
Over the past few weeks, I’ve grown increasingly frustrated with the two major weekly polls in women’s Division I college hockey. Because I track nearly all 45 DI teams closely, I have a pretty strong feel for each team’s true performance level. Yet after weekends filled with upsets or narrow wins over weaker opponents, I’m continually surprised by how certain teams — and some conferences — appear to be ranked higher than expected. While the NPI (see below for detailed explanation) is objective and used for at-large playoff selections, it also isn’t perfect.
As a result, I decided to apply my own analytics to create a new, objective Champs App ranking for DI women’s hockey. But before sharing that system, here’s a breakdown of the current major polls and ranking models.
🔢 Three Types of Ratings: Subjective vs. Objective
Women’s DI hockey currently uses three major rating systems, which fall into two categories:
- Subjective (human-voted):
- USCHO.com Poll
- USA Hockey / AHCA Poll
- Objective (mathematical):
- NCAA Power Index (NPI) — now the official NCAA selection metric
- PairWise Ranking (PWR) — the legacy system, replaced by NPI
Below is a clear summary of how each poll or model works.

1. USCHO.com Poll (Subjective)
The USCHO.com poll is a traditional, human-voted ranking composed of sportswriters, broadcasters, and coaches. Neither the list of voters nor their individual ballots are published.
PROS
Contextual Judgment
- Captures elements no algorithm can quantify: the “eye test,” injuries, momentum, travel fatigue, and lineup changes.
Media Relevance
- Drives debate, fan engagement, and weekly storylines.
Focus on Current Form
- Voters can quickly adjust for hot streaks or slumps, sometimes more rapidly than data-based systems.
CONS
Lack of Transparency
- No published criteria. Voters have full discretion, making results unpredictable and unauditable.
Inconsistency and Bias
- Subject to inertia (teams maintaining rank despite bad losses) and regional bias. It’s not difficult to guess which conferences receive the benefit of the doubt.
Weak Tournament Predictor
- Often diverges significantly from the objective NPI used to select NCAA tournament teams.

2. USA Hockey / AHCA Poll (Subjective)
This weekly poll is conducted by USA Hockey in partnership with the American Hockey Coaches Association (AHCA).
Methodology
- Human-voted, similar to USCHO.
- Voters include coaches and journalists from all NCAA women’s hockey conferences.
- Rankings are based on total points from submitted ballots.
While it provides valuable insight from actual DI coaches, it shares the same challenges as USCHO:
- Only 19 voters
- No transparency into who they are or how they vote
- Susceptible to the same regional biases and subjective inconsistencies
The coexistence of two separate human polls does help smooth out extreme opinions — and when they differ noticeably, it signals a lack of consensus that adds useful context that a single mathematical model cannot provide.

3. NCAA Power Index (NPI) and PairWise (PWR) (Objective)
The NCAA Power Index has fully replaced PairWise as the official NCAA tournament selection tool. NPI is a streamlined, strength-of-schedule-driven model that uses an opponent-based rating system and assigns bonuses for beating highly rated teams.
PROS
Pure Objectivity
- Removes human bias. Rankings come directly from win percentage and opponent strength, based on a fully transparent formula.
Improved Strength-of-Schedule (SOS)
- Uses opponents’ NPI ratings directly, replacing the more convoluted RPI components of the old PairWise system.
Rewards Quality Wins
- Includes a Quality Win Bonus (QWB) for beating strong opponents — and importantly does not penalize teams for beating weaker opponents (a major flaw of old RPI).
CONS
No Contextual Adjustments
- Ignores coaching changes, injuries, goalie slumps, or roster disruptions that human voters naturally account for.
Occasional Mathematical Oddities
- Any complex model can produce counterintuitive outcomes in specific cases.
Self-Referencing Structure
- Because a team’s NPI depends on opponents’ NPI — which depends on their opponents — the calculation must be iterated to find a stable solution.
NPI Statistical Engine (Simplified)
- 25%: Win Percentage
- 75%: Opponents’ NPI (Strength of Schedule)
- Quality Win Bonus (QWB): Extra credit for beating high-NPI teams
- Bad Win Treatment: Mechanisms to remove or neutralize extremely low-value wins
- Strength-of-Schedule (SOS): Directly uses opponents’ final NPI rating for a cleaner, more intuitive strength measure
🔜 What’s Next
In the next post, I’ll introduce the Champs App proposal for two new objective ranking models:
- A simplified, transparent Strength-of-Schedule Index
- An ELO-based model similar to the systems used in chess and tennis
Both provide intuitive, statistically robust alternatives to today’s polls — without the subjectivity of human rankings.
































































































