To complement the existing subjective polls and the NCAA’s NPI system, we’ve developed two fully objective ranking models built on modern analytics:
- ChampsRankSOS – Strength-of-Schedule Model
- ChampsRankELO – ELO-Based Rating System
Here are the first two iterations of the models as of games completed on November 30th, 2025:
ChampsRankSOS – Strength-of-Schedule Model
ChampsRankELO – ELO-Based Rating System
Both models are designed to be transparent, data-driven, and updated within a few hours of each game, making them some of the fastest-refreshing public rankings available.
Our intent is to provide a clear, unbiased picture of team strength that reflects actual on-ice performance while avoiding the opacity and inertia that can affect human-voted polls.
ChampsRankSOS — Strength-of-Schedule Model
This system is structurally similar to MyHockeyRankings, but with important enhancements that make it better suited for college hockey.
Key Features
- Strength-of-Schedule + margin of victory as core components
- ±7 goal cap on margin of victory (Note: There was consideration for increasing/ignoring a cap, but after the recent 17-2 Wisconsin win over Stonehill the current cap would not let these kind of anomalies bias the data)
- Full iterative recalculation of all ratings until stable
- 45-day decay model — recent games are weighted more heavily
- Games older than 45 days typically represent 10+ games ago, meaning they don’t accurately reflect a team’s current form.
- The decay ensures the rankings reward teams improving now, not teams who were hot months ago.
- Updated within hours of each game
Core Formula
Rating = GD + SOS
- GD: Average goal differential per game (capped at ±7)
- SOS: Average opponent rating (recalculated iteratively)
How It Works
- Compute each team’s average goal differential
- Initialize ratings
- Iteratively recalculate: Rating = GD + SOS
- Anchor the top team at 20.0
- Recompute until convergence
- Apply time decay to down-weight older games
ChampsRankELO — ELO-Based Model
Our second system is a modernized hockey-specific ELO model. The ELO model is a rating system that evaluates team or player strength by updating ratings after every game based on the expected outcome versus the actual result. It rewards upsets, penalizes underperformance, and naturally adjusts as more games are played. Originally developed for chess, ELO has since become widely used in sports such as tennis, soccer, basketball, and esports due to its ability to track performance dynamically and objectively.
Key Features
- Home-ice advantage bonus: +11 ELO points (based on historical win rates)
- 45-day half-life time decay on older games
- K-factor multipliers for early-season games, conference games, etc.
- Margin-of-victory multiplier for blowout wins
- Adjusts outcomes differently for regulation, OT, and shootout
- Updated within hours of each game
Core Formula
R_new = R_old + K × MOV_multiplier × (S − E)
- K: Sensitivity constant (base 32, with multipliers)
- MOV: Margin of victory multiplier
- S: Actual game outcome
- E: Expected outcome (based on rating difference + home-ice advantage)
How It Works
- All teams begin at 1500 ELO
- Process games chronologically (one pass)
- For each game:
- Compute expected outcome
- Apply home-ice advantage
- Update ratings using K-factor + MOV + decay
- No iteration required — ratings naturally evolve over time
Main Differences at a Glance
| Feature | ChampsRankSOS | ChampsRankELO |
|---|---|---|
| Basis | Goal differential | Win/loss outcomes |
| Processing | Iterative | Sequential (one pass) |
| Starting Point | AGD-based | All teams start at 1500 |
| Top Team | Fixed at 20.0 | Emerges naturally |
| Home Advantage | Not modeled | +11 ELO points |
| Time Decay | Yes — 45-day decay | Yes — 45-day half-life |














