Premier League — Análise de Cantos
Seis épocas de dados de cantos de todos os jogos da PL. Índice de chasing por equipa, cantos esperados na 2ª parte por resultado ao intervalo e o que o segmento de odds pré-jogo revela sobre o volume de cantos.
The chasing effect is real but asymmetric. Total 2nd-half corners stay nearly constant (~5.9) regardless of HT score. What changes dramatically is the split: when the home team trails at HT, they earn 3.56 H2 corners vs 2.84 when leading. The trailing team is systematically underpriced on individual corner handicap markets.
Perfis de Canto por Equipa
Ordenado por cantos médios ganhos por jogo. * = menos de 50 jogos em casa (menor confiança).
| # | Equipa | Casa | Fora | Total | Conc.C | Conc.F | Chasing Idx | Tempo atrás |
|---|---|---|---|---|---|---|---|---|
| 1 | Manchester City | 7.9 | 5.9 | 6.9 | 2.4 | 3.9 | 0.42x |
17% trailing
|
| 2 | Liverpool | 7.3 | 6.3 | 6.8 | 3.6 | 3.8 | 0.60x |
20% trailing
|
| 3 | Arsenal | 6.7 | 5.2 | 6.0 | 3.7 | 4.4 | 0.47x |
15% trailing
|
| 4 | Chelsea | 6.5 | 5.3 | 5.9 | 4.0 | 4.3 | 0.79x |
22% trailing
|
| 5 | Aston Villa | 6.0 | 5.1 | 5.5 | 4.6 | 5.7 | 1.46x |
32% trailing
|
| 6 | Bournemouth | 5.7 | 5.1 | 5.4 | 5.5 | 6.0 | 1.65x |
35% trailing
|
| 7 | Tottenham Hotspur | 5.9 | 4.9 | 5.4 | 4.4 | 6.0 | 1.29x |
31% trailing
|
| 8 | Luton Town * | 6.6 | 4.1 | 5.4 | 5.1 | 6.8 | 4.30x |
57% trailing
|
| 9 | Manchester United | 6.0 | 4.6 | 5.3 | 4.4 | 6.0 | 1.05x |
25% trailing
|
| 10 | Brighton and Hove Albion | 6.1 | 4.4 | 5.2 | 4.1 | 5.4 | 1.49x |
28% trailing
|
| 11 | Leeds United | 5.8 | 4.6 | 5.2 | 4.4 | 5.4 | 1.95x |
34% trailing
|
| 12 | Fulham | 5.4 | 4.7 | 5.0 | 5.0 | 5.2 | 1.54x |
28% trailing
|
| 13 | Newcastle United | 5.6 | 4.5 | 5.0 | 5.0 | 5.8 | 1.02x |
23% trailing
|
| 14 | Southampton | 4.7 | 5.2 | 4.9 | 5.6 | 5.6 | 2.26x |
42% trailing
|
| 15 | Leicester City | 4.9 | 4.6 | 4.7 | 5.1 | 5.8 | 1.77x |
37% trailing
|
| 16 | West Ham United | 5.0 | 4.5 | 4.7 | 5.0 | 6.2 | 1.78x |
33% trailing
|
| 17 | Brentford | 4.8 | 4.4 | 4.6 | 5.2 | 6.3 | 1.46x |
31% trailing
|
| 18 | Sheffield United | 5.3 | 3.9 | 4.6 | 5.5 | 7.4 | 3.59x |
41% trailing
|
| 19 | Everton | 5.1 | 4.1 | 4.6 | 5.2 | 6.2 | 1.65x |
29% trailing
|
| 20 | Burnley | 5.0 | 4.0 | 4.5 | 5.6 | 7.3 | 2.32x |
38% trailing
|
| 21 | Crystal Palace | 5.0 | 4.0 | 4.5 | 4.8 | 5.7 | 1.44x |
28% trailing
|
| 22 | Wolverhampton Wanderers | 4.7 | 4.2 | 4.4 | 4.9 | 5.8 | 2.58x |
36% trailing
|
| 23 | Watford * | 4.4 | 4.3 | 4.3 | 5.3 | 6.1 | 3.40x |
39% trailing
|
| 24 | Norwich City * | 5.0 | 3.6 | 4.3 | 6.6 | 7.1 | 3.51x |
42% trailing
|
| 25 | Nottingham Forest | 4.4 | 3.7 | 4.0 | 5.5 | 6.4 | 1.52x |
31% trailing
|
| 26 | West Bromwich Albion * | 5.7 | 2.3 | 4.0 | 6.2 | 7.7 | 1.28x |
25% trailing
|
| 27 | Ipswich Town * | 3.8 | 3.5 | 3.7 | 5.4 | 7.4 | 4.00x |
44% trailing
|
| 28 | Sunderland * | 3.8 | 3.3 | 3.6 | 4.8 | 5.8 | 2.33x |
27% trailing
|
Conc. = cantos sofridos (C = em casa, F = fora). Chasing Index = cantos quando perde ÷ cantos quando lidera. Valores >2.0 = a equipa depende muito de chasing quando está atrás.
