The inaugural outing of the predicted scores wasn’t an overwhelming success, but as I mentioned in the original post I knew this would probably be the case given the sample size. However, the formula did produce 7 correct results and, against the mold, rightly predicted Burnley to win at Everton. It’s the goal ratio accuracy that needs to improve.
In light of that, I’ve now added a further measurement to the calculation. In the above table it is only goals scored/conceded by each team that is considered, and not the quality of opposition. In order to rectify that, I looked at the points collated by each team so far this season and the standard deviation each had from the mean.
How does this work?
By taking the total number of points from all teams (193) we have a mean average of 9.65 per team. Man Utd have 19 points, so their standard deviation is 9.35 (19 – 9.65), whereas Huddersfield have 9 points, so theirs is -0.65 (9 – 9.65).
These numbers are still too high to use in the formula, though, so I’ve divided the results by 20 (number of teams) to give us a usable figure. Using the same teams as above, this gives Man Utd and Huddersfield standard deviations of 0.47 (19 – 9.65 / 20) and -0.03 (9 – 9.65 / 20), respectively.
The sample size is still small, but adding this gives the scores a better feel to me, and better represents the form teams in the league. I thought about using standard deviation of goals scored, but this would not be accurate – see Liverpool as an example of a team that currently aren’t producing results that match their level of goals scored.
Fantasy wise, there are few surprises in predicted clean sheets for Burnley, Chelsea, Man Utd, and Tottenham. Huddersfield are fancied away at Swansea too, although personally I feel the Swans have been a little unlucky recently.
Goals wise, Chelsea and Man City are, again unsurprisingly, backed, while the games at Southampton and Leicester are expected to be close affairs.