glacica

glacica

what the hell is going on around here? - an explainer

26.03.17

here i will try to explain myself, what this is, what this isnt, and what im going for. but i want to start with a few disclaimers

  • METRICS ARE BUILT FROM PWHL PUBLIC DATA WHICH SKEWS TOWARDS OFFENSIVE EVENTS. AS SUCH, THESE METRICS DO NOT FULLY REPRESENT OR ACCOUNT FOR MORE NUANCED CONTRIBUTIONS. THE CONTRIBUTIONS OF DEFENSIVE PLAYERS IN HARD MATCHUPS, NET FRONT PLAYERS THAT CREATE TRAFFIC, STRONG FORECHECKERS ETC WILL BE UNDERRATED UNTIL THE DATA GAPS ARE FILLED
  • MANUALLY TRACKED GAME DATA IS IN PROGRESS TO HELP FILL THESE GAPS BUT IS TIME CONSUMING AND ALSO OPENS A WHOLE OTHER CAN OF WORMS
  • ALL METRICS ARE BASED ON ALL SITUATIONS OF GAMEPLAY. POWERPLAY RINGERS CAN OVERSHADOW EVEN STRENGTH WORKHORSES. MORE REFINEMENT FOR GAME STATE SPECIFIC FILTERING IS IN THE PIPELINE.
  • SOME METRICS SHOWN MAY USE NHL WEIGHTING TEMPORARILY DUE TO A LACK OF SAMPLE SIZES, LACK OF CITABLE METHODOLOGY, LACK OF PUBLIC DATA, OR WHATEVER ELSE BUT THIS WILL BE COVERED IN EACH RELEVANT SECTION AND SUBSEQUENT EXPLAINERS
  • PLEASE TAKE EVERYTHING WITH A GRAIN OF SALT! I AM NOT A MATHAMAGICIAN I JUST HAVE WEIRD FIXATIONS AND IM DOING MY BEST

WHATS THE POINT THEN

currently the PWHL mainly publishes counting stats, but more can be done even with the limited public data available. this is a work in progress for recalibrating standard analytics methodology to women's hockey and hopefully build a robust foundation going forward.

NON-BOXSCORE STATS EXPLAINED

Rate Stats

GS/60 - gamescore/60 is a measurement of player impact based on the weighting of events per 60 minutes played. players are rewarded on a weighted scale for goals, assists, shots, hits, faceoff wins and deducted for faceoff losses and penalties taken. we are using an adaptation of Dom Luszczyszyn's framework, except i bumped the value of a secondary assist down. i will go into all of this deeper in its own post

P/60 - all points scored per 60 minutes of ice time. general metric of offensive production

G/60 - goals scored per 60 minutes of ice time. shows scoring rate

A1/60 - primary assists per 60 minutes of ice time. primary assists can be a stronger indicator of offensive contributions as secondary assists can be noisy

S/60 - shots taken per 60 minutes of ice time. shows the pressure a player generates even if it doesn't yield results

Quality Stats

s% - raw shooting percentage, goals per shot taken

xS% - stabilized bayesian shooting percentage that accounts for sample size. pulls extreme numbers toward the expected average so a player who scored on their only 2 shots doesn't look like a 100% shooter. flagged for players with limited sample sizes. currently the PWHL does not have a large enough data sample of shooting talent to reflect how skill distributes across the league, so the variance of shooting skill between players is borrowed from NHL metrics. we are using the current cumulative PWHL average shooting percentage to place the "average" level on the spread of NHL shooting talent. this is calculated for forwards and defenders separately. this overlap should be resolved in a few seasons when we have enough data. full explainer coming in a follow up post.

GASA - Goals Above Shot Average; uses the above bayesian shooting percentage to ask, given a player's shot volume, are they scoring more or fewer goals than expected? a positive number means scoring above expectation for their shot volume, negative means below

got all that? ok

HERES THE PLAN OVERALL

hax the mainframez ✅
- data displayed scraped from pwhl end points as per Isabelle Lefebvre's PWHL data reference project https://github.com/IsabelleLefebvre97/PWHL-Data-Reference

build out the datasets
- manual tracking is in progress to add zone entries/exits, shift times and pairings, and expand on scoring chances, possession time, and other untracked game events

add richer metrics
- once the quality game data is tracked we can really get into the madness! true xG, corsi, fenwick, wowy metrics, etc to come

fun stuff in the pipeline
- visualizations, literal and abstract
- ;)

stay tuned!
glaci