Decision Tree Match Predictor
Three yes-or-no questions, learned from 20 Scottish Premiership seasons. Enter both sides' points from the table and their recent results, and watch the match find its way down the tree to a group of past matches.
Last season's points are over 38 games. Promoted side? Use about 45, what a newly promoted side usually earns.
That makes the gaps for recent form, this season and last season, in points per game.
Each gap is points per game, home side minus away side: over the last five matches, over this season so far, and over last season. Positive means the home side has been better.
Try an example
How to read it
- Enter what each side has done: points from its last five games, this season so far, and last season. A league table and the last few results have all of it.
- The model only sees the gaps, home side minus away side. It doesn't know who is playing, just how far apart they've been.
- The three chances add up to 100%. A 60% home win still means the home side fails to win four times in ten.
- It knows only results. Injuries, suspensions, transfers and a new manager are invisible to it.
- Change one number at a time to see which ones the model cares about.
- Home win
- Draw
- Away win
This match lands in a group of training matches.
Try "Evenly matched": with no gaps at all the tree makes the away side favourite, because its group also holds plenty of weaker home sides. And change the points from the last 5 games: the tree never asks about recent form.
How it works
The tree asks at most three questions, each about one gap: is it at or below a cut? A yes goes one way, a no the other, and after three answers the match lands in one of eight groups of training matches. The chances are simply what happened in that group, with one imaginary home win, draw and away win added so nothing is ever 0%.
In plain football
- The tree chose its own questions and cuts from every Premiership match from 2001/02 to 2020/21, picking each time the question that left the two halves easiest to call.
- It asks about this season and last season. Recent form never made the cut.
- Because it works in groups, a tiny change in a gap can move a match to a different group and change the chances a lot; within a group, nothing changes at all.
On the five test seasons, 2021/22 to 2025/26, this tree scored a log loss of 0.960, against 0.950 for the logistic regression, 0.932 for the bookmakers and 1.099 for a model that knows nothing. Two questions deep it did slightly better, 0.954; the third question makes the groups easier to read, not the forecasts better.
Where it goes wrong
- It moves in steps, not slopes: sides just either side of a cut get very different forecasts.
- It knows only results. Injuries, suspensions, transfers and new managers are invisible to it.
- It was grown on the Scottish Premiership, where two clubs have usually been far ahead of the rest; another league would grow a different tree.
How it was built and tested: Three questions and a prediction. Decision trees
These figures come from a statistical model and are for analysis and education. Football remains uncertain and model predictions will frequently be wrong.