← Home France LNB · Sun 22 Nov 2026, 17:30

Paris – Gravelines-Dunkerque

Paris host Gravelines-Dunkerque. The model's total is 181.0 points, against the market's line of 177.5. Pick: Over 168.5 at 1.84, logged 165 hours before tip-off with an edge of +18.4%. The line has since moved to 177.5; the pick stays as logged.

In short

Figures in brackets: each one's effect on the model total, in points. Effects under 1 point are left out.

The model vs the market

Model totalParisGravelines-DunkerqueLineOver priceUnder priceOver chanceOver edge
Opening line181.0100.580.5168.51.841.8564%+18%
Line now181.0100.580.5177.51.851.8656%+4%

The two team columns: the model total split between the teams (Team totals below). Line now: the latest line, at the last hourly run; it becomes the closing line at tip-off.

How the model total is built

Points
League average, recent games counting more173.7
Paris+5.9
Gravelines-Dunkerque−2.5
Projected total177.1
Starting point, from past games+1.6
The league against its lines+0.1
The two teams against their lines−1.3
The line against the league average+1.7
Game pace+0.8
3-point luck+0.8
Home and away form+0.4
Model total, every signal in181.0

Each team's number is how many points above or below the league average its games produce, its scoring and its defence together, worked out from this season's games only. The rows below the projected total are the other signals: each one's effect on the chance of the over, put back into points. Together they give the model total, the total that on its own would give the same chance against the line.

Team totals

TeamScores a gameAllows a gameProjected points
Paris (home)97.285.9100.5
Gravelines-Dunkerque (away)79.189.580.5
Model total181.0
Expected margin, Paris+20.0

Each team's projected points: the league's 86.9 points per team, plus 1.7 for the home side (minus for the away side), plus how much more the team scores than that this season, plus how much more its opponent allows. Teams with few games count less. The two are then scaled to add up to the model total, every signal in. Display only: the pick rule uses only the total.

From the total to the chance of the over

SignalReadingEffectChance
Starting point, from past games––52.7%
Projected total against the line−0.4 points against the line−0.7 pts52.0%
The league against its lines+2.0 points a game against its lines lately+0.1 pts52.1%
The two teams against their lines+1.0 points a game against their lines−2.2 pts49.8%
The line against the league average+3.8 points against the league average+2.9 pts52.8%
Game pace+0.9 possessions a game against the league+1.3 pts54.1%
3-point luck−0.7 points a game from threes above the league rate+1.3 pts55.5%
Home and away form+1.5 points: home side at home, away side away, against all their games+0.7 pts56.1%

How much each signal counts is learned from past games, so a small gap to the line can still move the chance, and a big one moves it less than it looks.

These steps use the current line of 177.5.

From the chance to the edge

The edge is the chance times the price, minus 1. When the pick was logged: 0.643 × 1.84 − 1 = +18.4% at the line of 168.5.

The rules

RuleThis game
Edge above 8%+18.4%met
Regular seasonyesmet
Both teams 3+ games9 and 8met
Half-point line (.5)168.5met
3-point luck under +1.5−0.6met

First half

Points
League first-half average, recent games counting more86.7
Paris+2.2
Gravelines-Dunkerque−2.9
Correction: first halves ran this much above the projection, 2022–23 to 2025–26+0.8
Model 1H86.8
Model 1H1H lineOver priceUnder priceOver chanceOver edgeUnder edge
86.889.51.861.8639%−27%+13%

Built only from earlier first halves, the way the projected total is built from earlier totals: the league's first-half average and each team's first halves this season against it. It never sees the full-game line or the model total. The chance of the over comes from how widely first halves spread around this projection in the league (10.0 points). The 1H picks are a live test, apart from the record: the over when its edge is above 8% at a .5 first-half line.

Paris: last 5 league games

DateAgainstClosing lineTotalTotal − line
Sat 14 Nov 2026Limoges180.5179−1.5
Sun 8 Nov 2026JL Bourg182.5184+1.5
Tue 3 Nov 2026Cholet176.5184+7.5
Sun 1 Nov 2026Chalon/Saone180.5228+47.5
Sun 25 Oct 2026Roanne172.5171−1.5

Gravelines-Dunkerque: last 5 league games

DateAgainstClosing lineTotalTotal − line
Sun 15 Nov 2026Nancy164.5185+20.5
Fri 6 Nov 2026Boulazac163.5152−11.5
Sat 31 Oct 2026Roanne165.5158−7.5
Sat 24 Oct 2026JL Bourg169.5170+0.5
Sun 18 Oct 2026Dijon166.5175+8.5

Against the closing line this season

TeamGamesOver–underMeanMedianSpreadAt homeAway
Paris95–4+6.7+1.521.1+12.5 (4)+2.1 (5)
Gravelines-Dunkerque85–3+2.2+4.519.7+3.9 (5)−0.5 (3)

Each game's total minus its closing line: above 0, the game went over what the market expected. The closing line already allows for the opponent, the venue and the news, so this shows how far the market has been off on a team, not how high its games score. The mean is the long-run average; the median is the typical game, and a gap between them means a few unusual games. Spread is the standard deviation: single games miss the line by about that much either way, so a mean of a few points over a handful of games is description, not proof. The model already uses this, weighted by how recent and how many games: it is the signal "The two teams against their lines" above. Counting it again, as a mean, median, home and away split or last five games, did not improve the picks in the 2023–24 to 2025–26 backtest.

What it cannot see

No news: injuries, rest days and roster moves are invisible to the model until they show up in results.