We’ve established that marginal goals (or goals above replacement, if you will) will be the basis for how we allocate point shares to players. But how do we use marginal goals to credit players for their on-ice contributions? First, we’ll learn how to allocate offensive point shares to skaters. We need a simple way to measure the offensive contribution each skater made in a season. Points would be the primary measure most would suggest first – it factors in both goals and assists, and we know it is one of the few statistics tracked since the NHL opened its doors. But we also know goals and assists are not equal in generating offence; goals require a more direct impact on a scoring play and are awarded less frequently than assists. As a result, we will use a hybrid statistic – goals created – initially developed by Alan Ryder in his paper, Player Contribution. Ryder’s goals created awarded exactly 50% of the credit of a team’s goals to the individual goal scorer, the remaining 50% of the credit to the team’s assists, and distinguished even strength, power play, and shorthanded goals. We will use Hockey Reference’s version of goals created, which simply pools all of a team’s goals together in the calculation and does not force a precise 50/50 split. Nonetheless, each method rightfully awards the goal scorer more credit for creating the goal than those who earned assists.
|NEW STATISTIC:Goals Created (GC): An assignment of credit of a team’s total goals calculated by adding goals scored to 0.5 times assists, then multiplying by team goals divided by team goals plus 0.5 times team assists; the sum of player goals created is equal to the sum of player goals scored by team.|
Let’s illustrate by continuing the example of Duncan Keith in 2009-10. That year, Keith scored 14 goals and scooped up 55 assists. The Blackhawks as a team scored 262 goals and were awarded 464 assists. Using Hockey Reference’s goals created formula, Keith is credited with 22 goals created out of Chicago’s 262 goals scored, properly reflecting his split of goals and assists in the team’s offence.
So, from an offensive perspective, Duncan Keith created 22 goals in 2009-10. But we want to identify how many marginal goals for he created, Keith’s value being only what he contributes above what a replacement could achieve. To identify Keith’s marginal goals, consider two pools: one contains total goals created by defencemen in the NHL; the other contains every minute played by a defenceman. Knowing the rate at which goals are created by all defencemen in 2009-10 and that 5/12 of any team’s contributions are for offence, we can establish a baseline of what a replacement defencemen would produce with Keith’s ice time. In this case, replacement level is calculated at 6.5 goals created. Since Keith created 22.0 goals, we can say he created 15.5 marginal goals (i.e., goals created above replacement threshold). As we know the number of marginal goals and the number of standings points league-wide, we can easily convert Keith’s 15.5 marginal goals to 6.3 offensive point shares. Therefore, we can conclude that
|POINT SHARES REVIEW:#4: Goals created are the basis for the assignment of offensive point shares.|
Keith’s offensive contributions helped the Blackhawks earn 6.3 points in the standings.
|POINT SHARES REVIEW:#5: The rate at which skaters create goals can be used to establish replacement level by position for offensive contributions.#6: A player’s marginal goals (i.e., above or below replacement) can be converted to point shares based on the league’s total standings points and marginal goals.|
This method can then be applied to every skater in the NHL for every season. By using goals created specific to a particular year when calculating marginal goals, we eliminate the noise of offensive environment. In 1985-86, Chicago’s Doug Wilson created 22 goals, equaling Keith’s 2009-10 output; however, Wilson earns only 4.1 offensive points shares (to Keith’s 6.3), replacement level for offence being higher given the frequency of goals that year. When it comes to forwards, our pool of goals created and time on ice will be only those of forwards; this is necessary to calculate marginal goals on the basis that forwards create more goals than defencemen by virtue of their position. Keith’s 2009-10 teammate, Jonathan Toews, created three more goals than Keith that year in 661 fewer minutes. Given replacement level by goals created for forwards is higher than defencemen, Toews earns only 5.6 offensive point shares. Thus, the point share system smoothly adjusts for skater position, as well as offensive environment by season.
With our understanding of offensive point shares secured, let’s apply it to the 2009-10 Hawks to see how the system fits. Going forward, we will use point share abbreviations as follows: OPS (Offensive), DPS (Defensive), GPS (Goalie), and where totaled, simply PS
TOP 10 OFFENSIVE POINT SHARES, CHICAGO BLACKHAWKS, 2009-10
[+/- = Plus/Minus; GC = Goals Created; TOI = Time on Ice (i.e., Minutes Played)]
The order and point share values pass the smell test. It reads like a leading point scorer list, but with several critical improvements – greater value assigned to goals than assists; production by forwards and defencemen properly valued relative to position; rewarding efficiency by factoring total minutes played into earning value; and elite performance better rewarded, value being earned for productivity above replacement level, not simply for playing at a mediocre level.
|POINT SHARES REVIEW:#7: Prior to 1998-99, weighted games played by position are used in lieu of minutes played.|
Now, we referenced ice time as an input in calculating marginal goals. However, the NHL did not track individual player ice time until the 1998-99 season. So, we need to come up with a method for the league’s first 81 seasons, absent ice time data. Concessions of this nature are a requirement the further back we go; when the best information is not available, we need to work with the best that is available. Prior to the publication of ice time, we’ll use games played as a proxy for minutes played. We can safely assume that in his 1974-75 Art Ross-winning season Bobby Orr did not play the same number of minutes as teammate Al Sims (no offence, Al). Highly utilized players, such as Orr, will gain some advantage on 21st century players in earning offensive point shares. But as we work through the methodology, there will be other data points that create small advantages or disadvantages when assigning credit for some players. The system works with the best obtainable inputs, and we acknowledge and accept these caveats.