How Bawler's football model works
Bawler turns team data into match probabilities, and probabilities into one combined pick for each match: Bawler's pick. This page explains how, in plain English: the model, the inputs, how the pick is built and settled, and where the model is weak.
1. The scoring model
We model each team's goals as a rate: how many they are expected to score, and how many they are expected to concede, with a linking term that captures how the two sides' scoring moves together. From the resulting spread of possible scorelines we derive every outcome probability — win/draw/win, over/under goals, both teams to score, correct score, and more.
2. The inputs
- Rolling expected goals (xG) — time-decayed, so recent matches weigh more than old ones.
- Form & opponent strength — adjusted for the quality of opposition faced.
- Home advantage — league- and venue-aware.
- Team news — availability is taken into account where it is known when a call is made. Calls are locked at about 5pm UK time, typically 2 days before kickoff, so later news, including the confirmed lineups, is not in them.
- End-of-season motivation — title, European and relegation pressure in the final weeks.
3. Bawler's pick
On each match page you will find Bawler's pick: one combined pick for the match, made of a few parts (we call them legs) that all have to come in. The legs come from what we can measure well:
- The result, such as a team to win or draw, or both teams to score.
- Shots and corners, for one side or for the whole match, and shots on target.
- A named player: his shots, his shots on target, or the fouls he commits.
Each line is set below what the team or player usually manages, so each leg on its own is likely. The pick follows what we expect from the game: a match with a clear favourite gets a different shape from one where both sides should score.
How the chance is worked out. Each leg has its own chance from the model. Legs in the same match are linked: a side that dominates tends to win, take more shots and win more corners all at once. So we checked, on past matches the model had not seen, how often each combination of legs really came in together compared with simply multiplying their chances. That measured factor is applied to the combined chance. Combinations whose legs pull against each other were tested and left out.
The fair price next to the pick is 1 divided by that combined chance. It is what the pick would pay if the price matched our chance exactly, with no bookmaker margin.
Frozen before kickoff. Each pick is saved to the Bawler's pick record before the match starts, and the pick itself is never changed after that. It is settled from the final match stats: it wins only if every leg left in play comes in. If a named player doesn't start, his part is dropped and the pick is settled on the rest. If the match is not played, the pick is void. If a result is later found to break one of these rules (for example a match that was postponed), it is corrected, and the record shows what it was before.
4. The calls on each match
As well as Bawler's pick, every match page shows the model's own chances for each outcome, and up to three single calls:
- Most likely — the model's main call. If one side is a clear favourite to win outright, that win is the call. Otherwise it is the most likely of the outcomes the model rates well above even, which is often a goals call such as “over 1.5 goals” or a “win or draw”. Most of these calls have been stated at 69–85%.
- Close call — an outcome the model rates at around even, pointing the same way as its favourite and saying something different from the main call. Many matches have none. Most have been stated at 48–71%.
- Outside shot — a less likely outcome that still has a real chance, chosen so it does not clash with the other two calls. Most have been stated at 23–36%.
All three are free.
5. When calls are published
Calls are published and locked at about 5pm UK time, typically 2 days before kickoff. Results are settled automatically from live match data; we never mark our own picks by hand.
6. Weekly recalibration
The model is checked against actual results every week and adjusted, so that when it gives something a given chance, that thing happens about that often. The aim is honest chances, not just more wins.
7. Limitations & known failure modes
- Low-data leagues and early-season fixtures have noisier xG and wider error.
- Derbies and cup one-offs are less predictable than the model's base rates imply.
- Team news that arrives after a call is locked, including the confirmed lineups, isn't in the call.
- Player legs depend on the player starting. If he does not, his part is dropped and the pick is settled on the rest.
- Luck decides a lot in the short run — judge any pick over many matches, not single weeks.
Bawler publishes statistical estimates. Model probabilities are not guaranteed outcomes, and past results do not guarantee future ones.