devils_data
  • Articles
  • PlayerCards
  • Weekly Reports
  • About Me

Articles

Introducing: GameScore

2/22/2017

0 Comments

 
To find GameScore stats, navigate to the Playercards page.

INTRODUCTION
Introducing: GameScore, a “catch-all” statistic that gives a relative idea of how a player performed on any given night in one convenient metric.
The number is produced using a weighted formula that incorporates some statistics that are commonly found in box scores and some stats that some may consider fancy.  This idea of a catch-all statistic is not original and has been attempted by many before myself. In particular, I’d like to mention the work by Dom Luszczyszyn (@domluszczyszyn ) who writes for hockey-graphs.com. Just as I gained inspiration from his work, he acknowledges the work by 538 that details the origin of a basketball gamescore and work by Bill James that details the origin of a baseball gamescore. My formula is similar to Dom's with one major difference: Dom’s formula is intentionally narrowed in scope to include only simple stats included in any box score with one exception (corsi) whereas my formula incorporates a few statistics from the advanced community in conjunction with the simple box score stats and excludes faceoff stats (unfairly favored centers). Neither my work nor his is inherently “better” but it is important to recognize the differences and understand the stories that these stats attempt to tell.

METHODOLOGY
My GameScore metric uses 11 weighted stats to produce a final Y variable that I call "GameScore." There were two major decisions to be made when creating this metric: 1) What statistics to include and 2) How to assign weights.

I summarize both decisions in the chart below:
Picture
Perhaps the first thing you will notice is the color-coding. In my opinion, there are three tiers of statistics in the formula, separated by the degree to which they impact the game. To better distinguish these classes, I gave each one a color.

The first tier (blue) has 4 stats: Individual Goal Scored (Goal Scored by the Player Himself), Goal Differential (Goals For - Goals Against), Individual Primary Assist (Primary Assist Registered by Player Himself), and Individual Secondary Assist (Secondary Assist Registered by Player Himself). The most tangible way in which a player can affect the outcome of a game is to score a goal, therefore this stat receives the highest weight of 1.0. The next most influential way in which a game can be affected is by a player and his linemates outscoring their opponents while on the ice. Therefore, Goal Differential receives the next highest weighting of .75. Individual Goals are rated higher than Goal Differential because it is possible that a player may have little-to-no effect on a goal being scored by his linemates, but be credited with a Goal-For regardless. I want to reward a player who directly impacts the game over one who may benefit from talented linemates. Individual Primary and Secondary Assists receive the next highest weights at .70 and .60, respectively. They are the subordinates of goal-oriented stats because assists have a degree of separation from the actual goal scoring and therefore are not as impactful as the tally-maker.

The stats in the second tier (green) increase the likelihood of first tier stats occurring. They are: Scoring Chance Differential (Scoring Chance For - Scoring Chance Against), Individual Shots For, and Corsi Differential (Corsi For - Corsi Against). They are prioritized in order of their likelihood to produce a scoring event. Scoring Chances are more dangerous than Shots which are more dangerous than Corsi Attempts and the weighting progression is .25, .20, .15, respectively. There is a large weighting discrepancy between Tier 1 and Tier 2 stats because of the difference in magnitude of their overall affect on the game’s outcome. The only stats that truly affect the scoreboard are goals and assists, and therefore the formula must weigh the statistics with relative importance.

The third and final tier (yellow) contains measures that affect the outcome of Tier 2 stats. Each stat either decreases the likelihood of an opponent from producing a Tier 1 or 2 stat or increases the likelihood of the player or his teammates adding to their own Tier 1 or 2 tallies. The four Tier 3 stats are Individual Penalty Differential (Individual Penalties Drawn - Individual Penalties Taken), Individual Hits Differential (Hits Delivered - Hits Taken), Individual Takeaway Differential (Takeaways - Giveaways), and Individual Blocks. Penalty Differential has a higher weighting because of its prolonged effect over the course of the ensuing powerplay whereas hits, takeaways/turnovers, and blocks are static events with less likelihood to permanently affect the outcome of the game. Individual Penalty Differential receives a weighting of .10 and the final three stats receive a weighting of .05. 

The final formula reads:

GS = (
(GF-GA)*0.75) + (G*1.0) + (A1*0.7) + (A2*0.6) + ((CF-CA)*0.15) + ((SCF-SCA)*0.25) + (ISF*0.2) + ((IHF-IHA)*.05) + ((ITKA-IGVA)*0.05) + (IBLK*0.05) + (IPENDIFF*0.1)

SHORTCOMINGS
“Tell-All” Stats are rarely ever what they are designed to be. GameScore does not take into account Quality of Teammates/Opponents, Time On Ice, Score Effects, Home/Away Effects, etc. There is also no minimum or maximum score to achieve; all scores are to be judged relative to other scores. A score of 6.4 would mean nothing if one did not know that the mean score was .82, the upper quartile score is 2.3, and the lower quartile is -.85 (all facts). 

RESULTS
The best performance by means of GameScore this season was Travis Zajac's hat trick on December first which was supported by strong Tier 2 and Tier 3 stats.

The worst performance by means of GameScore this season was Kyle Quincey's December 8th outing where he had no points, allowed 3 goals, was buried in scoring chances and corsi, and had little success in Tier 3 stats. 


CONCLUSION
To summarize, this stat was created to provide a one-glance assessment of single-game performance by a player. It is intended to be used comparatively with other player's GameScores to gauge relative magnitude and is not perfect. GameScore has shortcomings when considering factors like Quality of Teammates/Competition, Time on Ice, and Score Effects, and Home/Away Effects. To see how each player on the Devils has been performing relative to GameScore, go to the Playercards page and play with the interactive tool.


*All stats were taken from Corsica.hockey.com/skater and GameScore can be calculated by you by following these directions:
  1. Go to corsica.hockey.com/skaters
  2. Click Custom Query in the top left
  3. Indicate a specific Team or Select All in the Team field, "All Situations" in the Strength State field, select all reports in the order they are provided in the dropdown menu, amd indicate the correct date (10/12/16-present).
  4. Uncheck Aggregate Games
  5. Download as CSV
  6. Open in Excel
  7. Create a new column and use the formula below:
    1. =((S2-T2)*0.75)+(AM2*1)+(AN2*0.7)+(AO2*0.6)+((G2-H2)*0.15)+((AA2-AB2)*0.25)+(AR2*0.2)+((AY2-AZ2)*0.05)+((BB2-BA2)*0.05)+(BC2*0.05)+(DJ2*0.1)
  8. Check the formula's cell alignment with the intended formula design:
    1. ​GS = ((GF-GA)*0.75) + (G*1.0) + (A1*0.7) + (A2*0.6) + ((CF-CA)*0.15) + ((SCF-SCA)*0.25) + (ISF*0.2) + ((IHF-IHA)*.05) + ((ITKA-IGVA)*0.05) + (IBLK*0.05) + (IPENDIFF*0.1)
  9. Done.
0 Comments



Leave a Reply.

    Categories

    All
    Ad Hoc
    Weekly Goalie Report
    Weekly Skater Report

Home

About

Articles

Playercards

Contact

  • Articles
  • PlayerCards
  • Weekly Reports
  • About Me