Applying Sabermetrics To Football Manager

Ever since FSG took the reins at Liverpool sabermetrics has been a hot topic in football. Even though I suspect it will be a great deal harder to apply to football than it is to baseball I suppose with the right analytical system and data it can be done. I suggest you read the wiki article I linked to if you don't know what sabermetrics is about. In short it is about finding undervalued players based on statistical analysis.

Fortunately FM is a data driven game. As FM players we have a lot more data readily available to us than managers do in real life. So, it's very possible that this approach can be used in FM.

What stat has the most to say about a players impact on a game?

I think it's obvious that the players match rating has a big impact. However, there's bound to be more to it than that.

First of all, you can't just go by the players average rating. Why not? Well, if he has only played one game and received a rating of 10 then you can't be sure this player will be a good investment. So, the average rating is more accurate if it is based on 40 games than if it is based on 5 games.

Does that mean all players with a high average rating and a high number of games played is a good buy. If you were managing Manchester United would you buy a Blue Square North player with an average rating of 8.5? Most likely the answer is no. The league has to be factored in as well. In Football Manager every league is assigned a reputation, beware that this reputation may change over time, the link discloses each leagues starting reputation as of 2009.

I'm not a huge fan of using editor data so what about using the leagues co-efficient? The advantage is that you can see the changes over time, but it doesn't say anything about the different tiers in each league.

Fortunately a bit of research, clicking around in the game, I found the screen I needed.

It even shows a star rating of each league. So that information is also readily available.

At this point the factors we will consider are average rating, appearances and league reputation. Appearances could be a problem though. Someone playing 55 matches a season would get an unfair advantage over someone having played 50 matches. And I think we would all agree that both players have reached a point where their average rating has reached a point where you can count on it. I would say that after 30 matches the average rating has reached its level. It is also a point where I would consider the player to have made a full contribution to the team that season.

This brings up another point. The player probably shouldn't be judged on his current season. You won't get the full picture before the season is over. And what about players in leagues you are not playing in? They won't play as many games.

The formula I'm going to use to evaluate players at this point is Average Rating*Appearances(max. 30)*League reputation.

There are still a ton of open questions and issues, but I think I'll save those for other blog posts. (That means I haven't sorted them out yet)