The model on 20 October: 27 out of 48
Of the 48 matches our model predicted before kick-off on Tuesday 20 October 2026, it called the result of 27 (56%). It named the exact scoreline 5 times. Across all 4,373 matches it has predicted since 19 July, the rate is 48%.
What stood out
At 56% the day ran 8 points above the model's running rate of 48%. The result it least expected was FC Porto 0-1 PSV Eindhoven: it had given a win for PSV Eindhoven 18%. Its best competition of the day was Pro League (Saudi-Arabia) (4 of 4), its weakest League Two (England) (4 of 12).
By competition
- League Two: 4 of 12
- League One: 7 of 11
- UEFA Champions League: 6 of 9
- Premier League: 3 of 4
- Primera A: 2 of 4
- Pro League: 4 of 4
- Championship (Scotland): 0 of 1
- Championship (England): 0 of 1
- 2 other competitions: 1 of 2
The most confident misses
The predictions it was surest about and got wrong:
- FC Porto 0-1 PSV Eindhoven: the model had FC Porto at 61%.
- Barnet 1-3 Swindon Town: the model had Barnet at 58%.
- York 1-1 Accrington Stanley: the model had York at 51%.
Exact scorelines
- Luton 2-1 Barnsley: the exact score it named, with Luton at 56%.
- Asyut Petrol 0-1 Zamalek SC: the exact score it named, with Zamalek SC at 49%.
- Al Diriyah 1-0 Al Riyadh: the exact score it named, with Al Diriyah at 47%.
How this is counted
A result counts as called when the outcome the model rated most likely (home win, draw or away win) is the one that happened. Every prediction here was published and logged before kick-off, and the log is public: the track record breaks it down by confidence, by competition and by club. How the probabilities are produced, including where the model is known to be wrong, is in the methodology.
Every evening on Telegram
This summary is posted each evening on our Telegram channel, t.me/fulltimestats_com, with a link back to the full recap.