Saturday, March 10, 2012

New Feature - Today's Games

I've added a new feature to the site called Today's Games (accessible via the top menu bar).  This is currently an NBA-only feature.


The purpose of this feature is to show how the point spreads and over/unders for the day's games deviate from standard market expectation, where standard market expectation is defined by my betting market rankings (here are the rankings and here is an overview of the methodology).

If the line deviates significantly from the prediction, that's a good indicator that something abnormal has happened that has led the market to re-evaluate the relative strength of the two teams.

In many ways, this is an automated version of the analysis I did on Carmelo Anthony in January.  


Here is a sample from today's version of this new feature:

The lines shown are for the home team.  So, a -3 point line means that the home team is a 3 point favorite.  The bar next to the miss points towards the team that has a bigger advantage, compared to the prediction.  For example, the biggest miss is on the Warriors-Mavericks game.  The rankings expected the Mavericks to be favored by a point on the road against the Warriors, but instead the Warriors are favored by three.  What could be driving this?
  • Injuries.  Both Brendan Haywood and Brandan Wright are day-to-day right now.
  • Drama.  There's Coach Rick Carlisle's apparent issues with Jason Terry.  And the ongoing Lamar Odom soap opera.  According to Jason Terry, "We’re not a team out there right now".
  • Rest.  Today will be the Mav's third game in as many days.  According to the analysis I shared in this post, I couldn't find any evidence that point spreads are adjusted for 3-in-3 situations, at least not above and beyond the adjustment for back-to back (worth 0.75 points).  But I think I may have to revisit that analysis.
A more straightforward example is the Nets-Rockets matchup.  The Rockets are favored by 2.5 points more than expected.  The most likely explanation is that the Nets' Deron Williams will miss tonight's game due to a calf injury.  So, the market appears to value Williams contributions at about 2.5 points.

Supplementing Stat-Based Models

A potential use for this new feature could be as a supplement for those of you that try to use stat-based models to beat the NBA point spreads or over/unders.  Because stat-based models do not handle injuries and other recent developments very well, this feature could be used to let you know when you should adjust your models, or when to simply lay off betting on a particular game.


  1. Please pardon me if I overlooked something in this article, but do you adjust expected point spreads according to rest days at all?

    An idea for a new type of players' +/-...*drumroll* Vegas +/-.

    Basically do what you did for C.Anthony for every player (except for those who haven't missed a game but if you stretch the window to a year back or so, you get fewer players who missed zero games). I'm not sure if you can automate this so that you easily can do this for every player, but it'd be fantastic because from what I've seen, Vegas point spreads continue to outperform adjusted (and regularized/ridge regression) +/- systems. Perhaps to go further, multiply a player' Vegas +/- by how many minutes he typically plays to arrive at a per-48 min figure. (e.g. +2.5 and 28 MP/G = 4.3 per 48 min)

  2. Hi Ben. I do adjust for rest days, but only if the team is playing back to back. Any team that is playing on zero days rest is docked 0.75 points in the predicted point spread (no differentiation between home and away). Here is a link to the analysis: link

    That's a great idea on the Vegas plus/minus. I'll have to give some thought as to how to implement it.