15 responses

  1. Matoo
    19/10/2011

    Great work Walter, really appreciate the effort put in to put up a not insignificant number of reviews so far!

    Hopefully your project can get more refs (hopefully of different alliegences) to review all the games, and then we’d have accurate stats for all the clubs. Because, with the current formula, you get about 10 games a season for most clubs, and this is probably not a large enough test size to see whether they are being favoured or not. However, it might put to bed the argument that refs always favour the big clubs (which I believe is not true)

  2. nicky
    19/10/2011

    Walter,
    A lot of work went into your stats and you should be complimented for that.
    One point strikes me. What appears to be missing is the “degree” of bad calls. There must be some that even top class officials could be excused for missing and it doesn’t seem fair to me that these should be included in the bad calls list.

  3. 037
    19/10/2011

    Solid work!
    This data will be useful for the improvement of British refereeing in the future I’m certain

  4. Eric
    19/10/2011

    Hi Walter! Great work as usual, but… I dont think you should count in the teams outside the top6 which you dont review often, simply because that just cant be accurate, seeing as the smaller teams are bound to get calls against them vs the bigger teams (unless you play arsenal). Now if you would review a smaller team vs an other smaller team you would get very different numbers.

    ps: i am interested in reviewing games if theres stll help needed, how can i contact you Walter?

  5. Shard
    19/10/2011

    Walter

    How many games of Everton have been reviewed?

  6. Mahdain
    19/10/2011

    nice work walter…i saw a table once and guess who was leading having wrong calls go against em? yep Arsenal but tell this to the AAA and they`ll tell you are making excuses and that our team is not good enough,the manager has lost the plot bla bla bla…and all thanks to the media..speaking of which look at this article i found looks like other blogs are also starting to see just how pathetic they are http://www.onlinegooner.com/article.php?section=blog&id=344

  7. Notoverthehill
    19/10/2011

    Some commentators on Walter’s article are more than somewhat off-putting!

    I would have thought another blogger could run through the reviews and keep a running total of the games involved for each team Walter puts into his graph.

    It is obvious that The Arsenal will be the most reviewed. It is a simple matter for anyone to check back on the reviews, is it not?

    Well done Walter and DO NOT allow certain commentators to nick pick because they are too lazy to do a bit of back research for themselves. Next they will say Webb has refereed 10 games, Dowd 9 games and so on.

  8. Dan Tree
    19/10/2011

    Brilliant analysis of the data collected. Another great method of showing the bias for/against each team. Maybe one day there will be enough ref reviewers to cover every match and we will have complete stats on this. Then there would be a great case for sending a few thousand strongly worded letters to Mr Riley.

  9. Eric
    19/10/2011

    @Notoverthehill

    Not at all, I think we are all greatfull for all the great work Walter is putting in (and the other reviewers ofc) and hope he continues for a long time doing so. But 1 has to be critical, otherwise you send the wrong message. Which happened in this article imo, its only a fair picture if all the data is there, but it aint since (only) the top6 gets reviewed. So then only the top6 can be compared to each other (for the fair picture).
    Im not at all trying to be negative, I support UA since a long time, love reading it everyday and usually never comment unless i see something dodgy.

    Hope this clarifies a couple of things!

  10. Gord
    19/10/2011

    I think there might be a better way to plot this data. I can’t say I can see what generated the order of the teams, but in terms of a visual presentation, the order of the teams is arbitrary (any permutation could be used). Which suggests “connect the dots” probably shouldn’t be used. It does lead one’s eyes to adjacent data points.

    The data looks more to me like some kind of histogram data. I wonder if plotting it as a histogram, with yerrorbars might not be a good thing to do?

