Opposition analysis football: using win probability and likely scorelines to frame the report
A pre-match probability and an exact-score estimate tell staff how big the task is, what game state to expect and which scenarios to rehearse. Here is how to use and judge them.
Most opposition analysis football reports begin with the opponent's habits: how they build, how they press, who they look for under pressure. That material matters, but it answers "what will they do" before anyone has said "how hard is this match". A pre-match probability and an exact-score estimate put that question first. They give the report a frame, and the detail then fills in the picture inside it.
Start with the probability. A home win, draw and away win split tells the coaching staff the size of the task. If a side is given, say, a 40% chance of winning, that is a match where the opponent is not a clear favourite or underdog, and the plan can be balanced. If the number is far lower, the report should accept that the realistic aim may be a point, and prepare the structure, set pieces and substitutions for a match that is defended into the final twenty minutes. The same number does a different job upstairs: it lets a sporting director or board hear "we are underdogs here" before the result, so a defeat is read against expectation rather than against hope.
The exact-score estimate adds the likely game state. A probability says who is favoured; the most likely scorelines say what kind of match it probably is. If the leading scores cluster around 1-0 and 1-1, the report should plan for a low-event match where one set piece or one transition decides it, and rotation decisions can lean on that. If they stretch to 2-1 and 2-2, the match is likelier to open up, and the staff should prepare for being behind and for protecting a lead. The useful habit is to take the two or three leading scorelines as scenarios, then write the plan for each: how we score first, how we respond if we concede first, what changes at 60 minutes if it is level.
Neither number replaces the film work. The probability is a prior; the video explains why the opponent might beat it. A good report holds both and says where they disagree. If the model rates an opponent as modest but the footage shows a side that has just changed shape and is winning the ball high, the analyst's job is to flag it, not to defer to the figure.
Judging a forecast like this comes down to a few questions. Is it a probability for all three outcomes, or only a pick? Does it update when the line-ups land, since an injury or a rested striker can move the picture? Is the record public, graded on the last call before kickoff, and measured over many matches rather than a few memorable ones? And is the benchmark honest? A forecast that beats "always back the home side" is a low bar, so ask also whether it beats a baseline using the sample's own home, draw and away rates, and whether its probabilities are calibrated, meaning that calls given 70% win about seven times in ten.
Kickwise offers a worked example of how such a record can be presented. Its pre-match probability blends a language model's reading of the match facts (form, team strength, head to head, injuries, suspensions and the confirmed line-ups) with the market's consensus where bookmaker prices exist, and its exact-score estimate starts from a goal simulation calibrated to the bookmakers' view of the favourite. Its public record keeps the last call before kickoff and grades it on the result after 90 minutes plus any extra time. Over the last 30 days, across every confidence level, the record shows 218 right out of 375 (58%) on three-way outcomes, with a Brier score of 0.551 against 0.642 for the baseline, where lower is better. At 65% confidence or higher over the last 90 days it shows 90 right out of 109 (83%), while the home side won 60% of those matches.
Those figures also show the limits, and so do the stated ones. The draw is the hardest outcome to call, and friendlies and matches with thin data are close to coin flips, so an opposition report for a low-data fixture should lean harder on footage and less on the number. A 70% call is still expected to miss about three times in ten, which means one wrong call says almost nothing about the model. Judge any forecast over many matches, and never rewrite a plan because a single result went the other way.
In practice, the frame belongs at the top of the opposition report as three short lines: the size of the task, the probable game state, and the scenarios to prepare. Treat it as the starting point of the discussion, not its conclusion. The best use of a forecast is to make the staff argue more precisely about the match, which is what data driven decision making in football should look like when it works.