Results Audit: Breaking Down dixon_coles_v1 Model Picks

In the evaluation window running from September 23, 2026, through October 10, 2026, the dixon_coles_v1 model compiled an overall record of 17 wins, 21 losses, and 4 pushes across 42 picks in 25 games. That performance produced a win rate of 0.447 and a net output of -1.38 units. Across a brief timeframe of just over two weeks, these figures provide a limited sample that highlights both high-payout wins and notable misses without establishing a definitive conclusion on model efficiency.

Market Breakdown and Summary Records

The 42 picks generated by dixon_coles_v1 were split evenly between spread and total markets, with 21 selections apiece.

Market	Picks	Wins	Losses	Pushes	Net Units
Spread	21	8	9	4	-0.38
Total	21	9	12	0	-1.00
Total Overall	42	17	21	4	-1.38

Spread selections concluded with 8 wins, 9 losses, and 4 pushes for -0.38 units. The totals market delivered 9 wins and 12 losses with zero pushes for -1.00 unit. Across all picks in the window, one graded pick lacked a final score, while two graded picks were omitted from the primary sample breakdown.

High-Divergence Winners and Outlier Hits

The largest positive unit outcome emerged in Brighton's visit to Sunderland. The model listed a 0.2609 model probability for the away spread of -1.5 against a decimal price of 4.400. Brighton secured a 0-2 victory, covering the line and producing +3.40 units.

Another significant gap appeared in Real Madrid's 1-0 win over Villarreal. The model set under 2.5 goals at a 0.2831 model probability against a decimal price of 3.600, yielding +2.60 units.

In SC Paderborn 07's 0-2 defeat to VfB Stuttgart, under 2.5 goals carried a 0.6442 model probability and a 3.000 decimal price, returning +2.00 units. Similarly, Manchester United's 1-1 draw with Tottenham saw under 2.5 goals hold a 0.4051 model probability at a 2.600 decimal price, adding +1.60 units.

High-Probability Setbacks and Match Performance

High model probabilities did not guarantee coverage. In Napoli's 3-0 win over Frosinone, the model posted a 0.8527 model probability on away spread 1.5 at a 1.680 decimal price, resulting in a -1.00 unit loss. In FC Augsburg's 2-2 draw with Bayern München, the model assigned a 0.2172 model probability to under 2.5 goals at a 5.500 decimal price, which lost -1.00 unit when four goals were scored.

Match-level actions mirrored these outcomes. In Borussia Dortmund's 2-2 draw with Werder Bremen, Dortmund spread -1.5 (0.5330 model probability, 1.990 decimal price) and under 2.5 goals (0.3942 model probability, 3.000 decimal price) both lost -1.00 unit. Nico Schlotterbeck and Daniel Svensson scored for Dortmund, while Ludovit Reis and Niclas Füllkrug scored for Werder Bremen, assisted by Oskar Wójcik and Marco Grüll.

In Lens's 2-1 victory against Lyon, under 2.5 goals (0.4412 model probability, 2.520 decimal price) lost -1.00 unit. Matthieu Udol and Florian Sotoca scored for Lens, driven by 2 assists from Florian Thauvin. In Málaga's 1-1 draw with Espanyol, Espanyol spread 0.0 (0.6493 model probability, 1.935 decimal price) resulted in a push at 0.00 units; Bryan Zaragoza scored for Espanyol via a Javier Hernández assist, while Rafael Garrido assisted for Málaga.

Evaluating Sample Size Context

A run of 42 picks over 25 games represents a very small sample size. Short windows in soccer modeling carry inherent volatility, where a small cluster of high-payout covers or multi-goal draws heavily influences net unit returns. These numbers summarize the window directly without offering a definitive judgment on the model's long-term edge.