7 responses

  1. Pete
    22/12/2013

    I’m glad you published this as was a bit tardy in getting around to commenting on the previous article… that you needed to divide the score by the number of games reviewed to get a comparable score for each team.

    And you focused on exactly the key conclusion. On average MU got net one important (erroneous) decision in their favour as opposed to Arsenal every single game. Even if the Untold numbers are slightly biased in Arsenal’s favour it is still likely that MU in effect, last season, had a colossal head start.

    As MU’s positive bias seems – anecdotally – to have declined this season that comparison won’t be so bad – but Arsenal still seem (anecdotally) to be getting the thin end of things as much as ever.

    A real pity that Untold no longer have the resources to continue this effort. I really hope someone else can pick up the baton because the only way to prove bias, in the absence of a smoking gun, is by deep statistical analysis where bias can be proved beyond reasonable doubt.

    If the PGMOL’s stats are correct (90%+ decisions correct and – obviously – no bias…) then this kind of outcome is many standard deviations away from what should be expected. To all intents and purposes this is “impossible”. So something is wrong. Incontrovertibly!

    Unfortunately, a large proportion of the population, including many who are otherwise educated, have a very limited education in statistics and probability and just don’t understand this.

  2. OMGArsenal
    22/12/2013

    Great work Tony & Walter. As a retired psychologist who worked continuously with sample populations, standard deviations, means,modes and statistical variations used to measure human behaviour and predict possible outcomes,it is important to remind our UA readers that statistics are predictive NOT certainties. The very nature of statistical sampling across a very large or small population/sample (the EPL is a small population with 22 teams playing 38 league games = 176 referee performances)changes the accuracy and usability of the results. what I mean here is that sampling 17 referees’ performances over a 38 game season does not require anyway near what Walter and Tony have done, to produce fairly accurate and highly predictive results. Most psychological testing requires thousands of members in a sample or population and the standard deviation and variance (how much one result varies from another)must still be taken into account when using the subsequent results.
    Therefore, reviewing every referee’s performance would NOt give us any more accurate a picture than doing what Walter & Tony have done. the principle here is that looking at 30-40-50-60% of the games will give more or less the same results.

  3. Mick
    23/12/2013

    I guess that even if the sceptics are forced to concede the results you present for 2013 are valid they would then argue that it is only one year and thus are not of any value as the sample rate is only one. They would say yes OK but things would even out if you were to look at the last 10 years say.

  4. gouresh
    23/12/2013

    off topic, but just watched Henry’s interview. just fantastic.

  5. Mandy Dodd
    23/12/2013

    Seems whatever index or table applied we lose out. Can only mean one thing

  6. Va Cong
    23/12/2013

    I’m fed up of this can I just kill all the refs is we can get new ones?

  7. finsbury
    23/12/2013

    re: City game.

    Ignoring the bizarre and rare spectacle of the linesmen getting so many calls wrong in the same game against the same opponent (must be a coincidence!) we can state with confidence that the extra days rest gave Abu Dhabi a clear advantage.
    Two vs. three days rest is a crucial difference, this data has been available for a while. It is why the German FA are bulding their own complex in Brazil so their players don’t have to faff about travelling between different complexes in-between games every three days. They want their players to rest and recover in those three days, as much as is possible:

    “research by the former Wales assistant manager Raymond Verheijen. Last year he analysed 27,000 matches – from seven top-flight European leagues, the Champions League and Europa League – and found teams playing after only two days’ recovery against teams who had enjoyed at least a three-day gap were 42% less likely to win. He has called for a three-day gap between matches mandatory in all fixture scheduling”

    http://www.theguardian.com/football/blog/2013/dec/22/premier-league-winter-break-england

Back to top
mobile desktop