Q&A about WhoScored Ratings with Cristiano Acconci
A few months ago, the 538 podcast “Hot Takedown”[1] prompted some Twitter discussion about the WhoScored ratings and what exactly they represent. WhoScored have a page of explanation. It’s not explicit, but does give a little bit of insight into how they are calculated. Each player starts out at a 6.0 on a 1–10 scale, and various events during the game raise or lower that rating. At the end of the game, the winning team gains points; GKs, Defenders, and Midfielders get some bonus for clean sheets as well.
I got a chance to ask Cristiano Acconci, Co-Founder of WhoScored, a few questions about their rating. Here’s a transcript of our quick conversation. Grammar is unedited; order of statements has been changed for clarity.
Conversation between Danny Page and Cristiano Acconci
CA: I’d be delighting to explain the ratings more if needed.
DP: Are the WS ratings intended to be descriptive or predictive?
CA: Descriptive, they reflect current statistics live only.
DP: Why are Full Time boosts given to players? (Individual WS Ratings are boosted if the player’s team wins.)
CA: To reward team work and team contribution.
DP: How much is that winning or clean sheet boost?
CA: It depends on position. Mids won’t get as much as Defs etc
DP: 55% players are bunched in the 6.X range. Intended? (See the attached image.)
CA: That was the 2012/13 PL average only.
DP: Since it’s descriptive, was it a subjective or objective method to determine stat weights?
CA: Objective
DP: Has the WS method been updated based on stat trends, and are these changes reflected retroactively?
CA: We have, and will plan a new update for next season
DP: Great, thank you so much! Do you mind if I use these answers in an article?
CA: Sure go ahead.
Concluding Thoughts
As an analyst, I’d prefer if a given player’s ratings were not given bonuses for the results of the match. The essence of the ratings appear to be all about the actions taken directly on the pitch and not the actual result. It seems strange to mix the two without explictly showing the amount of bonus points given. I’d also prefer some more transparency in showing how it determined the weighting of certain events.
I appreciate Cristiano Acconci answering some of the questions I posed to him, and I hope to hear more about the details of the rating in the future.
Addendum:
- Mike Goodman and Gab Marcotti further discussed analytics in footy on the follow-up episode: http://fivethirtyeight.com/datalab/podcast-a-premier-league-preview-the-nfl-gets-microchips/
- Screenshots of our discussion, sorry about the ordering:
