For years, the AB Score has been the foundation of everything on this website, from contract projections and GM evaluations to coaching rankings and team audits. Until this summer, however, it lived almost entirely inside the NHL. That changed after the 2026 NHL Draft in Buffalo. Since then, hockeyfreeforall.com has been building a prospect evaluation system that extends AB to 19 professional and junior leagues around the world, connects every one of them back to the NHL, and produces player cards, team views, and coaching evaluations for all of them.

The system has already powered several pieces on this website, including the Gavin McKenna projection, the comparison between McKenna and Ivar Stenberg, and the prospect cards for Alexander Command and Adam Novotny. This article explains how it works, step by step, and how it evolved into its current form.

Step One: Building the Data Foundation

Every AB Score starts with data. The prospect system draws its league data from InStat, covering the NHL and 19 other leagues: the AHL, ECHL, KHL, SHL, Liiga, DEL, DEL2, HockeyAllsvenskan, Mestis, the Swiss National League and Swiss League, Slovakia’s Tipos Extraliga, the MHL, Sweden’s J20 league, the three CHL leagues (OHL, WHL, and QMJHL), the USHL, and the NCAA. Each league is stored in its own workbook, and every player-season is scored using the same AB framework used for NHL players.

Combining 20 leagues into one database created problems that don’t exist in a single-league system. The same player can appear under different spellings across sources, such as Matthew and Matty Beniers, Oskar and Oscar, or names with and without accents. To solve that, every player is matched through a four-tier system: an exact match, a formatting-normalized match that strips accents, suffixes, and punctuation, a nickname match for common pairs like Matt and Matthew or Mike and Michael, and finally a near-identical spelling match that requires both the first and last name to be within a single character of each other. A separate check confirms that a player’s seasons age forward consistently, which prevents two different people who share a name from being merged into one career.

For college hockey specifically, the system also scrapes live rosters from all 63 Division I programs, so current rosters, midseason transfers, and incoming freshmen are always up to date. That matters because a true freshman’s most recent AB Score usually comes from a junior league, and a transfer’s comes from his previous school. The system searches for each player’s most recent season in any league, not just the NCAA, so the players who need evaluation the most are never dropped.

Step Two: Replacement Level by League

An AB Score only means something relative to the league it was earned in. A 1.00 in the WHL and a 1.00 in the SHL describe very different players. Every league therefore carries its own replacement level, which serves as the baseline for what a typical fringe player produces in that league. On the player cards, that baseline appears as a dashed line on the career trajectory. McKenna’s 8.85 AB Score in the WHL in 2024-2025, for example, sat more than seven times above that league’s replacement level, while his 2.23 at Penn State sat nearly five times above the NCAA’s.

Same-league comparisons, such as a player against his own league’s replacement level or a coach against his own league’s peers, only require this step. The moment two different leagues are placed side by side, a second correction is needed.

Step Three: The Deflator System

The deflator is the heart of the prospect system. Raw AB Scores earned in different leagues aren’t directly comparable, because league strength varies enormously. A score that looks dominant in a junior league can translate to a modest number against NHL competition. The deflator corrects for that.

For every player who appeared in a given league and later played in the NHL, the model compares his average AB Score in that league to his average AB Score in the NHL. The gap between the two, calculated separately for forwards and defensemen, is that league’s deflator. Converting a raw score to its NHL equivalent is then a simple addition:

NHL-Equivalent AB = Raw AB Score + Deflator

The Original Weakness

The first version of the deflators used a simple average of every matched player’s gap, with no filter on how much NHL time that player actually had. Two problems came out of that approach. First, the individual gaps are extremely noisy. In nearly every league, the spread of individual player gaps is two to four times larger than the deflator itself, so a handful of outliers could pull the average in either direction. Second, a meaningful share of the matched players, as many as a quarter in some leagues, had only a single NHL season on record. A one-season stint can reflect a player’s true level, or it can be a brief call-up dominated by small-sample noise, and the original system weighted both identically.

The issue surfaced most clearly with Liiga defensemen. The original deflator for that group was negative 0.272, which implied that Liiga and the NHL were nearly equal in strength at that position. That didn’t match any reasonable understanding of the gap between the two leagues, and a closer look at the data confirmed that the average was sitting inside a much wider and noisier spread than the single number suggested.

