01
Matrix
The independent layer between your tracking providers and your analysis.
It turns optical tracking, wearable GPS and event data from different providers into one harmonised dataset, in three stages: Normalise → Reconcile → Fuse.
The operating system for football tracking data.
One dataset from every tracking provider.
Matrix brings optical tracking, wearable GPS and event data into one harmonised, quality-scored dataset. It is independent of every provider, so you stay free to use any of them.
Ponta
What Ponta is
Ponta builds two products for football data.
02
Plaiq
Scores the decision, not just the outcome.
It measures off-ball decision quality, pressure and pass option value, from Matrix’s output or a single provider’s data.
Ponta works alongside every provider you use. Independent of all of them, Matrix treats each provider’s data on equal terms.
Matrix
Two systems. One match. Different numbers.
Clubs running more than one tracking system, or about to change one, know the pattern. Each system is right on its own terms.
Optical tracking
Optical tracking gives tactical positioning and team shape.
GPS wearables
GPS wearables give accurate physical performance metrics.
Event providers
Event providers capture actions on the ball.
Together they cause three problems:
- A switch breaks your history. Years of workload data stop lining up with the new numbers.
- Systems disagree. A high-speed run in one system may not be one in another, and substitution and training-load calls often hinge on thresholds.
- Analysts become the integration layer. Timestamps, coordinates, player IDs and sample rates all differ, so someone aligns them by hand.
Matrix Reconcile
Built for provider transitions
Switch tracking or GPS vendors and your historical workload data usually breaks: the old and new numbers aren’t comparable.
A typical switch
Old providerHard breakNew provider
With Matrix Reconcile
Old providerAnalytically continuousNew provider
Matrix Reconcile is designed to keep your multi-year workload history continuous through the switch.
Early access
Matrix · How it works
How Matrix works
From provider exports to one harmonised dataset, in stages you can adopt one at a time.
-
Step 1
Ingest
Optical tracking, wearable GPS and event data, from different providers.
-
Step 2
Normalise
Available now -
Step 3
Reconcile
Early access -
Step 4
Fuse
In development
GPS wearables
Optical tracking
Event data
Matrix
Harmonised dataset
Standard export format
Plaiq, IQ analytics
Matrix · Layers
Three layers, each useful on its own
Each adds to what you get, for every player in every frame.
01
Available nowNormalisation
These files describe the same concept differently.
- Player positions in one standard coordinate system, in metres
- Velocity and acceleration
- A quality and confidence score for each frame
Matrix Normalise: send one provider’s export, get it back standardised.
02
Early accessReconciliation
These systems measured the same phenomenon differently.
- Unified timestamps across providers
- Players matched across sources
Matrix Reconcile: for clubs with two providers. Comparison and continuity through a switch.
03
In developmentFusion
Each source adds what the other is missing.
- Source attribution (GPS, optical or fused)
- Complementary signals combined into one player representation
Matrix Fuse: for clubs with GPS, optical and event data.
ALL LAYERS
Confidence and quality
Position confidence scoring and drift and quality detection run across all three layers.
EXPORTAll of it in a standard export format.
Matrix · Proof
Normalisation across real vendor formats, on open data.
Metrica and SkillCorner describe tracking data in very different ways. Matrix normalises both into one schema: shared columns, one coordinate system, consistent units.
Metrica
- 3-header-row CSV
- 25 Hz
- Coordinates 0–1, top-left origin
- Pitch 106 × 68 m
SkillCorner
- JSONL
- 10 Hz
- Metres, per-period clock
- Pitch 104 × 68 m
After normalisation: one schema, both sources
Metrica
| timestamp | player_id | team | x | y | source | sample_rate_hz |
|---|---|---|---|---|---|---|
| 600.0 | home_1 | home | -6.52 | 4.76 | metrica | 25.0 |
| 600.0 | home_10 | home | 18.88 | 14.65 | metrica | 25.0 |
SkillCorner
| timestamp | player_id | team | x | y | source | sample_rate_hz |
|---|---|---|---|---|---|---|
| 300.0 | home_10 | home | -8.45 | 21.67 | skillcorner | 10.0 |
| 300.0 | home_12 | home | -46.28 | 1.88 | skillcorner | 10.0 |
What comes next: Matrix Reconcile brings two providers’ recordings of the same match together.
Data: open sample data published by Metrica Sports and SkillCorner.
Plaiq
Plaiq scores the decision, not just the outcome.
Plaiq judges every pass against the best option available when it was played, including passes that were intercepted.
-
Off-ball decision quality
Each decision is scored against the best option that was on, then placed in a band from Poor to Optimal.
-
Pressure
How tightly the opposition squeezes each part of the pitch, worked out from where every player stands.
-
Pass option value
What a pass to each teammate is worth at the moment the ball is played.
Every metric is computed on a hexagonal grid of roughly one square metre per cell.
What comes out
-
- Pass
- Carry
- Intercepted pass
Every decision
Passes, carries and intercepted passes, each scored on its own.
-
Four rating bands
Poor, Suboptimal, Good and Optimal, so a score reads at a glance.
-
Every player
A DQ Rating for each player, with a trend through the match.
Matrix’s harmonised output
A single provider’s tracking data
A single provider’s event data
Plaiq
Scored decisions
Player DQ Rating
Trend through the match
Plaiq runs on Matrix’s output, or on its own.
Plaiq · Worked example
One decision, scored.
Plaiq’s output on open sample data: one pass judged against the best option available, and one anonymised player’s decisions across a match.
- Passer
- Where the pass went
- Best option available
- Teammates
- Opponents
A pass, scored
Suboptimal0.58Decision Quality
Scored against the best pass option available when the pass was played, whether or not it was the one taken.
A carry, scored
Good0.83Decision Quality
Scored against the best carry available from the same start.
An intercepted pass, scored
Good0.94Decision Quality
Plaiq infers the intended receiver from the ball’s path (Player 10, strong inference), then scores that choice the same way.
One player, one match
Player 7
- DQ Rating
- 0.86
- Band
- Good
- Decisions
- 135
- Pass (62)
- Carry (57)
- Intercepted pass (16)
- Trend, last four minutes
Each bar is one decision, scored from 0 to 1.5. The shaded bands are the rating bands.
Real open tracking and event data, scored decision by decision. Plaiq does this for every pass, carry and intercepted pass in a match; the ones shown were selected to illustrate.
Data: open sample data published by Metrica Sports (Sample Game 3, anonymised).
FAQ
Frequently asked questions
Which providers does Matrix support?
Matrix reads Metrica and SkillCorner optical tracking and STATSports GPS exports into one schema. Plaiq works with StatsBomb event data, and Matrix is being built to bring event data into the same dataset.
Use a different provider? Tell us below. It shapes what we build next.
How does Ponta fit with our existing providers?
Alongside them. Your providers stay where they are, and Matrix is the integration layer that makes them work together.
Which parts of Matrix are available today?
Matrix Normalise is available now. Matrix Reconcile is opening to design partners, and Matrix Fuse is being built. Each stage is useful on its own.
What does “harmonised” mean?
One consistent, quality-scored dataset: every provider’s data in the same schema, coordinate system and units, with confidence scores alongside.
Get in touch
Using data from more than one provider, or about to switch?
Tell us which providers you use and we’ll show you what Matrix returns.