Fatum

Find live-match conditions that had often been followed by a goal in historical data.

≈10 years

Historical data

>50

Leagues in the first research stage

>100

Leagues after expansion

Historical match data and patterns identified by football analysts were compared with live match conditions. Incoming live data was processed in Google Sheets.

Co-founder and data analyst

• Worked with historical and live match data.
• Helped structure patterns identified by football analysts.
• Organised incoming data in Google Sheets.
• Reviewed signal outcomes and model performance.

1. Receive live match data.
2. Process incoming events in Google Sheets.
3. Compare current conditions with historical patterns.
4. Issue a signal when similar pre-goal conditions appear.
5. Check whether and when a goal follows.

≈90%

Signals followed by a goal in the same match

≈40%

Goals scored within 15 minutes after a signal

These figures apply only to the signals included in the internal test. They are not the model's overall accuracy and do not mean that the system predicted 90% of all goals.

• The published figures come only from internal testing.
• They do not show the model's overall accuracy or how many of all goals it could predict.

Signal count alone is not enough to evaluate a system. The signal rules, observation period, false signals, sample size and testing on new data all need to be documented.