Chasing Index — O que significa para apostas
Estas equipas geram mais cantos quando estão a recuperar. Evite 'mais cantos totais' quando estão a ganhar — o volume colapsa.
Estas equipas atacam de forma constante independentemente do marcador. O volume de cantos é previsível — tanto totais como handicaps individuais.
Modelo Live — Cantos Esperados 2ª Parte por Resultado ao Intervalo
Usar ao intervalo: cruza marcador HT (perspectiva da equipa da casa) com o segmento de odds. Cada célula mostra: cantos totais 2ª parte / split casa | fora / amostra. Escala: teal = baixo, vermelho = alto.
| Resultado HT | Forte fav. casa (>62%) | Leve fav. casa (50–62%) | Equilibrado (42–50%) | Leve fav. fora (32–42%) | Forte fav. fora (<32%) |
|---|---|---|---|---|---|
| HT -2 | — | — | — |
6.5
4.0 | 2.5 n=19 |
5.4
2.8 | 2.6 n=42 |
| HT -1 |
6.8
4.8 | 1.9 n=20 |
6.8
4.8 | 2.0 n=35 |
6.0
4.4 | 1.6 n=32 |
6.0
3.3 | 2.3 n=58 |
5.4
2.7 | 2.5 n=98 |
| HT 0-0 |
6.2
4.2 | 1.9 n=96 |
5.6
3.4 | 2.2 n=90 |
5.6
2.9 | 2.7 n=71 |
5.8
3.0 | 2.8 n=111 |
6.0
2.5 | 3.4 n=160 |
| HT +1 |
5.6
3.5 | 2.1 n=90 |
5.8
2.8 | 3.0 n=65 |
5.4
2.1 | 3.3 n=46 |
5.6
2.5 | 3.1 n=45 |
6.1
1.8 | 4.0 n=59 |
| HT +2 |
5.4
3.6 | 1.8 n=35 | — | — |
6.7
2.4 | 4.2 n=21 | — |
Como usar isto ao intervalo
- Anota o marcador ao intervalo — ex. 0-1 significa estado HT = HT -1 para a equipa da casa.
- Verifica as odds pré-jogo — categoriza em strong_home / slight_home / balanced / slight_away / strong_away.
- Consulta a tabela — encontra os cantos totais esperados na 2ª parte e o split casa/fora.
- Compara com o mercado ao vivo — se o mercado oferece Mais de 4,5 cantos 2ª parte e o modelo diz 6,0, 'mais' tem valor.
- Verifica o chasing index — se a equipa atrás tem um índice alto (>2,5), apostar ainda mais nos seus cantos na 2ª parte.
🛑 Efeito do Cartão Vermelho nos Cantos — Sinal Live
Baseado em 135 eventos de cartão vermelho com janelas pré/pós válidas em 2.464 jogos PL.
