Chapter 1 · 9 min read
Chapter 1 — How to analyze a football match BEFORE kickoff
Pre-match analysis answers one question: what do we know that the market hasn't fully priced yet? Eight categories of data, ordered by predictive weight. Every one of them already exists on PronoStats — the links at the bottom of each section take you to where that data lives, in two clicks.
1. Recent form — the last 5, read properly
Form is the most-read and the most-misread metric. A team on LLWWL isn't automatically in crisis: it depends on WHO they lost to and WHERE. Three rules: (a) weight opponent quality — beating the bottom side is worth less than drawing with the leader; (b) split home and away — many teams have two identities; (c) five matches is a tiny sample, don't over-weight it. PronoStats shows the 5-match string (WLWDL) next to every team name, and the home/away split in match detail.
2. Injuries and suspensions — the real weight
A starting striker missing isn't the same as a backup full-back missing. PronoStats syncs injuries every 6 hours and tags them by type (muscular, card-related, illness) and expected duration. The right question: does the replacement have minutes in their legs? If yes, expect a 10-15% impact. If they play once a month, it can be 30%. Suspensions from yellow-card accumulation often slip under the news radar but change the eleven. For internationals (World Cup, Euros), factor in travel fatigue and time-zone shifts.
3. xG vs goals scored — who deserves more than they're getting
Expected goals (xG) measures chance quality independent of who finished. A team averaging 1.8 xG but 1.2 actual goals is creating well and finishing poorly: the market underrates them because it reads only the scoreline. Long-term, xG converges to real goals — so it's a leading indicator of mean reversion. PronoStats integrates StatsBomb xG and shows home/away xG forecasts in Premium predictions. The rule of thumb: xG minus goals > +0.4 per match → a lucky team about to regress. xG minus goals < -0.4 → an unlucky team about to bounce.
4. Head-to-head (H2H) — when it matters and when it's noise
H2H is the public's most over-rated stat and one of the least useful when used raw. Three filters first: (a) same coach? If both teams changed bench twice, the 4-year-old games are different teams; (b) same context? A Champions final and a friendly shouldn't be averaged; (c) sample size — fewer than 5 recent games = noise. PronoStats shows last-10 H2H with dates and context. Use it to spot real bogey-teams (sides that historically struggle even with full squad turnover) — that signal is real.
5. Odds movement — why a dropping price says something
Bookmaker odds aren't prophecies but reactions: they move when money flow tilts one way (recreational) or when sharp news arrives (sharp money). An odds drop >15% in the hours before kickoff means someone informed is betting hard. PronoStats tracks the previous value of every odds line and flags 'Dropping Odds' in real time — one of the most underrated Premium signals. It doesn't mean that team will win, it means the market thinks so with more conviction than 30 minutes ago.
6. Probable lineups — reading the turnover
A week with three matches close together, a Champions tie on Thursday, an away trip on Sunday: every rational coach rotates. The probable lineup tells you whether they're playing squad A (10-15% extra strength) or B. PronoStats syncs lineups 5 minutes before kickoff, but 24h out you have indicators: did they play 96 hours ago? Big tie in 3 days? All this ends up in the model's feature engineering, but for a human reader it's enough to know a City facing Real Madrid plays differently at Bournemouth.
7. Standings pressure — who has something to play for
In May, a mid-table team with nothing to play for plays at 20% less intensity than a side fighting relegation. Motivation is real and ignored by pure-stats models. Three high-value setups: (a) three points between relegation and safety with 4 games left; (b) head-to-head Champions race between 4th-5th-6th; (c) already-crowned champions facing desperate sides (upset potential). PronoStats shows the live table with point gaps from threshold positions.
8. Glicko-2 — the rating that beats ELO
ELO assigns a fixed score to each team. Glicko-2 adds two things: an uncertainty (RD — rating deviation) and a volatility. What does that mean in practice? A freshly-promoted team has a high rating but huge RD: the model knows it doesn't know, and tells you to be cautious. A Champions League team with 100 games of data has low RD: the rating is reliable. PronoStats shows Glicko-2 + RD + volatility in every Premium prediction. When Glicko-2 is high but RD is also high, the scenario reads as 'market expectations without confirming data' — potential value bet territory.
Chapter quiz
Check if the concepts stuck — nothing tracked, just for you.
1.Is the last-5-games form a good predictive indicator?
2.What does an xG of 1.8 with actual goals averaging 1.2 mean?
3.Which pre-match signal carries the most asymmetric info (what the market doesn't know)?
Frequently asked questions
Which pre-match data point matters most?
None alone. The PronoStats model blends 50+ features; no single one explains more than ~12% of variance. Sharp odds movement and xG differential are the two most asymmetric signals (information the public ignores).
How long before kickoff should I do my analysis?
24-48 hours before. Earlier, you miss injury and lineup news; later, the odds have already adjusted. Sweet spot: the night before.
Are the model's 50+ features public?
Yes, the list is documented on /methodology. The exact weights aren't — they're learned from data and change on each nightly retrain.