Sports and Numbers

Current Power Ratings

Thru games played Mon Feb 8, 2010

(Scroll to bottom of page to see methodology)

Prediction Comparison

Upcoming Games

Predictions Tracker

(not sorted chronologically - first sorted by home team, then road team)

Methodology

The LS80 methodology employs the concept of knowing when a game is "statistically over."  It is based on a formula devised by Bill James.  If a team with a lead meets the criteria in the formula, that lead is 100% safe.  At Statsheet.com, each game has the "Lead Safe" metric posted in the "Flow" section of the game recap.  Next to that is a number between 0 and 100 to indicate how safe the lead is.

The underlying principle of the LS80 Power Ratings is that the difference in strength between two teams is directly correlated to the time left in the game when the lead is 80% safe.  I also believe it is correlated to the point at which it is 100% safe, but so many games are only over at the :01 mark or :05 mark that it doesn't provide enough heterogeneity to establish a clear ranking system.  The ratings will be a best-fit solution for the results of each game, namely the time left in the game when the winner has reached the 80% mark for "Lead Safe" according to Statsheet.com.  Usually there is no specific time point with 80%, so I will do a linear estimation of that point between two given points (i.e., if 1:40 left is lead safe 60% and 1:10 is 100%, then the halfway mark 1:25 is the 80% point I will enter as data).

In order to test this hypothesis a bit, I will do a simple experiment for the Big 12 regular season.  I will do a comparison between the LS80 Power Ratings and a traditional "Pure Points" style power rating (e.g. Sagarin Predictor), where the only thing that matters is the final ratio of scores in the game (as opposed to the simple margin - adjusts for the tempo this way).  The metric I will use to do the comparison is the prediction of the RATIO of the scores for each team.  I will use only Big 12 regular season data so that both systems have the same amount of information.  That is one of the reasons I'm not comparing it to KenPom.com or some other system that has several months of data already stored up.

Example:  If the LS80 ratings for the two teams after home advantage are 60 and 40, the prediction will be that Team A will score 60% of the points (calculated as Team A's rating divided by the sum of the two ratings).  If the "Pure Ratio" ratings for each team after home advantage are 55 and 50, the prediction will be that Team A will score 52.4% of the points in the game (same calculation method).  ACTUAL RESULT - Team A (78pts) and Team B (55pts).  So, Team A actually scored 54.5% of the points, meaning that the LS80 system was off by 5.5% and "Pure Ratio" was off by 2.1%.

"Pure Ratio" won this hypothetical match-up, and that will be tracked.  But for the purposes of the overall winner at the end of the season, I will look at the average of the sum of the squares of each result.  That's what you'll see in the tracker below the ratings.

Additional Notes:

• Overtime games will count as 0:00 time left on clock for purposes of the LS80 ratings, and for Pure Ratio ratings they will count as 50/50.
• Sometimes, I will not be able to update this page every time games are played.  However, when I "skip ahead" I will always be doing predictions retroactively to the data that would have been available the day a game was played.  You will see these predictions in the predictions tracker listed just as any other prediction would be.