Skater Similarity Scores

The similarity scores found on the site are determined by calculating a weighted euclidean distance between one player and all other players who:

  • Play the same position (Forward / Defenseman)
  • Are +/- 2 years of the player’s age.

The following metrics are used for determining the similarity of two players:

  • All Situations:
    • GP, Take GAR, Draw GAR, SPAR, xSPAR
  • Even-Strength:
    • EV TOI, EV TOI/GP, EV Zone % (OZS / (OZS + NZS + DZS + OTF))
    • EVO GAR, xGAR EV RF, xGAR EV QF, xGAR EV Shoot, RAPM EV GF60, RAPM EV CF60, RAPM EV xGF60
    • EVD GAR, xGAR EV RA, xGAR EV QA, RAPM EV CA60, RAPM EV xGA60
  • Powerplay:
    • PP TOI, PP TOI/GP, PP Zone % (OZS / (OZS + NZS + DZS + OTF))
    • PPO GAR, xGAR PP RF, xGAR PP QF, xGAR PP Shoot, RAPM PP GF60, RAPM PP CF60, RAPM PP xGF60
  • Shorthanded:
    • SH TOI, SH TOI/GP, SH Zone % (OZS / (OZS + NZS + DZS + OTF))
    • SHD GAR, xGAR SH RA, xGAR SH QA, RAPM SH CA60, RAPM SH xGA60

All 3-year and 6-year periods available for each player are summed and converted to Z-Scores (around the positional mean of the respective summed datasets).

We weight the metrics in each strength state by the percentage of league time on ice each strength state makes up:

  • All Situations: 100%
  • EV: 80%
  • PP: 10%
  • SH: 10%

The final calculation looks like this (where “Player” is the selected player and “Comp” is player being compared to the selected player):

    sqrt(
      (Player GP - Comp GP)^2 * 1.0 + ... 
      (Player EV TOI - Comp EV TOI)^2 * 0.8 + ...
      (Player PP TOI - Comp PP TOI)^2 * 0.1 + ...
      (Player SH TOI - Comp SH TOI)^2 * 0.1 + ...
      )

The resulting distance measures are then re-scaled to be displayed from 0% – 100% using the following calculation (where the min and max are taken from the entire respective 3-year and 6-year datasets):

  • (x - min(x)) / (max(x) - min(x))

 

← Return to Glossary Home