Sunday, 21 March 2010

So you think you can Jump?

It is with great interest I read the posts in the FML beta forums where people say jumping and speed are the only attributes needed to get a good team nowadays in FML.

Well, the proof is in the pudding. Over the weekend I created some scripts for my MBP to easily grab screenshots from FML and convert them to hard numbers to play around with.


The results were found through regression analysis, with (almost) all attributes set up as independent variables and average rating as the dependent variable. Criteria for the filters were 20+ major appearances and only players playing in a specific position were included to avoid false data.

In this scenario, a positive outcome means that an attribute has a higher chance of increase the average rating, a negative outcome means that the attribute is less important for the average rating.

Here are the Jumping B-coefficients, note that these numbers are by no means decisive, but I think they offer an indication of the ranking of attributes for specific positions:
GKs: 0.084 - 8th most important attribute
DCs: 0.357 - 1st most important attribute
DL/Rs: 0.395 - 1st most important attribute
MCs: 0.067 - 9th most important attribute
ML/Rs: 0.032 - 14th most important attribute
STs: 0.105 - 7th most important attribute

The conclusion? If you want to get high average ratings for your defenders, you definitely need someone with high jumping attributes! :)

More to follow.. Perhaps some stats around pace and acceleration. I'm waiting for players to get enough matches in Miller so I can get some updated numbers from there too..

Oh, and one more thing. I'm not really going to hand over to you the list of the important attributes.. Where would the fun be in that? ;)

3 comments:

  1. I thought of doing something like this before. How were you able to import all that data from images?

    ReplyDelete
  2. I set up some scripts on my mac..
    First normal screenshots, then run pictures through ImageMagik and in the end the open source OCR called tesseract.

    I think (?) this will work on windows too if you install it correctly

    Here are the commands if you want them :)

    convert *.png -crop 1170x433+50+248\! crop%03d.png
    convert crop*.png -resize 200% gifstor%03d.gif
    convert *.gif -gravity center -append alt.gif
    convert alt.gif alt.tif
    tesseract alt.tif ut -l eng
    convert *.png -crop 1170x18+50+228\! header.png
    convert header-0.png -resize 200% header.gif
    convert header.gif header.tif
    tesseract header.tif ferdig -l eng
    cat ut.txt >> ferdig.txt
    rm crop*.png
    rm ut.txt
    rm *gif
    rm *.tif
    rm header*.png

    ReplyDelete