Why Draws Are So Hard to Predict
About a quarter of football matches end level, yet the draw is almost never the most likely outcome. The maths behind it, and how to read draw probabilities sensibly.
By the GoalStatsHub team · · 5 min read
Roughly one league match in four ends level, give or take a few points depending on the league and season. Yet if you look through any set of predictions, the draw is almost never the outcome the model rates most likely. That isn’t a flaw in the model. It follows from how football scores are spread, and understanding it will change how you read the 1X2 figures.
A draw is many outcomes squeezed into a narrow band
A home win can happen in dozens of ways: 1-0, 2-0, 2-1, 3-1 and so on. A draw needs the two sides to finish on exactly the same number. As soon as a match is expected to produce a fair number of goals, the chance that they land on the same figure falls.
Here is what a simple Poisson model gives for evenly matched teams at different scoring levels:
| Expected goals (each side) | Home | Draw | Away |
|---|---|---|---|
| 1.3 – 1.3 | 36.8% | 26.4% | 36.8% |
| 1.0 – 1.0 | 34.6% | 30.9% | 34.6% |
| 0.8 – 0.8 | 32.3% | 35.3% | 32.3% |
Even when the two teams are identical, the draw only becomes the single most likely result in a very low-scoring match. In a typical game with 2.6 goals expected, each side’s win is more likely than the draw, even though a quarter of such games end level.
Add a favourite and the draw falls further. With expected goals of 2.0 against 0.8, a simple model gives 65.4% home, 20.5% draw and 14.1% away.
Why “pick the most likely outcome” never picks a draw
Suppose you take every match and back whichever of 1, X or 2 the model rates highest. Because the draw is rarely on top, you will almost never pick it, even though about 25% of matches end that way. Your record will look reasonable on wins and miss every draw. This is a property of the strategy, not of the probabilities: the model is saying “draw 27%” and meaning it.
The figure that matters is the draw probability itself, compared with the price. A 30% draw is a meaningful chance, and on our match pages a near-even split is called out in the preview as “closely matched” rather than presented as a win for one side.
When draws become more likely
- Evenly matched sides, especially when both defences are strong and goals are scarce.
- Low-scoring leagues and tight fixtures: the lower the total expected goals, the higher the draw share.
- Game situations a pre-match model can’t see: a point that suits both teams late in a season, or a second leg where a draw takes one side through.
A known weakness of simple models
Textbook Poisson models slightly underestimate 0-0 and 1-1. In real matches the two teams’ goals aren’t fully independent: a side protecting a lead plays differently from one chasing the game. The correction published by Dixon and Coles in 1997 nudges these low-scoring draws up, and many modern models adjust for the effect in some way. Our guide to how prediction models work covers the method.
Markets built around the draw
- Double chance (1X, X2, 12) covers two of the three outcomes. On our pages 1X is the home win plus the draw, so it is always the sum of those two figures.
- Draw no bet returns the stake if the match is drawn, which is the same as Asian handicap 0. Our Asian handicap table shows it with the push chance listed separately.
- Half-time draw: at half-time the level score is far more common than at full time, often the single most likely half-time state. Our half-time markets show it next to the full-time figures.
18+. These figures are information, not betting advice. Please read our responsible gambling page.