Reading the clean-sheet rings
Each ring in the chart above is a 5-segment donut. Each segment = one of GW1-5. Brighter segment = higher clean-sheet probability that week. Centre stat = cumulative 5-GW average.
The Onside model rates clean-sheet probability as a function of: opponent xG-against, recent form, home/away venue, and projected starting lineup. Arsenal leads at 53% average CS/GW because their opening 5 includes 3 home games against bottom-half + the structural quality of Saliba/Gabriel + Raya behind them.
Clean sheet probability matters because every CS = 4 points for goalkeepers + 4 points for defenders + 1 point for midfielders. Across 38 GWs the difference between a 50% CS club defender and a 30% CS club defender = +30 points/season. Picking your defensive double-up around the top-CS clubs captures structural value before any individual stat enters the equation.
The 3 CS-tier-1 clubs (50%+ avg)
Arsenal (53% avg CS/GW). Gabriel + Timber + Raya all benefit. The 3-defender Arsenal exposure is the model's highest-confidence template — captures both the attacking BPS leadership AND the structural CS bonus.
Liverpool (46% avg). Van Dijk + Alisson are the only Tier-1 picks; supporting defenders (Konaté, Robertson) are squad fillers post-TAA exit. Single-Liverpool defensive exposure is the model recommendation.
Manchester City (50% avg). Gvardiol + Akanji + Ederson all qualify. City's structural xGA is lower than Arsenal's but the CS conversion rate is comparable. Single Man City defender + the Haaland exposure works.
The 3 CS-tier-2 clubs (40-50%)
Newcastle (46% avg). Pope at £5.0m is the standout goalkeeper differential — 46% CS probability at the cheapest top-7 club price point. Tonali + Burn are the supporting cast but don't qualify as Tier-1 picks individually.
Aston Villa (42% avg). Cash + Konsa pair behind Watkins. Emery's back-line is structurally sound but the home/away split is wide — Villa's away CS probability is just 32%. Pick Villa defenders for home fixtures only.
Chelsea (40% avg). James (when fit) + Jackson + Sánchez. Post-Cucurella, Chelsea's defensive depth thins; the CS probability is fragile. Single Chelsea defender at £5.5m is the model exposure.
The 2 CS-tier-3 clubs (35-40% — avoid)
Brighton (38% avg). Verbruggen + Van Hecke. Premium Brighton picks don't qualify as Tier-1 at the £4.5-5m price point.
Crystal Palace (35% avg). Mateta + Eze + Olise are the model's attacking Palace exposure; the defence isn't structurally tight enough to justify CS-tier picks.
Below these two, the next-tier clubs (Forest, West Ham, Wolves, Brentford, Bournemouth) all sit at 30-35% CS probability — bench-only territory for FPL defensive picks.