Numbers Game – Statistical Analysis & Management

Working with a company that has been family-owned and operated for several decades has provided me with a sort of second-hand history lesson in business-practice and management. Through the lunch-room and water cooler stories of the family's octogenarian patriarch and his son (the company’s current President), all with a heavy coating of my father’s 35 years of overlapping experience on both sides of the retail game and his tendency to comment on and contextualize everything I am exposed to, I’ve been privy to a mini crash-course in the experiences, culture and evolution of a rapidly changing and increasingly challenging business landscape.

A micro sample even within our respective market, the lessons I have learned are nonetheless universally applicable to competitive business models, which must perennially react, evolve and strive to be ahead of the curve in their given sector. Among the leading trends of the past couple decades, in direct correlation with the advancement and proliferation of computer capabilities of course, is the increased emphasis on and sophistication of market research and statistical analysis. Gone are the days in which sales was a game of throwing mud at the wall to see what would stick; today, every item must be carefully studied and micro-managed in order to exploit every shred of opportunity available in a crowded and inherently limited marketplace.

Similarly, NHL hockey has become increasingly professionalized at all levels (from management to coaching, scouting, even player personnel) in response to a progressively more competitive market for all involved. Magnifying and intensifying these pressures, the CBA and its Salary Cap structure place even greater onus on roster decisions, as shortcomings and mistakes can no longer be corrected by throwing money at the best available free agents. Where assets remain limited and funds are suddenly capped, a system that was once dominated by deep wallets suddenly and necessarily converts to an obsessive search for value.

Evinced by the emergence of a burgeoning field of study devoted to the statistical analysis of hockey (see Hockey Analytics, Puck Prospectus, Behind the Net, Hockey on Paper, Objective NHL, Hockey Numbers, etc.), the trend towards more sophisticated means of talent evaluation has also spawned the solicitation of NHL organizations by consulting groups such as Coleman Analytics. The pioneering endeavour of statistician, Richard Coleman, and former Winnipeg Jets and Chicago Blackhawks GM, Mike Smith, Coleman Analytics is breaking ground in what could prove to be a viable long-term resource for all NHL organizations.

The key to analytics, which Smith himself is careful to distinguish, is that they are not simply statistics. Stats are available in near infinite variations, limited only by the ability to objectively measure some facet of the game, but raw data can only tell you so much: the significance of analytics, on the other hand, lies in the ability to shape, contextualize, and draw relevance from the numbers. Analytics look to dig beyond the surface to discern the objectivity behind the numbers and to engineer complex new criteria by which a player’s relative impact or value can be assessed. A terrific complement to the work of an NHL GM (already tasked with formulating such guesses), analytical studies effectively present mathematically supported conjectures.

Numbers on a spreadsheet can never replace the eyes and instincts of managers and evaluators though, nor should they; according to Smith himself, his services are not to be relied upon independently of traditional forms of evaluation, but as an added layer of intelligence and confirmation (or refutation) of what you already think you know. The bottom line is that NHL organizations, like other businesses, have a responsibility to seek out every possible edge they can in order to best position themselves within their marketplace, and this new breed of analytical consultant may offer just that (Smith claims that, in 3 seasons in the NHL, 17 of 20 teams using his services have reached the playoffs).

Most infamously instituted by Okland A’s manager, Billy Beane, in the form of Sabermetrics, the analytical movement has been slower to take hold in hockey than in other major team sports. This, Smith suggests, is at least in part the result of a ‘conservative’ hockey world that is often slow to change, but one whose general acceptance of analytics as a practical tool is bound to increase as the field of study reaches new levels of depth and refinement and, more importantly, as it becomes associated with success.

At the heart of it all, however, the Salary Cap is the catalyst that made analytics a viable and desirable resource for NHL teams in the first place: putting the onus on GM’s to assemble a winning roster with capped funds, the new structure of the NHL turns GM’s from simple spenders (particularly in the larger markets) to high-risk investors who must very selectively divvy up their resources in order to obtain the highest possible returns. Players thus effectively become stocks, thoroughly studied in terms of their relative market value, expected yield and cost-effectiveness.

Pivotal in determining a solid investment? Structured analytics, of course.

Numbers Game - The Curious Case of Mikhail Grabovski

In these, the Dog Days of the NHL off-season, I’ve come across multiple Toronto  Maple Leaf fans with a rather optimistic view of Mikhail Grabovski. Knowing the player fairly well, and having watched him play over a period of time, I found myself rather surprised by the idealistic perception of his forward potential in the NHL and with the Leafs.

