| Method | Based On | Starting Point | Update Behaviour | Weighting | Evidence / Stability | Best Use |
|---|---|---|---|---|---|---|
| Elo V1 (Points + Division Prior) |
Average result (0–1) from points; symmetric pairings. | Base 1.00 + division prior (evidence-damped) + small experience lift. | Iterative Elo with evidence-based K (low evidence moves more). | Points-based weight, boosted and capped (max 2.5). | Evidence = points/25; flags low evidence (< 3). | Best when division strength should influence starting ratings. |
| Elo V2 (Points + Games Played) |
Average result (0–1) from points; symmetric pairings. | Base 1.00 + GamesPlayed × 0.03. | Iterative Elo with games-based K. | Points-based weight with non-linear scaling. | Stability driven by games played; flags < 3 games. | Simple model, ideal when divisions are not present. |
| Glicko-2 (Points + Uncertainty) |
Point share per matchup; symmetric aggregation. | μ from averages (P10–P90), plus uncertainty (ϕ). | Updates μ, ϕ and volatility (σ) per rating period. | Points-based weight, boosted and capped (max 2.5). | Flags low evidence or high uncertainty (ϕ > 0.85). | Best for uneven data where uncertainty matters. |