The Fix

Two changes were applied to every league and position group. First, a player now only counts toward a league’s deflator if he has logged at least 20 NHL games, which removes the brief call-up problem directly.

Second, rather than switching every league to a median, which would discard useful information in leagues where the average was already reliable, the estimator is chosen individually based on the data. If a league has fewer than 15 qualifying players, the median is used automatically, since there isn’t enough data to trust an average against a few outliers. If a league has 15 or more, the model compares the mean and median relative to their standard error. If they differ by more than half a standard error, something is distorting the average and the median is used. If they’re close, the mean is kept. Across all 38 league and position combinations, that rule placed 30 on the median and eight on the mean. Those eight, DEL forwards, NCAA forwards, OHL forwards, QMJHL forwards, both USHL groups, and both WHL groups, were all cases where the two estimators already agreed closely. Under the corrected methodology, the Liiga defensemen deflator moved from negative 0.272 to negative 0.421.

Every deflator also carries a confidence label based on its sample size: Good for 30 or more qualifying players, Moderate for 15 to 29, and Low for fewer than 15. Low-confidence deflators should be treated as rough placeholders rather than precise corrections.

The Full Deflator Table

The current deflators for every league are listed below. The recommended value is added to a raw AB Score in that league to convert it to the NHL-equivalent scale, and n represents the number of qualifying players with at least 20 NHL games.

League	Pos	n	Estimator	Deflator	Confidence
AHL	D	292	Median	-0.377	Good
AHL	F	597	Median	-0.670	Good
DEL	D	19	Median	0.005	Moderate
DEL	F	92	Mean	-0.626	Good
DEL2	D	2	Median	-0.418	Low
DEL2	F	9	Median	-1.304	Low
ECHL	D	12	Median	-0.321	Low
ECHL	F	26	Median	-0.731	Moderate
HockeyAllsvenskan	D	12	Median	-0.145	Low
HockeyAllsvenskan	F	30	Median	-0.328	Good
J20	D	9	Median	-0.451	Low
J20	F	12	Median	-1.259	Low
KHL	D	57	Median	-0.233	Good
KHL	F	197	Median	-0.519	Good
Liiga	D	33	Median	-0.421	Good
Liiga	F	75	Median	-0.542	Good
MHL	D	3	Median	0.225	Low
MHL	F	14	Median	-1.106	Low
Mestis	D	3	Median	0.253	Low
Mestis	F	7	Median	-0.862	Low
NCAA	D	71	Median	-0.315	Good
NCAA	F	108	Mean	-0.597	Good
OHL	D	28	Median	-1.765	Moderate
OHL	F	50	Mean	-2.826	Good
QMJHL	D	8	Median	-1.495	Low
QMJHL	F	17	Mean	-2.616	Moderate
SHL	D	68	Median	-0.210	Good
SHL	F	152	Median	-0.482	Good
Swiss League	D	1	Median	-0.450	Low
Swiss League	F	8	Median	-0.694	Low
Swiss National League	D	34	Median	-0.776	Good
Swiss National League	F	127	Median	-0.708	Good
Tipos Extraliga	D	6	Median	-0.629	Low
Tipos Extraliga	F	22	Median	-0.468	Moderate
USHL	D	25	Mean	-0.258	Moderate
USHL	F	51	Mean	-0.663	Good
WHL	D	21	Mean	-1.844	Moderate
WHL	F	43	Mean	-2.331	Good

The table reveals several important patterns. The three CHL leagues carry by far the largest deflators, with OHL forwards at negative 2.826, QMJHL forwards at negative 2.616, and WHL forwards at negative 2.331. That reflects how much higher raw AB Scores run in major junior compared to the NHL. By contrast, the SHL (negative 0.482 for forwards), the KHL (negative 0.519), and the NCAA (negative 0.597) sit much closer to NHL level. For prospect evaluation, that difference is critical. A dominant CHL season and a solid college or SHL season can look very different on the surface and translate to nearly identical NHL-equivalent values.

The Deflators in Action

Every prospect card displays the calculation directly at the bottom of the card. A few recent examples show how it works.

