Premier League 2015-2016 Prediction league table based on the expected goals data model
Expected Goals ForPrediction metric
Expected Goals AgainstPrediction metric
Expected Goal DifferencePrediction metric
Expected Points™Predicted final points tally of the season
PDOShot accuracy metric
Expected PDOShot accuracy metric based on expected goals
Total Shots RatioOverall dominance metric, higher is better
Prediction Rank™Expected performance
|15||West Bromwich Albion||38||31.74||48.29||-16.55||27||903||909||0.37||926|
|13||West Bromwich Albion||17||5||5||7||17||23||-6||138||239||20|
Expected Goals (xG)
This is the number of expected goals a team will score based on shot data from previous games. Shot zones, types, attempts, accuracy and other stats are all taken into account. By calculating expected goals for (xGF) and against (xGA), we reach an expected goal difference (xGD). This is then compared to the team's current goal difference to calculate performance.
Expected Points™ (xPts)
Our own formula based on a number of factors. The team's current points-per-game is multiplied by 38 to serve as the baseline. We then boost that number using our own predRk metric (see below) and then adjust further using team performance according to xGD.
This formula originated in ice hockey and was translated into football terms by James Grayson. It gives an indication of how well a team is performing in terms of goal conversion. Higher is better.( (goals for / shots for) + ( 1 – (goals against / shots against) ) ) * 1000
Expected PDO (xPDO)
The same formula as PDO, but with expected goals (xG) instead of current goals. The end result is a baseline number that should give an indication of form: if the xPDO is lower than the PDO, it means the team is overperforming and scoring more goals than it should.
Total Shots Ratio (TSR)
Before the emergence of expected goals (xG), the preferred prediction metric was total shots ratio (TSR). It's essentially a dominance formula, calculated by dividing the number of shots taken by the total shots overall. The closer a team is to 1.0, the more dominant the team has been in the season so far.(shots for / (shots for + shots against)
Prediction Rank™ (predRk)
This is our own formula to score teams based on a number of different factors, including all of the above mentioned metrics. This score is then used as the basis for our league prediction.
The base numbers on this page all originate from Opta via WhoScored and Squawka. The xG and predictive models are partly opinionated, which is why our xG data is an average of several different sources:
Paul Riley (@footballfactman)
Paul's data is a great source based on his own shots model explained here. This data is regularly updated and readily available for everyone.
Michael Caley (@MC_of_A)
Despite being a Spurs fan (boo), Michael's xG model has great detail. His model is explained here.
Danny Page (@DannyPage)
Danny is a developer that has created a few interesting match prediction tools, both for the outcome of a specific game and also a long-term xG simulator. To use his tools you need data, which can be lifted from the aforementioned sources.