Effect by red card minute
| Momento do vermelho | Jogos | Ganho 11-jog./10min | Perda 10-jog./10min | Impacto prático (30min) |
|---|---|---|---|---|
| Cedo ≤35' | 40 | +0.469 | −0.149 | ≈ +2,6 cantos extra para a equipa de 11 em 30 min |
| Meio 36'–65' | 55 | +0.109 | −0.170 | ≈ +0,3 cantos nos 15 min restantes |
| Tarde >65' | 40 | +0.339 | −0.276 | Maior queda da equipa de 10 — muda para defesa pura |
⚖ Posse → Calibração Cantos
4,735 team-match records. Pearson r = 0.465 (possession vs corners). Shots are an even better predictor: r = 0.544.
| Intervalo de posse | Jogos | Cantos médios | Visual |
|---|---|---|---|
| 25–30% | 246 | 2.93 | |
| 30–35% | 342 | 3.58 | |
| 35–40% | 478 | 4.00 | |
| 40–45% | 595 | 4.44 | |
| 45–50% | 670 | 4.89 | |
| 50–55% | 665 | 5.47 | |
| 55–60% | 606 | 5.89 | |
| 60–65% | 490 | 6.50 | |
| 65–70% | 361 | 7.29 | |
| 70–75% | 282 | 8.35 |
ⓘ Key insight: shots (r=0.544) beat possession (r=0.465) as a corner predictor. Teams that generate shots — not just passes — earn more corners. Use expected shots (from xG models) as a better pre-match corner estimator.
Eficiência dos Cantos — Volume vs Perigo
Winning more corners doesn't mean scoring from them. Manchester City average 7.05 corners/match — the highest in the dataset — yet rank last (17.8/100) on danger score. Their corners rarely reach the six-yard box (only 11.3% into-box delivery rate). Nottingham Forest average just 4.27/match but convert nearly every corner into a genuine scoring chance: 46.7% produce a direct shot, 53.1% generate a header.
| # | Equipa | C/Jogo | Remates/C | Na Grande Área | Cabeceamentos | Cab. Enquadrada | Pontuação Perigo |
|---|---|---|---|---|---|---|---|
| 1 | Nottingham Forest | 4.3 | 46.7% | 39.2% | 53.1% | 11.4% |
88.8
|
| 2 | Sunderland * | 3.5 | 43.0% | 39.0% | 55.0% | 15.0% |
84.8
|
| 3 | Everton | 4.6 | 44.7% | 31.4% | 51.3% | 12.8% |
75.2
|
| 4 | Ipswich Town * | 3.9 | 45.5% | 30.3% | 46.2% | 11.7% |
70.6
|
| 5 | Brentford | 4.7 | 43.0% | 24.1% | 56.6% | 15.6% |
70.1
|
| 6 | West Ham United | 4.9 | 43.1% | 27.0% | 50.2% | 12.7% |
65.3
|
| 7 | Crystal Palace | 4.7 | 42.8% | 35.4% | 41.7% | 9.8% |
63.9
|
| 8 | Fulham | 5.1 | 46.0% | 23.6% | 47.5% | 10.0% |
63.6
|
| 9 | Leeds United | 5.3 | 44.4% | 29.4% | 42.5% | 9.9% |
62.0
|
| 10 | Norwich City | 4.5 | 39.6% | 42.7% | 35.1% | 9.1% |
58.0
|
| 11 | Newcastle United | 5.5 | 42.9% | 25.9% | 44.3% | 10.6% |
56.5
|
| 12 | Burnley | 4.5 | 38.7% | 31.1% | 47.4% | 11.5% |
55.4
|
| 13 | Luton Town * | 5.4 | 39.2% | 25.0% | 52.0% | 10.8% |
52.8
|
| 14 | Wolverhampton Wanderers | 4.6 | 39.3% | 27.3% | 43.2% | 11.4% |
49.1
|
| 15 | West Bromwich Albion * | 4.1 | 34.5% | 42.3% | 35.9% | 8.5% |
44.9
|
| 16 | Sheffield United | 4.7 | 34.9% | 40.1% | 35.3% | 10.0% |
44.6
|
| 17 | Southampton | 5.1 | 40.7% | 23.9% | 37.5% | 11.6% |
44.4
|