One of the simplest and most invaluable lessons learned in the early stages of my career in business (as a vendor of domestically retailed product) is that results rule the day. As my tremendously experienced father/boss/mentor repeats time and time again, “there are a lot of stories in the naked city, but numbers never lie.” Organizations (like the one I represent in sales) can claim as many reasons as they like as to why their product is going to be outstanding, or as many excuses as they like as to why their product didn’t perform as well as anticipated, but one thing stands above all the stories and substantiates reality—numbers.

More interesting still, these numbers don’t change; a product’s performance in one setting will very closely dictate its performance in any comparable setting and, with proper historical knowledge and a small subset of relevant data, can be forecasted to within a few percentage points of its results for the total season based on a few simple statistics and trends.

Though admittedly far more complex than the retail sales of a given product, the development and contributions of a given hockey player can be roughly forecasted based on extremely similar principles as those found in the retail game. Pierre McGuire’s comments regarding Cristobal Huet immediately come to mind, as he repeatedly suggested that Huet’s suddenly All-Star worthy play as the Canadiens’ goaltender, in his 30’s, was probably more of a blip than an indication of his true level of play. A product, my father says, does not suddenly become good; it either is good or it isn’t, and the proof is in the pudding.

Bringing this all back to the case of Grabovski, it struck me as being completely out of line that segments of Leaf-nation believe he may score 30+ goals next season, though a quick glance at his career stats reveals that he has never scored more than 20 in any single season as a professional.

“Yes,” Leafs fans will protest, “but he’s just getting started at this level.” To that I must defer to Gabriel Desjardins and his work on League Equivalencies, where he begins by offering the statistical tidbit that NHL players actually peak in terms of point-production between the ages of 23 and 26, and that their point-per-game production reaches 90% of its peak value even earlier (Grabovski is 25). The depth of his research, however, focuses on the average statistical translation of players’ point-production from one league to another. As with comparing results at Wal-Mart versus those at Zellers, for example, different settings or leagues have different standard values by which similar data may thus be extrapolated. For players with little NHL experience, this study and its findings therefore serve as a crude barometer of what may be expected of a given player based on his production at another level.

For AHL players making the jump, a point in the American Hockey League will equal, on average, 0.45 points at the NHL level. Grabovski’s 25 goals and 74 points in 78 total games in the AHL therefore suggests he should score, per 82-game NHL season, about 13 goals and 37 points. As for his experience in the Russian Elite League, where one point equals 0.91 points in the NHL, Grabovski’s 10 goals and 27 points in the 2005-06 season (his best of 3, statistically) suggest an output of about 16 goals and 44 points per 82 games at the NHL level. Having already reached the age at which most players see their production peak, and with a relevant set of historical stats that suggest 20 goals may in fact be a relatively high output, the numbers are clearly stacked against Grabovski ever attaining the 30-goal mark.

“Yes,” Leafs fans will protest, “but he’ll see more time on the PP this year.” While potentially impactful on his overall point totals, Grabovski’s elevation within the ranks of the Toronto Maple Leafs’ roster has to cut both ways. Yes, he will now very likely centre the first PP unit, but he must also now contend with opposing teams’ #1 defensive matchups on a nightly basis. In the end, the issue of dealing with such uncertain variables is both tricky and misleading, and so I will turn once again to a slightly more empirical method of projection of the expected output of a player, by way of ‘Triumph’ at Hockey on Paper. Determining the estimated production of a player moving from one team to another by comparing the two teams’ offensive stats, Triumph’s formula can be adapted as “a reasonable ballpark metric” for players entering another season with the same team.

In this case, changes to the Maple Leafs’ roster from 2008-09 to 2009-10 become relevant in the projection of individual goal and point totals, as the Burkeian shift in team identity and personnel may have a significant impact on the Maple Leafs’ overall offensive output. Building from the backend with big, physical, stay-at-home defensemen, Burke and the Maple Leafs have temporarily abstained from acquiring top-end offensive talent and, with the departure of their #1 centre in Nik Antropov, have let go of one of their top point-producers. Should Tomas Kaberle be moved, as rumoured, it would appear increasingly unlikely for this Maple Leafs squad to generate much in the way of offense, with very little reason to believe they will (as a sum of their parts) match last year’s 250 goals for. While this last look does nothing to necessarily discount Grabovski’s potential production next season, it certainly puts a caveat on it: a player’s production may spike when introduced to a stronger overall setting or when suddenly surrounded by greater support, but the reverse is also applicable.

There exists, of course, an as of yet undetermined myriad of factors that can and will play a role in Grabovski’s end-of-year goal and point totals; the bottom line, however, is that there is at this point nothing more than stories to validate the opinion that Grabovski has the potential to reach 30 (or even 25) goals next season. Until those stories are played out in next season’s scoresheets and bylines, they remain nothing more than fantasy and unsubstantiated hype.