Sunday, February 24, 2013

Which stats REALLY matter for a position?

I have been wondering for a while which stats are the most important to target and what type of players are most likely to produce high quantities of those stats. My main purpose for this is to find individual players who underperformed for us in fantasy last season that might be able to give us better returns this season if they have similar performances. In other words, what players got screwed over last season because their teammates didn't live up to average?

What I've done is look at the correlation between PP90 and every statistic that provides fantasy points and a couple that don't provide points such as shot conversion percentage and key pass conversion percentage. I've done this on a positional basis only, so we know which stats are most important for each position.

Defenders:

Defenders are the most inconsistent group of players in terms of finding a stat that highly correlates with PP90. CBI had the highest correlation at 0.34. Goals and assists follow with correlations of 0.29 and 0.28 respectively. The big shocker, is that a team's clean sheets has a surprisingly low level of correlation - a mere 0.19 - with a defender's PP90.

So, what does this mean? It's tough to draw any firm conclusions about what defenders we should choose. Since CBI and goals are the two stats with the highest correlation with PP90, I am inclined to say that we want to pick center backs who get forward on set-pieces. This is fairly accurately reflected in the players found at the top of my defender projections. We see center backs like Collin, Bernardez, Ianni, Taylor, Berry, Olave, and McDonald all feature as good targets according to my PP90 projections.

Midfielders:

There are a couple stats with much higher correlation with PP90 for our midfielders. Assists correlate with midfielder PP90 at a strong 0.77. However, there are several other stats that correlate at better than 0.5: key passes (0.72), goals (0.68), and crosses (0.62).

There's a couple things to take away from this. First, is the reason there is such high correlation for several statistics is that there are some midfielders (pretty much every defensive midfielder in the league) who simply don't rack up any of the above categories. Even with the added defensive points for CBI and recoveries available, defensive midfielders simply don't score enough of them to make them viable alternatives to their attacking counterparts.

We know we want attacking midfielders, but do we want wingers or central midfielders? Both versions of an attacking midfielder has about equal opportunity for goals, assists, and key passes. It isn't until we get to the fourth most highly correlated statistic where we find our answer. A decently high correlation with crosses and PP90 means we want wingers. Zusi, Rosales, Davis, Pontius, Chaves - all are in the top level of my midfielder projections and all are wingers. It should be noted that there are several central playmakers who also rank high in my projections (Morales, Ferreira, DeRosario) who take freekicks, corners, and/or PKs and are also valid targets.

Forwards:

Surprise, surprise, goals is the stat that best correlates with PP90 for forwards at 0.77. Assists (0.69) and key passes (0.64) also had high levels of correlation with PP90.

In our forwards, we are looking for goalscorers. I'm sure everyone who has even heard the term "fantasy soccer" could have told you that. However, I was surprised at how highly correlated both assists and key passes were with PP90 which gives me second thought about choosing guys who are poachers (Bengston, Bruin, Cooper and to a lesser extent, Wondolowski) instead of more well-rounded players (Higuain, Henry, Keane).

5 comments:

  1. PP90 is projected points assuming the player plays 90 minutes. I defined it in prior posts. Guess it's time to start making a glossary page.

    ReplyDelete
  2. Hey, I've been checking out your blog, and it's been very helpful! Great work. I had a thought you may want to consider: When considering a player's cost, I like to look at it as how much above the minimum are you paying for that player. Because you are required to have 15 players, you have a certain minimum that you could possibly pay. The cheapest defenders are at $4, so to get Collin at $6, I'm really paying $2 above what's required, whereas the cheapest forwards are $4.5, so to get Wondo I'm paying $6. I think it would be insightful to consider their production vs this cost.

    ReplyDelete
  3. Interesting thought, Jeff. I'm not really sure that comparing to the lowest possible price makes sense since those players almost surely won't see playing much playing time at all. I think it makes more sense to compare them against the cheapest starter and see how much greater their production is.

    Unfortunately, I don't have time for another 10 days to post anything on the subject as my coaching season begins next week and I am playing in a tournament this weekend.

    ReplyDelete