17/18 Season Simulation Results

Objective: We set out to see if we could use each player's average PB score last season and the variability in their scores combined with the fixture data to estimate each player's Performance Buzz dividends for last season.

We used this data to simulate each gameday for last season 100 times and recorded the winners of each game day every time. We then aggregated these results to find each player's average dividend returns for last season across the simulations. This is the model's expectations as to how much each player would have paid out in dividends last season for the games in which they were eligible to win Performance Buzz. 

This is useful to help identify the player's who were unlucky not to win as much as their performance buzz scores last season deserved.

Limitations

The main limitations of this model are as follows: The first is that this is only based on 1 season of fixture data so to use these results to estimate future dividends would bias those who had a preferable fixture schedule last season (ie competed on days with fewer than usual players). The second is that this simulation only uses actual PB scores from last season so for some players who only recently IPO’d the performance data is limited. We also didn’t adjust for the varying difficulty of each fixture which leads to a slight bias towards the top PB players when it comes to the aggregate results. Finally, using the results to estimate future dividends would also favourably bias those players who were IPO'd at the start of the season when there was less competition for Performance Buzz wins.

We plan to run a simulation next week that uses several seasons of fixture data and performance data for a more robust prediction for upcoming seasons. We will also use a larger variance for each player to give a more realistic spread of Performance Buzz winners. More details on the next simulation will be released on the website and Twitter shortly.

Results

We also felt you may want to cross-reference each player's results with their actual PB scores. To enable this if you hover over each players scores you are able to see a distribution graph of all their PB scores.