Gavin McKenna’s 2.23 raw AB Score at Penn State, combined with the NCAA forward deflator of negative 0.597, produces an NHL-equivalent AB of +1.63.

Ivar Stenberg’s 2.14 in the SHL, combined with the SHL forward deflator of negative 0.482, produces an NHL-equivalent of +1.66. Although McKenna’s raw score was earned in college and Stenberg’s against men in Sweden, the deflators place their 2025-2026 seasons almost exactly level. The difference between them only emerges over a longer sample, where McKenna’s career NHL-equivalent average of +2.32 far exceeds Stenberg’s +0.74.

The Devils’ Alexander Command and the Canucks’ Adam Novotny show the effect of the deflators even more dramatically. Novotny’s 3.22 raw AB Score in the OHL was the higher of the two, but the OHL forward deflator of negative 2.826 brings him to an NHL-equivalent of +0.40. Command’s 2.22 in Sweden’s J20 league, adjusted by that league’s deflator of negative 1.259, lands at +0.96, more than double Novotny’s figure.

Translating Between Any Two Leagues

Because every league connects to the same NHL-equivalent scale, the system can also compare two leagues directly, even when very few players have ever moved between them. The NHL serves as the hub. A raw score in League A converts to League B’s terms by adding League A’s deflator and subtracting League B’s.

For example, a WHL defenseman with a raw 2.5 AB Score converts to 0.656 on the NHL-equivalent scale (2.5 plus negative 1.844). Translated into KHL terms, that becomes 0.889 (0.656 minus negative 0.233). The same underlying level of play requires a much smaller raw number in the KHL because the WHL runs so much hotter relative to the NHL.

Step Four: The Prospect Player Card

With the data and deflators in place, the system produces player cards for any player in any of the 20 leagues, built on the same design as the NHL cards. Each card displays the player’s current season snapshot, his career AB trajectory against his league’s replacement level, his NHL-equivalent AB with the deflator calculation shown, and a set of career comparables.

Jackson Smith’s card is a good example of how the pieces fit together. The 19-year-old Penn State defenseman, drafted 14th overall by Columbus in 2025, posted a 2.77 AB Score in the WHL in 2024-2025 before recording 27 points in 34 games and a 0.81 AB Score in his first college season. Applying the NCAA defenseman deflator of negative 0.315 produces an NHL-equivalent of +0.50. His trajectory also shows how different the replacement levels are across leagues, with the WHL baseline above 1.0 and the NCAA baseline below 0.5.

The comparables engine has also evolved. The original version selected comparables from players within a few years of the target player’s exact age, which kept shrinking the eligible pool and caused the same handful of names to appear repeatedly. The current version offers two modes. General mode, used for established players, draws comparables within a set age gap of the player’s own age. Prospect mode, used for draft-eligible players, draws comparables from a fixed age band, such as 16 to 18, and only uses seasons played before a comparable player’s own draft year, meaning his D-1, D-2, or D-3 seasons. That ensures a prospect is compared to what other players looked like before they were drafted, not after.

Prospect mode also produces a Projected Draft Round, which averages the actual draft round of the entire comparable pool rather than just the six comparables displayed on the card. Command’s comparables averaged Round 1.20, while Novotny’s averaged Round 2.12, which is why his card projected him as a second-round talent despite being selected 24th overall. For drafted prospects, the comparison cards go one step further and project an NHL outcome, averaging the first three NHL seasons of 20 comparable players. McKenna’s projected NHL outcome is 1.45 AB, compared to 1.36 for Stenberg.

Step Five: Team and Roster Views

The same framework scales up to entire rosters. For college programs, the system builds a full roster chart showing each player’s career AB Score projected onto the NCAA scale.

UMass’s chart is a good illustration. Players with direct NCAA history are shown on their own NCAA numbers, while players marked with an asterisk, primarily incoming freshmen and transfers with no NCAA history yet, are projected onto the NCAA scale from their previous league using the deflator system. That allows a coach, scout, or fan to see how an incoming recruiting class compares to returning players before a single college game is played. For the Minutemen, the newcomers stand out immediately. Melvin Novotny (3.02) and Maxim Masse (2.13), both projected from their previous leagues, rank first and second on the entire roster, and fellow newcomers Jasper Kuhta (1.42) and Egor Barabanov (1.03) also rank inside the top eight. Among returning players, Jack Musa (1.81), Jack Galanek (1.48), and Francesco Dell’Elce (1.45) lead the way. In total, 15 players on the roster sit above the NCAA replacement line of 0.33, and the chart makes it clear that UMass’s incoming class could reshape its lineup from day one.

