How Our Ratings Work

KRACH is a pure win/loss rating built on the same mathematical foundation used in chess ratings and sports analytics. It answers the question: "Based on who you beat and who those teams beat, how good are you really?"

How it works

Every team starts with an equal rating. The algorithm then asks: if Team A (rated Ka) plays Team B (rated Kb), the expected probability of A winning is Ka / (Ka + Kb). Ratings are adjusted repeatedly — hundreds of iterations — until each team's expected win total exactly matches their actual win total. That convergence point is the final KRACH score.

Because the math chains through every opponent and every opponent's opponent, a team's KRACH automatically reflects the strength of its schedule — no manual adjustment needed.

Full method, formulas and sources →

Reading the numbers

• KRACH — The core rating, scaled so all teams in a division sum to 10,000. Higher is better. Because of that scaling it is comparable within a division but not between divisions.
• SOS — Strength of Schedule: a weighted average of your opponents' KRACH ratings, where each opponent's weight is the games you played against them divided by the sum of the two ratings. That weighting is what makes KRACH = (wins/losses) × SOS hold exactly.
• RRWP — Round Robin Winning Percentage: the winning percentage a team would be expected to earn if it played every other team in its division an equal number of times, based on their KRACH ratings. Higher is better; .750 means an expected 75% winning percentage.
• Adj. KRACH — A reliability-adjusted version that scales a team's rating down when it has played only part of a season (by the square root of games played against a 27-game baseline), so a 3-0 team in October isn't ranked above a 20-2 team in February.

PairWise was the system used by NCAA Division I hockey to select and seed teams for the tournament for roughly three decades. The NCAA replaced it with NPI (below) beginning with the 2026 tournament, but it remains a useful head-to-head lens. It answers the question: "In a direct comparison between any two teams, which one has the stronger case?"

How it works

Every team is compared with every other team in the division, scoring points on three criteria:

1. RPI — One point to the team with the higher Rating Percentage Index (see below).
2. Head-to-head — One point for each game a team won against the other. A tied game scores nothing.
3. Record vs. common opponents — One point to the team with the better record against the opponents both teams share. A winning percentage is worked out against each common opponent, and those percentages are averaged, so an opponent met five times counts no more than one met once.

The team with more points earns a PairWise Win. If the points are level, the team with the better RPI takes it, so a comparison only ends as a tie when the RPIs are level too. Criteria that don't apply score nothing — if two teams have never met, there are no head-to-head points, and the same goes for common opponents when they share none. A team's final PWR Points = PairWise Wins + 0.5 × PairWise Ties. Teams are ranked by PWR Points descending.

What is RPI?

RPI (Rating Percentage Index) blends three win-percentage numbers, with NCAA hockey's weights: 25% of your own winning percentage + 21% of your opponents' winning percentage + 54% of your opponents' opponents' winning percentage. Opponents count once for every game you played them, and their records leave out their games against you. Three-quarters of the weight is on schedule, so who you played matters more than that you won.

Reading the numbers

• PairWise # — Your rank by PWR Points among all teams in the division.
• PairWise PWR — Raw PWR Points (wins + 0.5 × ties in pairwise comparisons).
• PairWise Adj PWR — PWR Points pulled toward the division midpoint (half the number of other teams) for teams with few games played. The pull shrinks as games accumulate: a team's distance from the midpoint is scaled by games played ÷ (games played + 16).
• PairWise RPI — Your RPI score (roughly 0.0–1.0; higher is better).

NPI is an iterative rating that combines your winning percentage with a schedule-adjusted strength of schedule. It answers the question: "How does your win percentage hold up when we account for who you played — and who they played?"

How it works

The formula is: NPI = (Win% × 25%) + (Strength of Schedule × 75%)

The 75% weight on Strength of Schedule makes NPI much more sensitive to schedule difficulty than a plain winning percentage. Crucially, your opponents' SOS component uses their NPI ratings — which in turn depend on their opponents' NPI ratings — making the whole thing iterative. Ratings are recalculated in a loop until they stabilize.

A team that goes 10-0 against weak competition will score lower on NPI than a team that goes 8-2 against strong competition.

Reading the numbers

• NPI # — Your rank by NPI score among all teams in the division.
• NPI — The raw NPI score on a 0–100 scale. Higher is better.
• NPI SOS — The strength-of-schedule component that fed into your NPI.

SRS (used by myhockeyrankings.com under the name "Average Goal Differential + Strength of Schedule") is the only rating here that uses goals scored rather than just wins and losses. It answers the question: "How many goals better or worse than average are you, after accounting for how tough your schedule was?"

How it works

The formula is: SRS = AGD + SOS

• AGD (Average Goal Differential) — The average margin of victory or defeat across all your games, capped at ±7 goals per game. The cap prevents a single blowout win from inflating a team's rating.
• SOS (Strength of Schedule) — The weighted average SRS rating of your opponents, weighted by how many games you played against each one.

Like KRACH and NPI, SRS is iterative — your opponents' ratings depend on their opponents' ratings — and it loops until the ratings converge. Once converged, an additive shift is applied so the top-rated team in the division lands at exactly 99.9.

Example: A team that wins every game 5–0 against average opponents has an AGD of +5. If their schedule was slightly below average (SOS = −0.5), their SRS would be 4.5 before normalization. A team that wins 3–2 every game but only plays elite opponents (SOS = +3.0) could rank higher despite a smaller margin of victory.

Reading the numbers

• SRS # — Your rank by normalized SRS rating in the division.
• SRS — The normalized rating (top team = 99.9). All other teams are shifted by the same offset, so the gaps between teams remain meaningful in goals-per-game units.
• SRS AGD — Your Average Goal Differential per game (capped at ±7).
• SRS SOS — The average SRS of the opponents you played, on the same scale as SRS, so SRS = AGD + SOS.

All four ratings are updated every Monday morning and recalculated whenever new game scores are imported. Ratings only include games played through the current week's cutoff date.

Further reading: the full KRACH method · NPI primer (College Hockey News) · RPI