    It is binomial data. If some team had 20 calls made against them, of which 15 were bad, the probability of a wrong call is 0.75 (and the probability of correct call is 0.25). The mean number of wrong calls (Np) is 15. The variance associated with that is Np(1-p), which is 3.75. The square root of 3.75 is a little less than 2, so we’ll round the estimated standard deviation to 2 (although there are reasons to not round the standard deviation). We could then redescribe the number of wrong calls as 15 +/- 2 (or the fraction wrong as 0.75 +/- 0.10) (nominally 67% confidence, but not really if we round the s.d.).

    In the case of Wolves, where all the calls were wrong, using Np(1-p) to estimate the variance can’t be correct (the formula gives 0 for the variance). I suspect the problem comes from specifying the total number of calls. If the referee had to do the game all over again, would the referee make the same number of calls? I don’t know of an easy way around this.

    I don’t know what package you are using for making plots. If you happen to be using Gnuplot, the following page at Los Alamos National Labs talks about making plots with yerrorbars.
    http://t16web.lanl.gov/Kawano/gnuplot/plot5-e.html

  11. Cyberian
    19/10/2011

    I appreciate your graphical presentation in this post and the work that goes into it. Please consider some suggestions for making your visual presentation even more effective based on the work of individuals such as Edward Tufte (www.tufte.com) and Howard Wainer (http://www.statlit.org/Wainer.htm ). Tufte, Wainer, and others discuss what makes graphical displays more or less effective.
    First, your graph connects the points and implies a series or sequence when there is none. It would be better to use a bar graph intended for discrete events. Second, the bar graph allows the use of error bars or markers to indicate the range of scores (lowest to highest). In short, some measure of variability is helpful to the viewer. Third, you have the option of a horizontal bar (like the one you used) or a vertical bar (one less familiar) in which the labels go on Y axis and percent goes on the X axis (see below). The vertical bar is often easier to read for many people. Fourth, you do not need to include the number for the percentage value because the purpose of the graph is a relative comparison easily determined by looking at the relevant axis (e.g., Y axis in your graph). Instead you could insert a useful number such as the sample size (e.g., 5/6 might reveal the number of reviews of games of the number of EPL games played). Finally, graphical display experts recommend organizing data in tables or figures along some dimension. For example, you could rank order scores from the least to most favorable result. Such a ranking system also makes comparisons across time easier because you can average the rankings. Below is a crude horizontal bar graph with partial data for illustration. Please consider revising your figures to make your hard work even more effective. Note that the name (Wolves) data (xxx) and information (reviews/games) should appear on a single line; however they have been wrapped around in this narrower space: name xxxxxxxxxxxxxxxxxxxx information

    Wolves xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx reviews/games
    WBA xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx reviews/games
    Arsenal xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx reviews/games

    Blackburn xxxxxxxxxxxxxxx reviews/games
    Newcastle xxxxxxxxxxx reviews/games
    0 10 20 30 40 50 60 70 80 90 100

  12. WalterBroeckx
    19/10/2011

    Thanks for the tips. I will see if I can change it for the future. Now I must confess I’m not the smartest one with knowing how to present it in graphics. But I will take it on board and try to see how I can make things more complete.

    Like adding the number of games which is important of course.

    I will see if I can change the graphic part for other articles.

  13. WalterBroeckx
    19/10/2011

    Eric a mail is on its way

  14. Jas777
    19/10/2011

    For this to be accepted by others you need to clarify who is doing the reviewing and what connection they have with the teams being reviewed.

    Also the manner of reviewing. Is it off TV or is it at the ground? Is the reviewer off TV seeing mutliple replays and are they going back to see what happened or just viewing replays as appears on coverage. Are all the reviewers following the same rules

    Other things you have to clarify is
    Are the reviewers supporters of one of the teams playing? This is very important as most people have preconceived ideas about things and when there are 50/50 decisions to be made.

    Also are they using the rules to the letter of the law or do they have a regional basis? ie A english reviewer will probably have a different minset to someone from the continent.

  15. Walter
    20/10/2011

    Jas777,
    maybe I will answer to your questions in an article. Because the questions are important and the answers also.

Back to top
mobile desktop