Saturday, May 5, 2012

The Ticker - Playoff Edition (updated daily)

The two daily features of this site (Today's Games and The Ticker) were dark for the past week, ever since the NBA regular season ended.

I have now relaunched The Ticker with a focus on the playoffs.  Rather than focus on team rankings and Generic Points Favored (GPF), the focus is now on win probability: the probability that a given team will make it through the first round, the probability that a team will win the NBA championship, etc.

Here is an abbreviated description of the new features (I hope to go into more detail later, time permitting):


The GPF (Generic Points Favored) rankings are still updated on a daily basis, but the update is now more of an Elo-style update along the lines of what I used for College Basketball.  I then use these rankings to simulate the remainder of the playoffs 10,000 times.  I can calculate the win probability of any matchup by first calculating the predicted point spread and then converting that into a probability according to the following formula:

Win Probability = 1/(1+exp(-(points favored)/6.5))

This simulation is run every morning based on the latest game outcomes and point spreads and the results are summarized in the following three formats:

  • PWP (Playoff Win Probability) - This is similar to the old Ticker graphs, in that it tracks a team's daily chances to win any given round of the playoffs
  • BirdsEye - I'm fond of this one.  This summarizes the NBA Championship win probability for all teams on a single chart, and shows how each team's share of that probability changes day to day.  I tried to color code the graph with team colors, but you can also hover over the graph for the team detail (or hover over the legend to see the team highlighted).  Note how the Bulls' share of the pie shrinks dramatically after the Derrick Rose injury.
  • Ranking Table - This summarizes, on a daily basis, each team's GPF and probability of winning each round.
The methodology is still a work in progress.  I'm thinking that the Elo-style update to the GPF rankings may be too sensitive, in that the probabilities shift more due to changes in the rankings, rather than to the game outcomes.

No comments:

Post a Comment