| 18 | Brighton and Hove Albion | 5.4 | 41.1% | 21.0% | 41.8% | 10.1% |
44.1
|
| 19 | Manchester United | 5.3 | 41.4% | 23.1% | 38.9% | 9.2% |
43.8
|
| 20 | Tottenham Hotspur | 5.6 | 38.2% | 26.2% | 36.3% | 9.3% |
37.4
|
| 21 | Bournemouth | 5.5 | 37.6% | 26.2% | 39.3% | 8.1% |
37.0
|
| 22 | Liverpool | 6.9 | 43.5% | 12.4% | 39.9% | 8.2% |
36.9
|
| 23 | Watford | 4.4 | 34.2% | 30.3% | 38.8% | 7.6% |
32.2
|
| 24 | Aston Villa | 5.7 | 37.7% | 23.4% | 35.5% | 7.5% |
30.4
|
| 25 | Arsenal | 6.0 | 41.2% | 17.2% | 34.2% | 6.3% |
29.9
|
| 26 | Chelsea | 6.1 | 37.1% | 16.2% | 37.2% | 8.5% |
23.4
|
| 27 | Leicester City | 5.0 | 32.6% | 28.8% | 32.0% | 6.4% |
19.8
|
| 28 | Manchester City | 7.0 | 39.3% | 11.3% | 31.8% | 7.4% |
17.8
|
* menos de 50 aparições — tratar com precaução
👉 Betting angle: when a low-danger team (City, Arsenal, Chelsea) is generating corner volume against a physical side, the corners-on-target market may be overpriced. When Brentford, West Ham, or Forest earn even 4 corners, each one carries twice the headed-shot threat of a City corner.
Formação Tática vs Volume de Cantos
Formation is the single most actionable pre-match corner predictor. You know the formation before kick-off. A 4-3-3 playing at home averages 6.45 corners — nearly double the 5-4-1's 3.41. Wide formations (4-3-3, 4-2-3-1, 4-1-4-1) generate +0.97 more home corners than defensive/narrow setups (5-4-1, 5-3-2, 4-4-1-1). The effect is consistent: it holds both home and away.
| # | Formação | Ganhos | Conc. | Net | N |
|---|---|---|---|---|---|
| 1 | 4-3-3 | 6.45 | 4.14 | +2.31 | 496 |
| 2 | 4-1-4-1 | 5.98 | 4.46 | +1.52 | 113 |
| 3 | 4-2-3-1 | 5.65 | 4.66 | +0.99 | 897 |
| 4 | 4-4-2 | 5.40 | 5.32 | +0.08 | 232 |
| 5 | 3-4-1-2 | 5.38 | 4.42 | +0.96 | 40 |
| 6 | 3-4-2-1 | 5.35 | 4.41 | +0.94 | 210 |
| 7 | 4-4-1-1 | 5.12 | 5.41 | -0.29 | 59 |
| 8 | 3-5-2 | 5.09 | 5.09 | 0.00 | 120 |
| 9 | 3-4-3 | 5.01 | 4.40 | +0.61 | 88 |
| 10 | 4-5-1 | 4.93 | 6.17 | -1.24 | 30 |
| 11 | 5-3-2 | 4.87 | 4.97 | -0.10 | 38 |
| 12 | 5-4-1 | 3.41 | 6.63 | -3.22 | 46 |
| # | Formação | Ganhos | Conc. | Net | N |
|---|---|---|---|---|---|
| 1 | 4-3-3 | 5.41 | 4.84 | +0.57 | 476 |
| 2 | 3-4-1-2 | 5.05 | 6.05 | -1.00 | 43 |
| 3 | 3-4-3 | 4.99 | 5.59 | -0.60 | 81 |
| 4 | 4-2-3-1 | 4.75 | 5.45 | -0.70 | 836 |
| 5 | 4-4-2 | 4.70 | 6.08 | -1.38 | 233 |
| 6 | 4-1-4-1 | 4.48 | 6.06 | -1.58 | 111 |
| 7 | 3-5-2 | 4.26 | 6.56 | -2.30 | 135 |
| 8 | 3-4-2-1 | 4.13 | 5.55 | -1.42 | 243 |
| 9 | 4-4-1-1 | 3.92 | 6.78 | -2.86 | 63 |
| 10 | 5-3-2 | 3.65 | 7.63 | -3.98 | 54 |
| 11 | 5-4-1 | 3.28 | 7.07 | -3.79 | 58 |
👉 How to use: check both teams' confirmed formations before betting corners. A 4-3-3 vs 3-4-2-1 matchup averages 9.3 corners — the lowest matchup type. A 4-1-4-1 vs 4-3-3 matchup averages 12.3. That's a 3-corner swing driven purely by shape, before you even consider team quality or odds.
