We have received some emails from some of our subscribers asking if we can explain a little about the Four Factors and how we use them to construct ratings. While our algorithm is our secret recipe, and we obviously cannot reveal the numbers, we can show you a little about how it works, which is the basis of our Red Rating.
Pirate Red: An algorithm using the Four Factors adjusted to strength of schedule and home/neutral/road splits.
The Four Factors include:
Effective Field Goal Percentage by offense and defense: (FG+(3pt*0.5))/FGA
Example: Oklahoma Offense– (674+(252*0.5))/1426 = 56.1%
Oklahoma Defense– (605+(169*0.5))/1514 = 45.5%
Oklahoma Effective Field Goal % Difference 56.1-45.5 = 10.6%
Oklahoma EFG% Adjusted for SOS = 10.6% *1.076 = 11.4% better than average or about 13.2 raw points better than average based on our algorithm. We adjust this rating for home and road games, as the Sooners enjoy a home court advantage while suffering a little road court disadvantage.
Offensive Rebounding Rate & Opponents’ Offensive Rebounding Rate: Off. Reb./(Off. Reb. + Opponents Def. Reb)
Example: Michigan State Offensive Rebounding Rate: 313/(313+528) = 37.2%
Michigan State Opponents Offensive Rebounding Rate: 234/(234+752) =23.7%
Michigan State Rebounding Rate Difference 37.2-23.7 = 13.5%
Michigan State Rebounding Rate % Adjusted for SOS = 13.5%*1.064 = 14.4% better than average or about 10.6 raw points better than average based on our algorithm. We adjust this rating for home and road games, just like above with Oklahoma.
Turnover Rate for each team and their opponents: TO/Possessions
Example: Stephen F. Austin Turnover Rate: 279/1553 = 18.0%
Stephen F. Austin’s Opponents’ Turnover Rate: 413/1551 = 26.6%
Possessions differ by one or two per game depending on who wins the opening tip and who has the last possession of a half.
Stephen F. Austin Turnover Rate Difference 26.6-18.0 = 8.6%
Stephen F. Austin Turnover Rate Adjusted for SOS = 8.6*.924 = 7.9% better than average or about 5.2 raw points better than average based on our algorithm. As you can see in this example, SFA’s schedule is considerably weaker than average, so some of their advantage has been removed.
We then adjust for home/road splits like above.
Free Throw Rate: FT/Possessions
This is where we differ from most of the other Four Factors followers. The regular formula used by others is FTA/FGA. We believe this formula overstates the actual rate due to many college games where teams are forced to foul at the end. Thus a team might hold onto a 7 to 10 point lead with 5 minutes left in the game and then work the clock down to the last few seconds. In that same game, the opponent might be forced to foul in the last 2 minutes. Thus, field goal attempts would be down and free throw attempts would be up. What if the leading team blows a lot of free throws, allowing the trailing team to come from behind to win?
The stats for the team that blew the lead would be inflated. What if instead of taking 55 field goal attempts with 22 free throw attempts had they not slowed the pace down and not gone to the foul line almost every possession in the end of the game, they ended up taking just 49 field goal attempts and 32 free throw attempts, making just 4 of 10 of those foul shots under pressure when they blew the game?
Using the regular FT Rate, the team that blew the game had a 65.3 FT Rate but had the game been normal without the slowing down and multiple fouls, the team’s FT Rate would have been 40.0. That’s a difference of more than 25%, which is huge, and it led you to believe that the eventual losing team was better off to slow the game down and force fouls.
We use Free throws made per possession. Let’s look at that same game again. First, instead of each team having 70 possessions in the game, the slowdown lowered that number to each having 68 possessions, not much, but still worth mentioning.
The team that ended up taking 32 free throw attempt made 20 in the 68 possessions. Using our method, their FT Rate was 29.4%. Going under the theory that they would have taken 10 fewer free throw attempt (the 4 of 10 they shot when the other team fouled at the end), they would have been 16 of 22 in a 70-possession game, which figures to 22.9%. Instead of 25% difference, we now have just 6.5% difference, and that is about what the extra fouls by the opponent benefited this team. Even though they hit just 4 of 10 of the extra foul shots, there were 5 extra fouls committed, and somebody on the other team probably fouled out, hurting the other team.
Once again, the PiRate Rating Formula for FT Rate is: FT Made/Possessions
Example: South Carolina: 480/1793 = 26.8% as opposed to 50.4% the other way
South Carolina Opponents: 365/1793 = 20.4% as opposed to 39.1% the other way.
USC FT Rate Difference = 6.4% our way as opposed to 11.3% the other way, which overstates the Gamecocks’ advantage due to their numerous games in which they were fouled at the end of close games. In their most recent game, USC went to the foul line five different times in the last 57 seconds of the game, due to LSU fouling.
South Carolina FT Rate adjusted for SOS = 6.4*1.038 = 6.6% better than average or about 2.8 points better than average before adjusting for home and road splits. Had we used the standard FT Rates, USC’s adjusted points would have been inflated by 2.2 additional points, which is 2.2 points more than they deserve to have.
Putting It All Together
We plug each of the Four Factors’ results for the two teams playing and then we adjust by giving the home team their advantage based on their data while deducting from the road team based on their data. In some cases, the home team might actually have an overall disadvantage in some cases, but it is rare.
We then come out with a base number for each team. The difference is not the pointspread just yet. We have to include a constant that we correlate based on back-testing to come up with the most accurate predictive model.
How has it performed so far? This is just the second season we have done this, and because it is quite labor intensive, as we have to plug in new stats for each team for each game, we can only do this for the power conferences–ACC, Big 12, Big East, Big Ten, Pac-12, and SEC, plus the occasional big game, like SMU vs. Gonzaga this weekend.
Last year, in Todd Beck’s Prediction Tracker, our Red and White Ratings finished 1-2 overall. To date this season, the White-Red Ratings are currently 1-2 overall.
Here are this week’s selections.
|Games Schedule for:||Saturday, February 13, 2016|
|North Carolina St.||Wake Forest||8||8||4|
|Oklahoma St.||Kansas St.||2||1||3|
|Games Schedule for:||Sunday, February 14, 2016|
|Florida St.||Miami (Fl.)||-1||-1||-1|