Quick Q&A on dingoR…
Q1. So what happened to old dingoR with scores from 75-150 or so?Why are you publishing points now?
A: I thought points reveal more than a rating that did not have much significance other than comparing players. Total points or points per 90 are more intuitive. Let me explain. Baseline is ~50 points for an average team in a 38 game season. So an average player who plays all the minutes contributes 50/11=~4.5 points per season. This number can be scaled by minutes played. So for every 90 minutes a player should contribute 4.5/38=~0.12 points.
Q2. So how do you become above/below average?
A: Depends on what you are doing on the field and how your team is doing. We have a lot of data on offensive output(goals, assists, shots, etc.) dingoR also emphasizes possession. Not % possession but whether you are obtaining possession or giving the ball away. Many of these offensive and defensive actions we can count contribute to an additive score(no regressions) and after adjusting for your team strength we get a raw score for you that can be added or subtracted from points associated with your minutes.
If we look at Suarez this year he won the MVP award with 16.9 points. An average player would have contributed 3.9 points in the minutes he played. So his contribution was 13 points more than an average forward. Players in different positions have similar average point contributions but forwards and goalies have the most variance. Backs the least.
Q3. If you are not using regressions then how do you know how each action is worth?
A: A bit of logic and some math. I am sure I have wrong assumptions and mistakes but overall I am very happy with dingoR as far as capturing what I think matters on the field. Whether what I consider does really matter is a completely different question.
Q4. Isn’t per 90 a better metric to start with?
A: Depends. In case of injuries yes. Less so in suspensions as the player got suspended for something negative. Also some kind of players have inflated per 90 numbers. For example, super subs(attacking players in dominant teams) have very high per 90 contributions. Morata would have been our MVP this season by that metric. Now, he might be as good as that number even if he played every minute of every game but I doubt it.
Q5 What bothers you about dingoR?
A: Quite a bit actually but these are due to data constraints:
1- Goals are the main output. So players who have had a lucky streak can be overrated and vice versa. But this is mitigated by the fact that I have 5 years of data for major leagues so it is easy for me to see if a player deviated a lot from his average or trend. If one day I have the date I could also create expG version.
2-On the defensive end there is not as much data just because it is tougher to count prevented actions. Therefore defenders on the same team are more correlated with each other than other players. So perhaps we are not assigning the benefits/errors to right players. Again I would need more data to overcome this.
3- On the team level, clubs that have disproportionate points associated with their goal differences cause problems. For example, Spurs this year had a pretty bad goal difference for the number of points they got. This can be due to few reasons: too many/few draws, few lopsided games or luck. In Spurs’ case they had a terrible run against big clubs losing by large margins and then winning most of their other games by 1 goal. So combination of few games and good luck kept their points high. If we were to predict for next year we have to consider the context which makes it a more subjective exercise.
Q6 What is dingoR being used for?
A: Nothing. No betting, no commercial use or any other use is envisioned. One final note is that dingoR formula has not changed since beginning even though I can see how it can be improved slightly. But until there is big breakthrough or significant data I would rather keep it consistent rather than tinker.