57.

Part	Open men	Open women
Wall balls 100 reps	7:00	6:26
Burpee broad jumps 80 m	4:45	5:47
Sandbag lunges 100 m	4:09	4:01
Sled pull 50 m	3:36	4:02
Sled push 50 m	2:18	1:56
Farmers carry 200 m	1:34	1:30
Row 1000 m	1:13	1:22
SkiErg 1000 m	0:57	1:12
All eight runs	22:01	23:45
All eight stations	22:08	22:51
Roxzone (transitions)	7:41	7:37

The 90th percentile minus the 10th, for each part on its own. HYROX season 2025/26: 272,693 results with all 16 splits.

Two things follow. The run estimate moves a prediction most: the runs are the largest part of the race and spread the field as widely as all eight stations together. And because the stations spread it just as widely, a prediction from a run time alone has to assume you are typical on all eight of them. Rivo’s Number asks for three numbers: a 5 km time and your sled push and sled pull times, the two heaviest loads in the race at 152 kg (335 lb) and 103 kg (227 lb) for Open men and 102 kg (225 lb) and 78 kg (172 lb) for Open women, sled included.

Why does a HYROX predictor give a range instead of one time?

Because the same athlete can finish minutes apart on two race days, and one time hides that. The course alone moves finishes: comparing the 44,243 athletes who raced at two or more venues in 2025/26 with themselves, those who raced at Warsaw finished 5.5% faster there than at their other venues, and those who raced at Puebla 11.9% slower, about 4:57 faster and 10:43 slower on a 90-minute race.

More doubt comes from three places:

• Run 1 and run 8. Every venue lays out its start and finish differently, so these two runs swing from course to course while runs 2 to 7 keep their shape. The running pace guide shows by how much.
• Stations you haven’t timed. An estimate from athletes who run like you is right for the typical athlete, not for you.
• A first race. It has no past result to check the prediction against.

A range from the 10th to the 90th percentile reads like this: the model expects one finish in ten to beat the fast end and one in ten to miss the slow end. The more of your race is measured rather than estimated, the narrower it gets. The venue guide ranks every course by what it did to the same athletes’ times.

What does Rivo’s Number show?

A predicted finish time with its p10–p90 range and a label for how sure it is, and under it the stations costing you the most time, in seconds. It is free, needs no account, and starts from a 5 km time and your sled push and sled pull times, or from a published HYROX result you claim by name.

• The finish, its range and its label, always together: never a time on its own.
• What it’s costing you, in seconds: the parts of your race that cost you the most time.
• Where you sit in your age group and division, from real race results: your running as a percentile of your field, your stations as how many of ten finish-group medians you beat.
• A card to share with the prediction on it.

From three numbers it builds the finish the way the steps above describe: the runs from your 5 km time and the way real finishers’ runs slow through a race, the sled push and sled pull from your own times, the other six stations from athletes in past results who run as fast as you, and the roxzone from athletes with a similar finish. The range is wider for every part it had to estimate; claim a past race and all 16 splits come from your own result.

What it doesn’t do: the free Number never reads Apple Health, Strava or a watch. It uses what you type, or your published result if you claim one. Measured training moves it only inside the paid Plan, which has a 14-day free trial; what is free and what isn’t is set out in free HYROX apps.

How accurate is a HYROX time predictor?

Only a test on races the predictor never saw can tell you, and the range is the part to test. Ask any predictor three things: how many real finishes it was checked against, how far off it was on average, and how often the real finish landed inside its range.

A prediction with no range can’t be checked that way, because it never says how wrong it could be. This page quotes no accuracy figure for Rivo’s Number; every Number it gives carries its range and its label.

Can you predict a HYROX time before your first race?

Yes, from benchmarks: a 5 km time gives the runs, and timed sled pushes and pulls give the two heaviest stations. Expect a wider range than a repeat racer gets, because nothing from a past race narrows it.

Public results don’t mark anyone’s first race, so no results table can say what first-timers run. The fair benchmark is the median for your division, sex and age group on what’s a good HYROX time, and the splits guide shows what real finishers of each five-minute bracket ran on every segment.

Where do these numbers come from?

From 702,095 public HYROX results of the 2025/26 season, each with all 16 splits; how Rivo’s Number is built is described above, and no accuracy figure is quoted.

• Source: 702,095 public HYROX results from 98 venues in the 2025/26 season, read through pyrox-client, an independent open-source client, and loaded on 2 September 2026. Only results with all 16 splits and a finish between 40 and 300 minutes count.
• Spreads are the 90th percentile minus the 10th of each part on its own, so they don’t add up across parts. Shares are of the average finish.
• The venue comparison uses Open and Pro singles who raced at two or more venues in the season: each venue’s effect is what it did to their finish against their own finishes elsewhere. Championship and invitational races are left out of it.
• Nothing here describes one athlete: only counts, percentiles and averages of groups, never a name or a single result.