Step Six: Coaching Evaluations Beyond the NHL

The AB Coaches study that ranks NHL coaches has also been extended to every league in the system, including a dedicated NCAA version that evaluates coaches on three components.

The first is Coaching AB, which measures how players develop under a coach. It combines Return IMP %, the percentage of returning players who improve in consecutive seasons (25% weight), Advantage %, the percentage of players who improve relative to their previous coach (20%), Average IMP (15%), Average Player AB (15%), and Average Range Score (25%). Because some coaches have far more data than others, each component uses a Bayesian adjustment that blends the coach’s own results with the league average based on how much evidence he has. A coach with only a handful of returning players is pulled toward the league average, while a coach with years of data is judged almost entirely on his own record. That also ensures a brand-new coach with no returning players resolves to the league average rather than an error.

The second is an NHL Quality Score, which measures what happens to a coach’s players after they leave. It combines the average NHL AB Score of his former players (40%), their average post-entry-level salary (35%), and their performance relative to the NHL replacement level (25%).

The third is a Draft Positioning Score, a leading indicator that credits a coach for the current strength of his roster based on where his players were drafted, using point values calibrated on draft classes old enough to have matured.

The three components are weighted equally to produce a Combined Score and Combined Rank.

Penn State’s Guy Gadowsky shows how the components can tell different stories. His Coaching AB of 0.2798 ranks 30th out of 100 NCAA coaches, and his Quality Score ranks 44th, as none of his former players have yet become established NHL players. However, his Draft Positioning Score ranks sixth in the nation. Seven of the 21 players on his current roster are NHL draft picks, including McKenna, the first overall pick in 2026. Combined, Gadowsky ranks eighth out of 100. His Return IMP % of 58.7% also exceeds both the study average and the Top 10 average of 57.4%, and his average player AB has climbed to 0.77 in 2025-2026, the highest of his tenure in the dataset. His key players include Aiden Fink (2.834), McKenna (2.482), and Charlie Cerrato (2.179).

Step Seven: Testing the System Against History

The final piece is accountability. The Draft Class Export calculates the pre-draft career AB Score for every skater in a given NHL draft class, applies the deflators, and produces a ranking of how the class would have been ordered by the NHL-equivalent AB alone, regardless of whether those players ever reached the NHL. Comparing those rankings to where players were actually selected, and to how their careers unfolded, is how the system is being tested and refined. The export is limited to the 2019 draft class and later, as earlier classes are missing too many of their key pre-draft seasons in the available data.

The same philosophy drives the new points projection model introduced in the McKenna article, which was backtested on the last four NHL seasons using only data that would have been available at the time.

Limitations

No system is perfect, and there are several limitations worth understanding. Each deflator inherits its own confidence level, so any translation through a Low-confidence league, such as Mestis, DEL2, or QMJHL defensemen, should be treated as an approximation. The system also assumes the NHL-equivalent scale means the same thing regardless of which league produced it, which is a working assumption rather than something that has been independently verified.

Finally, the deflators capture league strength, not role. A player who cracks an NHL roster often steps into a very different role than the one he had in his previous league, with less power play time and tougher defensive assignments, and that change affects his AB Score independently of the quality of competition.

Conclusion

In just a few months since the 2026 NHL Draft in Buffalo, the AB system has grown from a single-league NHL evaluation tool into a global prospect system that covers 20 leagues, translates every one of them to a common NHL-equivalent scale, and evaluates both players and coaches along the way. The work is ongoing, and the deflators, comparables, and projection models will continue to be refined as more players move between leagues and more data becomes available.

The goal remains the same as it has always been: to evaluate every player as objectively as possible, regardless of where he plays. Whether it’s a first overall pick at Penn State, a teenager in Sweden’s junior league, or a freshman at UMass, every player can now be measured on the same scale as the NHL according to AB.