Como os Cantos São Gerados
Corners come almost entirely from two sources: goalkeeper saves that are parried wide, and outfield blocks that deflect the ball out of play. Quantifying these conversion rates across all 2,464 PL matches reveals the mechanical link between shot volume and corner volume.
Team variation: the rate at which a goalkeeper's saves produce corners varies between 15.7% (Sunderland) and 30.0% (Southampton). Teams with high-quality goalkeepers who catch or hold shots (rather than parrying) concede fewer corners from saves — Chelsea (22.5%), Wolves (22.1%), Aston Villa (22.0%). Teams with shot-stopping GKs who push wide concede more — Southampton (30.0%), Nottingham Forest (29.3%), Everton (28.8%).
ⓘ Implication: teams that generate shots on target — not just corner-forcing set pieces — are the primary corner generators. Expected shots on target (xSOT) is a better predictor of corner volume than pass count or even possession.
Mitos sobre Cantos — O que os Dados Desmentem
Three common bettor beliefs tested across 2,400+ matches. All three fail to reach even r=0.03 correlation with corner counts.
Metodologia
- Source: premierleague.com commentary JSON — 2,464 matches, GW1–GW38, seasons 2019–20 through 2024–25.
- 25,674 corner events extracted and tagged with score state at moment of corner.
- Odds segments computed from normalised implied probability (bookmaker margin removed).
- Chasing Index = (corners earned while trailing) ÷ (corners earned while leading). Neutral = 1.0.
- Live model cells with n<10 are hidden (insufficient sample).
- Backtest (2024–25 season, n=291): MAE = 2.57 corners — consistent with the natural variance of corner distributions.
- Red card analysis: 135 events with valid 30-min pre/post windows. Corner rates computed per 10 minutes, normalised for window length.
- Possession calibration: 4,735 team-match records from clean_stats.json. Pearson r computed over all records. Possession brackets at 5% intervals.
- Corner efficiency: team_stats.json fields wonCorners, attCorner, attHdTotal, attHdTarget aggregated per team across all 2,464 matches. accurateCornersIntobox read from defending team entry (Opta stores it as opponent's in-box deliveries). Danger score = 0.35×shots + 0.35×into-box + 0.20×headers + 0.10×headers-OT, normalised 0–100.
- Formations: lineups.json formation field extracted for 2,461 matches. Min 30 appearances per formation for table inclusion. Wide formations defined as 4-3-3, 4-2-3-1, 4-1-4-1, 4-4-2, 3-4-3.
- Save→corner rate: 14,486 'attempt saved' events tracked for next-event corner within same minute. Block→corner: 18,260 blocked shots. Save-to-corner rate per team computed from defending team perspective.
- Weather: weather_summary.json provides temperature, precipitation (mm), wind speed (km/h) at kickoff for 2,329 matches. Pearson correlations computed against total match corners. Result: r<0.03 for all weather variables — not significant.
- Yellow card effect: 6,368 yellow card windows (±20 min) from 9,011 yellow card events. Corner rate computed pre/post per 10 minutes. Net swing +0.028/10min = 6% of red card effect.
- Rest days: advanced_metrics.json context.home_rest_days/away_rest_days tested against corners across 2,424 matches. Pearson r<0.02 for all rest day variables — not significant.