Q&A

To date, we’ve established a thorough understanding of adjusted games played, goals, assists, points, and pace. We recognize the purpose. We know the mechanics. We see the impact. We are in a position to deploy these measures toward player evaluation. To close the book on Adjusted Statistics, we’ll take a moment to challenge Hockey-Reference’s system by reviewing some potential questions and answers you may have built up. No model, of course, is perfect, and certain assumptions and exclusions are intrinsic to developing one.

The view at a Tampa Bay Lightning game in the 21st century has changed considerably from the four-team NHL established over a century ago. "Game Winning Goal Celebration" by Photomatt28 is licensed under CC BY-NC 2.0

Even before adjustment, hockey’s basic stat system creates many questions. Why is an assist as valuable as a goal in total points? Why are empty net goals equal in value to overtime goals? Why is a primary assist equal to a secondary assist? Why are two assists awarded on a scoring play? Why not three, or one? The existing universal hockey point scoring model can easily be challenged or nitpicked. Ultimately, the guidelines are put in place, and hockey fans have largely chosen to accept the results. The adjusted point scoring model is no different, only there has been meaningful consideration to improve familiar box score stats via the four adjusting factors.

We’ll focus on trying to poke holes on the approach to adjusted statistics both to improve our understanding and justify the decisions. So, let’s challenge the system:  

  • How can a level playing field be created given the NHL of 20, 50, or 100 years ago is unrecognizable today?

No one can say the statistics subject to adjustment have been perfectly neutralized. We can say that carefully considered steps drive the methodology, and that significant context has been added to the raw numbers in an efficient way. The four-adjustment process addresses the obvious shortfalls in the uncontextualized counting system in the NHL’s records. These adjustments neutralize information and add value to any discussion in a way not previously attempted at this level of detail or scale.

  • Why aren’t advancements in player training, equipment technology, power play frequency, overtime format, or goaltending ability factored into the adjustments?

Oh, but they are. The first two adjustments for scoring environment comprehensively account for all the above. No matter how the players themselves look on old grainy footage versus 5K quality, bringing every player’s output to a six-goal, 10-assist average game equalizes the level of difficulty in scoring points in that specific season. This adjustment effectively balances the generational skater-goalie equilibrium in two calculations.

  • Why isn’t average time ice factored in for seasons it is available (1998-99 to present)? Wouldn’t this provide a better proxy for available minutes over roster size adjustment?

The purpose of the adjustment system is not to create a “per 60-minute” stat. Adjusted statistics are still counting statistics, only contextualized for era. Whether Paul Kariya played 20 or 25 minutes per night, so long as his Mighty Ducks had 18 available skaters, then his scoring output is not subject to discounting. Roster size adjustments account for league rules that limited the number of skaters in uniform pre-1982-83, thus creating inequity given each player comprised a greater share of the roster allowed at the time.

  • Recognizing the disparity in scoring and assist environments, is it reasonable to assign roughly 7.5 times more assists to the 1928-29 season than to the 1985-86 season?

One of the most noticeable figures in adjusted statistics is the representation of the period from 1925 to 1932 on the leaderboards; in fact, 13 of the top 20 single-season adjusted assist totals can be found in this seven-season span. Any instance of a period being overrepresented brings into question the system in place, as neutralization would normally balance population representation. Frank Boucher’s league-leading 16 assists in 44 games in 1928-29 translating to 119 adjusted assists in 82 games appears exaggerated at first glance. We must consider, though, that an assist in that season was unfathomably scarce – only 385 were issued in 220 total games; in 1985-86, in the same number of games, you would see an average of 2,889 assists!

So, the math and logic certainly work, but in the case of the first two decades of the NHL, the assist figures are so heavily extrapolated that anomalies are inevitable. Much like a hot start in October by a rookie (say 6 goals in 12 games), pundits start projecting 40-goal seasons across a small sample. So, when dealing with a scarce commodity like a 1928-29 assist (or goal for that matter), the adjusted single season outputs must be taken with a grain of salt. The relatively low number of games played, players on a team, and goals and assists in the earliest days of the NHL yield some eye-popping figures. These figures average out over a player’s career – only three of the top 30 adjusted assist per game leaders played in the NHL’s first quarter century. A multi-season view is more appropriate when analyzing player production from any era, but the impact is pronounced in the NHL’s earlyyears.

  • How are players credited for the lost 2004-05 season that was not played due to the NHL labour issues?

As there are no in-season statistics to adjust for, this lost season is not included in adjusted statistic figures. There could be an argument to assign the adjusted results of the 2003-04 or 2005-06 seasons that straddled the lockout. However, when no games are played at all, there is so much unknown about what could have happened. Would a player get hurt or abstain from injury? Would a player have debuted or retired? What would the curve of development or decline look like? Did a player ultimately extend their career by virtue of the rest? Adjusted statistics add context to an actual season’s performance, but to manufacture an entire season is outside the reasonable scope of the system.

While the Carolina Hurricanes won the 2006 Stanley Cup, how should adjusted statistics treat the complete lost season before? "71116006PS078_Edmonton_Oile_1_32_29_AM" by justinhenry is licensed under CC BY-SA 2.0

Certainly, you may have other questions about adjusted statistics and any of the concepts discussed or created on this site. Twitter is a great place to reach out my way, both to enhance understanding and to (respectfully) challenge hockey information.

Much like a player having one more secondary assist or one more empty-net goal in a season or a career, one more adjusted point is not intended to break ties on debates of all-time greats. The adjusted results are intended to provide much-needed context to raw numbers in a systematic way. The care taken in introducing, illustrating and illuminating the adjustments in the preceding sections of this site help strengthen the methodology so it can be applied seamlessly across decades of performances.

We still have a number of exciting ideas to discuss on this site, following along the anchored menus at the top of each page. The thrilling aspect of any new or different tool is using it to shine light on the fascinating outcomes of a level playing field. On the main Home blog, you’ll regularly see articles posted where the adjusted results of over 100 years of NHL history are analyzed, discussed, and explored, including entries into the Adjusted Hockey record book. If you’re following along sequentially, we’ll next explore the concept of Adjusted Point Shares, starting with an interesting look at NHL eras.

All data from Hockey-Reference.com.