Versus · Metodologia

Últimos 60 dias · até 100 partidas válidas por jogador

Somente partidas coletadas. Remakes excluídos antes do limite. Funções desconhecidas contam no global. Cobertura incompleta suspende o veredito. Wilson descreve taxas individuais; intervalos separados são uma regra prudente, não certeza de 95% sobre duelos ou habilidade. Partidas comuns criam dependência; região, elo oficial, função e atualização importam.

V / (V + D) · N = V + D · Wilson 95% · z = 1.959963984540054

Winner: N₁ ≥ 25 ∧ N₂ ≥ 25 ∧ Wilson₁.low > Wilson₂.high (P1); Wilson₂.low > Wilson₁.high (P2).

Empate observado: V₁ × N₂ = V₂ × N₁. Nenhuma vantagem clara: Wilson₁ ∩ Wilson₂ ≠ ∅.

Melhores resultados: N ≥ 10, Wilson.low ↓, N ↓, champion ID ↑. Lanes: N₁ ≥ 5 ∧ N₂ ≥ 5.

Wilson · NIST · Riot · Game Integrity

1. Sources and data status

In demonstration mode, every match and profile is fictional and labelled as such. Demo records show the interface and exercise calculations; they are not real results.

In Riot live mode, data comes from official Riot APIs, including Champion Mastery, and Data Dragon. Rank is the latest available snapshot. Update time and sample size are shown whenever they affect interpretation.

2. Samples and exclusions

• Personal statistics use only matches in the visible or stated sample.
• Remakes are excluded from win rates, performance averages, personal facts and comparisons.
• Matches with zero or invalid duration are excluded from per-minute comparisons.
• Queue, role, champion, period or patch filters recalculate metrics on that subset.
• A small sample describes observed matches only; it does not establish a player’s general skill.

3. Core formulas

For numerical comparison, a zero-death match uses 1 as the technical KDA denominator. Its individual display still says “Perfect” rather than presenting it as an ordinary KDA.

Win rate

wins ÷ completed games

KDA

(kills + assists) ÷ deaths; “Perfect” when deaths = 0

CS/min

(lane minions + neutral monsters) ÷ minutes

Gold/min

gold earned ÷ minutes

Damage/min

champion damage ÷ minutes

Vision/min

vision score ÷ minutes

Kill participation

(kills + assists) ÷ team kills, clamped to 0–100%

Damage / gold share

player value ÷ team total, clamped to 0–100%

4. Performance comparisons

A completed match is compared with the same player’s other completed matches; the target never enters its own baseline. FFGGWP first uses same queue and role when at least three comparable games exist, then same role with the same threshold, then all eligible games.

The displayed delta is (match value − baseline average) ÷ baseline average. Metrics remain independent and are never combined into a proprietary skill score.

Trend compares the latest 20 completed matches with the previous 20, one metric at a time. It is unavailable when either group contains fewer than five matches.

5. Benchmarks and champion statistics

• The preferred benchmark uses the same champion, role, tier, queue and patch, with at least 100 eligible matches and 20 unique players.
• If that threshold is not met, the first fallback combines current and previous patch at the same rank. The second keeps champion, role and queue but combines all ranks and labels that scope. With no sufficient sample, no comparison is shown.
• Benchmark averages cover KDA, CS/min, gold/min, damage/min, vision/min and kill participation.
• A percentile of 80 or higher is a strength; 95 or higher is labelled Elite. A percentile of 20 or lower may appear in the collapsed Next improvements area. Middle values receive no judgement.
• Across the latest 60 ranked games, OTP requires at least 50%, 20 total games and 15 on the champion; Main requires at least 30% and 10 champion games; Comfort requires at least 10%. Rare pick means 0–2 games and outside the top-three masteries. Below 20 games, only top mastery may be shown.
• An indexable champion page must use verified Riot data and at least 1,000 matches for that role.
• Champion win rate is wins ÷ sample; pick and ban rates use the total games in the displayed population.
• Player rankings use the latest rank; within a tier, league points then wins determine order. This is not an official Riot ranking.

6. What FFGGWP does not calculate

• No hidden MMR, ELO or private win-probability estimate.
• No conclusion about a player’s personality, behaviour or worth.
• No cross-patch or cross-queue comparison unless explicitly labelled.
• In a live game, no public weakness is shown: at most three positive or factual badges use rank, mastery and already-cached history. FFGGWP does not deep-fetch all ten histories.
• No insight is produced for a hidden or anonymous profile.
• No demonstration number represented as real Riot data.

7. Quality, corrections and versions

Results depend on API completeness, refresh delays, role identification and sample size. A data or formula correction may change a metric retroactively.

A material formula or threshold change creates a new methodology version and updates this page. Report a non-reproducible result to contact@ffggwp.com with the URL, filter and value.

Visual performance guide

Green and ✓ mean favorable against the stated reference; orange and ≈ mean a central range or inconclusive sample; red and ! mean room to improve against the reference; grey and — mean no usable reference. These are neither official ratings nor predictions.

Distribution bands reuse the insight P20/P80 boundaries: at least P80 is favorable, at most P20 indicates room to improve. Deaths reverse direction. Cohorts require at least 100 matches and 20 players and a matching context or stated fallback. Profile summary metrics use the available previous personal window (after the latest 20 completed matches, at least 5 observations), not a skill comparison across roles.

Individual win rates use a 95% Wilson interval against the descriptive 50% win/loss balance. An interval containing 50% is orange: this does not mean the player is average. Remakes are excluded. The estimate assumes independent games and does not predict the next game.

Post-match comparisons use the displayed reference: the same-role opponent or available average. Equality is orange, a favorable difference green and an unfavorable difference red. There is no universal KDA threshold or added 5% tolerance. Team gold, damage and objectives retain team identity colors. As team histories can overlap, their pooled win/loss balance is descriptive against 50%, without a statistical interval.

Duration, pick/ban rates, mastery and volumes are not graded without a meaningful reference. Demonstration data remains fictional even when an indicator is green.