Competition analytics

Golavo keeps four kinds of competition analysis separate because they answer different questions and carry different time guarantees.

Existing-data analytics

Domestic leagues and UEFA club competitions can show competition-local strength trends and rest/congestion derived from the indexed matches. A value of 100 is that competition’s own baseline; it is not a cross-league power ranking. Report cards use chronological held-out seasons and seeded match-bootstrap intervals.

Schedule difficulty is not calculated until the remaining fixture list is complete. An absent schedule is a blocked capability, never a set of zeroes.

Tournament and season outlooks

The World Cup 2026 outlook exactly enumerates the resolved four-team bracket for the Ratings voice, Goals voice, and equal-chance baseline. The voices stay separate and the disclosed knockout rule resolves regulation draws. This is a current simulation, not a sealed prediction and not part of the forecast ledger.

Domestic standings use verified competition-specific rules, including known points adjustments in the validation season. The seeded 10,000-run outlook can start only after Golavo certifies every team, ordered home/away pair, past result, and remaining fixture. All five bundled leagues now pass that certificate for 2026/27 — their full published schedules are pinned from the OpenFootball Football.TXT country repos — so the outlook runs. The certificate is still checked on every request, and a league that ever fails it produces the blocked state and no probabilities rather than a guess.

For an available domestic outlook, a conditional run can assign scores to 1–12 unique future fixtures and apply them together through the same three model voices. Golavo shows the returned engine values beside the canonical verified values for the selected voice; the interface does not invent a probability delta. The response is stamped hypothetical_only, persisted: false, and model_input: false; it exists only in the current view, cannot be sealed, and does not change the canonical verified outlook.

The run-in: projected points and what a fixture moves

An available domestic outlook also reports, for each remaining fixture, how far its result moves the two clubs’ season stakes. Golavo does not simulate the season again to get this. It partitions the runs it already made by that fixture’s simulated outcome and compares the branches: the swing is the gap between a club’s title, top-four, or relegation chance when it wins that fixture and when it loses it. The reported number is the largest of those three swings, in percentage points.

Two honesty limits apply. A conditional branch thinner than 200 of the 10,000 runs is sampling noise, so the fixture reports insufficient_coverage and no number at all — that is why a fixture one club is overwhelmingly expected to win may carry no swing, and why a fixture fixed by a hypothetical scenario never carries one. And because the council voices are never blended, every swing comes from a single named voice, which the payload states in importance.voice_id.

Alongside it, each club’s projected points is the mean final points across the same voice’s runs. The run-in table bands upcoming opponents by that same projected finish rather than joining in a second rating source, so one table never mixes two models.

For a top-five fixture, Match Cockpit can show the same certified current-season context for both clubs: exact table position and points, one named voice’s projected points and season chances, and the fixture’s guarded stakes when its exact match id is in the remaining schedule. It carries the outlook’s as-of time, source ids, and index hash and remains a descriptive simulation—not a sealed match forecast.

My Teams

Source main can collect exact competition-and-club favorites in a local-only My Teams room. Each available club card reuses the current table and one disclosed outlook voice to show current/projected points, season stakes, and the next five fixtures with guarded importance plus follow/pick state. It does not blend voices or create a new simulation.

Favorites are browser preferences, not a profile. If the active outlook no longer contains the exact saved identity after a rename, promotion, or relegation, Golavo preserves it and shows the mismatch instead of guessing. See My Teams.

World Cup history archive

Model Lab → World Cup lab starts with a read-only tournament archive: the women’s 1991–2019 record first, with the men’s 1930–2022 record available beside it. It shows titles, finals, appearances, each edition’s final four, and recorded player awards.

Every row comes from the pinned Fjelstul World Cup Database snapshot under CC-BY-SA-4.0. Golavo verifies the pack hash, filters and summarizes its rows, and shows the source, license, and pinned revision in the interface. The archive cannot train a model, alter a probability, enter an outlook, or become a seal.

Golden Boot and shootouts

An international competition shows a leading-scorers table and a penalty-shootout ledger, built from the bundled martj42 goalscorers and shootouts records. Both are leak-safe: they count only matches played by the requested cutoff, so a completed player’s tally is fixed and rewinding the clock removes later goals. Own goals are never credited to a scorer. Neither table carries a competition column upstream, so each is joined to the match index by date and teams to attach the competition.

This data ships only for men’s internationals, so a club competition reports a typed “no scorer data” state rather than an empty table — the same first-class unknown the rest of the app uses.

Historical team research

The isolated Pappalardo/Wyscout CC-BY-4.0 pack contains seven team-only summaries: the 2017/18 Premier League, La Liga, Bundesliga, Serie A, and Ligue 1; Euro 2016; and World Cup 2018. Together they cover 1,941 matches and 3,251,294 events.

Each collapsed panel names its competition and era before exposing pass completion, progressive passes, shots, goals, a disclosed same-team event-run proxy, and Golavo’s own 12×8 research-xT calculation. It does not ship raw events or player identities, does not claim observed xG, and never enters a current forecast or simulation.

Conditions and weather

Conditions Snapshot is display-only. Pinned GeoNames and Natural Earth resources can provide exact city context, local kickoff, elevation, rest, travel distance, and an offline route map. Stadium remains unknown without stadium-level evidence.

Weather is explicitly blocked. Golavo will not replace the forecast that was available before kickoff with observed weather gathered afterwards. Until a licensed source preserves issued-at history, the contract has no weather source id or value and states model_input: false.

See Coverage and Sources & licenses for exact eras, licenses, and exclusions.