Simulating match day results and expected dividends for the rest of the season

Guest blog post by the popular IndexMachine

Where’s the value left in August?

There’s plenty of value to be found on the index for the remainder of August, with gold days galore. While IPD’s offer great returns with the summer bonus, i’ve focussed on PB here and simulated the remainder of the season to look at expected payouts across all players.

These simulations are built using the suite of Index Machine models – simulation-based machine learning algorithms that model individual player scores for every gameday on a range of contextual variables such as where the games were played, locations, opposition and form. These models are updated as new games come in and then projected onto future games. The following projections are for the remaining games in August – the EL, CL and start of Ligue 1.

My game simulations have an Inter-United EL Final and City-Athleti CL Final, so other teams value obviously increase if the knockouts go very differently.

By simulating the games probabilistically, we can back out measures of expected value. Since each gameday is modeled independently, the expected PB dividends are the weighted average of simulated gameday win probability* dividends on offer across the fixture window. Therefore value increases with number of games played, the simulated probability of winning any of those gamedays and the divs on offer.

What are the projections for the rest of August?

Top GK predictions:

  • Oblak
  • Handanovic
  • Neuer

Top Defender predictions:

  • Kimpembe
  • Marquinhos
  • Alaba

Top Midfielder predictions:

  • De Bruyne
  • Veratti
  • Fernandes

Top Forward predictions:

  • Mbappe
  • Messi
  • Lewandowski

We can also back out value by looking at the relative yield of those expected divs. Here is a visual way of doing that. We plot expected divs (Y-axis) versus price (X-axis). Then we fit a (loess) regression of expected divs on price. Anyone above the line has higher relative value.

One massive value anomaly here – who represents orders of magnitude more PB value than any other player on the index for the rest of August. No surprises there:

Share This Post
Index Machine
Written by Index Machine
Machine learning and advanced analytics. Simulating player performance on the index.
Have your say!
1 0
NEW: Introducing Spread Alerts & Notifications on Slack. Find out more here.
Holler Box