How our numbers work.
Every number in Innoraze can be traced back to the data it came from and explained in plain words. This page sets out how we get from raw match data to the figures your staff see.
Expected goals, in one picture.
Expected goals (xG) is the chance that a shot becomes a goal. Distance matters, and so does how much of the goal the shooter can see. Choose a spot on the pitch to see what a shot from there is worth.
This is a simplified version: a right-footed shot in open play, using only distance and angle. The full model also uses the body part, the type of pass before the shot, the situation and, where tracking data is available, the position of the defenders and goalkeeper.
0.30 xG
11.0 m from goal, 37° of the goal in view
Scored about 3 times in 10.
It starts with clean data.
Most errors in football analysis come from the data, not the model. Before anything is calculated, every match goes through the same checks.
- 1
Match the players
Each provider names and numbers players differently. We map every player to one identity across event, tracking and GPS data.
- 2
Line up the clocks
Event times, tracking frames and video are synchronised to the same match clock, so a pass in the data is the same pass on screen.
- 3
Catch the gaps
We flag missing tracking frames, duplicated events and impossible values, such as a player moving faster than anyone can run.
- 4
Follow corrections
When a provider corrects its data after a match, reports update and we tell you what changed.
The models, in plain words.
These are the numbers that appear most often in Innoraze reports, and how each one is built.
- Expected goals (xG)
- Built from a large history of shots, each labelled with whether it was scored. The model learns how location, angle, body part, the pass before and defensive pressure change the odds. We check that it is well calibrated: shots it rates at 0.2 should go in about one time in five.
- Expected threat (xT)
- The pitch is split into a grid of zones. Each zone gets a value: how likely a team with the ball there is to score within the next few actions. A pass or carry that moves the ball to a more valuable zone adds expected threat.
- PPDA
- Passes allowed per defensive action, a measure of pressing. It is the number of passes the opponent makes in their own 60% of the pitch, divided by your tackles, interceptions, challenges and fouls in that same area. Lower means a harder press.
- Field tilt
- Your share of the touches in the final third, out of both teams’ final-third touches. Above 50% means you spent more of the match in the opponent’s end than they did in yours.
- Physical output
- From GPS or tracking data. High-speed running counts distance above 19.8 km/h and sprinting counts distance above 25.2 km/h, the thresholds most clubs already use. You can set your own.
- Player similarity
- Each player is described by a profile of per-90 numbers in their position. Recruit compares those profiles, weighted by what matters for the role you are searching for, and shows which numbers drive each match.
How we check our work
Tested on past seasons
Every new model version is run against previous seasons before it reaches a club, and we compare it with the version it replaces.
Every number is traceable
Click a number in a report to see the events behind it and, where you have video, the clips.
Definitions you can share
Every metric comes with a one-line definition your coaches can use with players.
Your questions, answered
If a number looks wrong to your staff, we look into it and explain what we found.
Help shape Innoraze.
We’re building Innoraze now. Join early access to see each part first and tell us what your staff need from it.
Or follow the build
We write up how Innoraze is built, from the cloud platform to the match models. Leave your email and we’ll send you each write-up.
Get build updates