Goalence

Football Glossary

Every metric, model term, and prediction market we use, defined in one place. Pi-Ratings, Poisson, xG, Drama Score, BTTS — what they mean and how we calculate them.

GOALENCE-EXCLUSIVE METRICS

Forward-Tracking

The commitment to only count predictions toward accuracy stats if they were logged before kickoff. No back-fill, no after-the-fact corrections, no 'what the model would have said' — only what it actually said in advance. The opposite of survivorship bias.

Goalence's accuracy figure on /methodology reflects exactly the predictions stored in predictions_log.json before each match's kickoff. The Statsman audit checks for any resolved entries with a future kickoff timestamp.

Drama Score

A Goalence-exclusive composite measure of how unpredictable a team's matches are. It combines comeback rate, late-goal rate, lead conversion, and second-half dominance into a single read on entertainment value. Higher = more drama per 90 minutes.

Galatasaray's drama score sits near the top of the Süper Lig — 50% comeback rate plus six derby matches per season. Manchester City has a different profile: 62% second-half dominance but lower drama.

Comeback Rate

The percentage of matches in which a team comes back from a losing position to win or draw. Computed minute-by-minute from event timelines, not just final scores.

Galatasaray 2025-26: 50% comeback rate over 10 matches behind. Bayern Munich: 28% — strong front-runners but rarely behind to begin with.

Lead Conversion

Of all the matches a team led at any point, the percentage they actually won. Previously called 'choke rate' (the inverse); reframed as lead conversion for positive framing. Higher = better at managing a lead.

Real Madrid: 88.9% lead conversion in 27 matches ahead this season. Lower-table sides typically convert leads at 55-65%.

Late Drama

The percentage of a team's total goals scored in the 75th minute or later. A measure of late-match impact and stamina.

Real Madrid late drama: 18% — nearly one in five goals comes after the 75th. Lower-tempo sides typically sit at 10-12%.

First-Goal Effect

A pair of conditional win rates: how often a team wins when it scores first, and how often it still wins after conceding first. A spread between these two reveals psychological / front-running character.

Real Madrid: 87.5% win rate when scoring first, 33.3% when conceding first — a wide gap shows front-runner DNA.

On-Pitch Metric (Sahadayken Gol)

A strict plus/minus-style measure: while a player was actually on the pitch (between substitution times), how many minutes did their team spend ahead vs behind. Goals are credited only when the player was physically on the field — not just selected to the squad.

Goalence computes this minute-by-minute from /fixtures/events. Vinicius Jr 2025-26 UCL: 66.4% winning-minute rate while on the field.

Elite Pick

A Goalence prediction with confidence ≥ 70%. The model only flags a match as 'elite' if the Poisson + Pi-Ratings stack agrees on a dominant outcome.

Real Madrid vs a relegation-zone side at home with 72% home-win confidence: flagged elite. A coin-flip derby: never elite.

Win Rate per Minute

Of all the minutes a player spent on the pitch this season, the percentage during which their team was leading. A more honest impact metric than match-level win rate because it strips out 'I was on the bench while we scored' artefacts.

Vinicius Jr 2025-26 UCL: 66.4% winning-minute rate. Bellingham: 62.8%. Both are top of the Real Madrid squad on this metric.

Second-Half Dominance

The difference between a team's win rate in the second half (minutes 46-90) vs the first half (1-45). A high positive gap is the signature of a fitness-led, late-pressing side.

Manchester City: 1H win rate 31%, 2H 62% — a +31 second-half gap. This pattern is rare and points to elite conditioning.

Derby Index

A Goalence sub-aggregate of drama metrics computed exclusively from a team's derby matches (hand-curated rival pairings). It exposes whether a team plays differently in marquee fixtures than in regular league rounds.

Galatasaray 2025-26: six derbies played, 50% comeback rate, 67% lead conversion. Compare to non-derby drama profile to spot 'rises to the occasion' patterns.

Confidence Threshold

Goalence groups predictions into three tiers by the probability of the picked outcome: Elite (≥ 70%), Safe (55-70%), and Normal (< 55%). Accuracy is reported separately for each tier so users can see how the model performs at its strongest signal.

On /app/stats, Elite tier sits around 65-70% accuracy across resolved matches, Safe around 55-60%, Normal closer to the baseline.

STATISTICAL MODEL

Pi-Ratings

A ratings system for football teams that separates a team's home and away strength into two distinct numbers, with home and away ratings updated after each match based on the match result and pre-match expectation. Developed by Constantinou & Fenton (2013).

Goalence's ratings file contains a home rating and an away rating per team across 32 leagues. Real Madrid's home rating sits 0.4 points above their away rating — a typical home-field gap in La Liga.

Poisson Distribution

The probability distribution that describes how many independent events occur in a fixed interval — perfect for modelling football goals, where each minute on the pitch is an independent shot at scoring. With each team's expected goals (lambda) plugged in, the distribution returns the probability of every plausible scoreline.

Goalence's model: if lambda_home = 1.8 and lambda_away = 1.2, then P(home wins 2-1) ≈ 9.4%, P(0-0) ≈ 5.5%, P(home wins 3-0) ≈ 4.4%. The model then sums by outcome to get 1X2 probabilities.

λ (Lambda)

In Goalence's Poisson model, lambda is the expected number of goals a team will score in a single match. Two lambdas are computed per fixture — one for the home side, one for the away — and then plugged into the Poisson distribution to derive scoreline probabilities.

Real Madrid vs Barcelona at the Bernabéu: lambda_home ≈ 2.1, lambda_away ≈ 1.3. The most likely scoreline (2-1) follows directly from these two numbers.

Round-Robin League

A league format where every team plays every other team a fixed number of times (usually twice — home and away). Total matches per team = 2 × (number of teams - 1). Bundesliga: 18 teams, 34 matches each.

Goalence's Statsman uses the round-robin formula `2 × fixture_count ÷ team_count` to validate per-team match counts. A team showing 35 matches in an 18-team Bundesliga means a playoff fixture leaked in.

PERFORMANCE METRICS

xG — Expected Goals

A statistical estimate of how many goals a team or player should have scored based on the quality of chances created (shot location, angle, body part, situation). Each shot gets a probability between 0 and 1; a 0.30 xG shot is one most strikers would convert about 30% of the time.

Manchester City 2025-26 average: ~2.4 xG per home match. Liverpool: ~2.1 xG. Goalence uses xG-adjusted lambda values inside the Poisson scoreboard prediction.

PREDICTION MARKETS

BTTS — Both Teams To Score

A prediction market that resolves Yes if both teams score at least one goal in the match, No otherwise. Goalence outputs a BTTS probability for every fixture by integrating the Poisson distribution over scorelines.

Goalence prediction example: BTTS Yes 65%, No 35%. A high-attacking matchup (Liverpool vs Manchester City) typically lands above 70% Yes.

Over/Under 2.5

A prediction market that resolves Over if the total goals in the match are 3 or more, Under if 2 or fewer. Computed from Goalence's two lambdas: P(home_goals + away_goals ≥ 3).

Real Madrid vs Atlético with lambdas 1.8 and 1.4: P(Over 2.5) ≈ 55%, P(Under 2.5) ≈ 45%.

Match Result (1X2)

The traditional outcome market: Home win (1), Draw (X), or Away win (2). Goalence outputs three probabilities summing to 100% for every fixture.

Real Madrid vs Barcelona: 1 = 72%, X = 10%, 2 = 18%. The model picks the highest as the headline outcome.