Let me begin by saying, unlike Kenneth Massey, Jeff Sagarin and several other computer ranking guys, I am not a mathematician. I like math. I am pretty good at it but I am not in the same league as those guys. Compared to them, I’m Fred Flintstone. They are geniuses. They utilize extravagant statistical formulas to come up with their rankings. I use an excel file and have developed fairly straight forward formulas and macros using simple mathematics to come up with my rankings. Additionally, my process of ranking teams can become quite convoluted and someday I may have to hire a computer programmer to assist me. Yet, in the mean time I’d bet if you checked with those ranking gurus, they’d tell you the “Lazindex” is a pretty decent ranking system.
I began ranking college football teams over twenty years ago using paper and pencil. The concept was simple. If Team A was rated 80 and Team B was rated 70, Team A should win by 10. Adjustments were then made based on the actual outcome. If Team A won by 10, no adjustment was made. If they won by 20 a minor adjustment was made to both teams. If they won by 30 or if they lost, a more drastic adjustment was made.
This process worked “o.k.” for time but I found it didn’t adjust quickly enough and its accuracy was often questionable. I needed something that would reflect what was really happening. It was taking too long to find out if Team A was a decent team because it took so long to judge just how good every other team was. For example if Team A beat team B by 10 in week one but Team B didn’t win a game all season, the system couldn’t tell soon enough that Team A wasn’t as good as initially indicated. This is because it could not see the interaction of all the other teams and the impact those teams had on Team A in a timely fashion.
When it comes to ranking teams, most sportswriters and pollsters or what I call the “human” element have but two criteria at their disposal:
1-What were the “head to head” results?
2-Who were the opponents?
Several of these voters will also use “who beat whom” scenarios. Unfortunately, from the human perspective, these “who beat whom” scenarios seem to be unable to work when it comes ranking teams. The reason these scenarios don’t work well for humans is that we can’t possibly see and/or remember all the interactions is because there are thousands upon thousands of them. A computer however, sees everything.
Every single game played has an impact on every team, even if they are idle that week. This is called “connectivity.”
Most people don’t even realize these connections exist.
For example:
Did you know, in 2005, #1
Bacone vs. Haskell
Haskell vs. Trinity Bible
So did 1AA #1 Appalachian State. Here is one of theirs:
Appalachian St. vs.
Drake vs. Waldorf
Waldorf vs. Luther MN
Luther MN vs. Trinity Bible
Here is one for D2 #1
S Nazarene vs. Haskell
Haskell vs. Trinity Bible
Believe it or not, every game
It should be noted, in college football there are 10 schools in the New England Small Conference that are “disconnected” from everyone else. This is because they only play against themselves. It is impossible to know how good or how bad these teams are compared to everyone else so this conference must be tracked separately. I have found these “disconnections” to be quite rare and an exception to the norm.
In any event, it’s very important to remember the “connections” concept. Hopefully, it will begin to make more sense as we go through the ranking process.
In order to begin the process, the first step is to determine a power rating for each team. The power rating is a value that determines a team’s potential. It is established based on actual game results.
These values are created and re-created every week and averaged together to create the average power rating for each team. The process usually takes place after all the week’s games are played but can be run at any time if for some strange reason, it was desired.
To create a team’s power rating for the week we must first establish a game value. To do this, simply add the beginning power ratings together and divide by 2. If we use Team A vs. Team B, we come up with a game value of 75.
Team A opening Rating = 80
Team B opening Rating = 70
150
150/2
Game Value =
75
How much emphasis should be placed on the final outcome? I
feel a game such as
For example:
Team A opening Rating = 80
Team B opening Rating = 70
Delta =10
100-10 = 90
Game Significance 90/100 = .90
In the case of Team A vs. Team B, the game significance is .90. This means the final point difference will be multiplied by 90%.
Once these levels are established, we need to create a min/max movement level or “cap” for both teams. This is done by multiplying a value of 10 by the game significance. The number 10 is the maximum amount of possible movement allowed for each game by the Lazindex. Note: in order to achieve maximum range both teams would have to have identical opening power ratings.
Maximum Possible = 10
Game significance
=x.90
Maximum range for
Team A vs. Team B = 9
(plus or minus)
Now that we know the game value (75), game significance (.90) and the maximum range (9) or (-9), all we need is a final point spread margin to determine the power ratings of each team for that particular game.
Let’s assume Team A wins by 13 points. Simply multiply 13 times .9
This value is divided by 2 then added to 75 for the winner and subtracted from 75 for the loser.
Final Margin = 13
Game Significance
x .90
Adjusted Margin = 11.7
Winning Team -
Adjusted Margin Assignment 11.7 / 2 = 5.85
Game Value = 75
+ 5.85
Team A rate = 80.85
Losing Team -
Adjusted Margin Assignment -11.7 / 2 = -5.85
Game Value = 75
- 5.85
Team B rate = 69.15
In summary, by virtue of its 13 point victory Team A played at a power rating level of 80.85 and Team B played at a level of 69.15. These values are “averaged” in with all previous weekly values for each team to come up with a new updated power rating. Seems easy right?
Here’s where it gets a bit more involved. As we discussed earlier, it is important that every “CONNECTION” is recalculated and constantly adjusted to ensure each school’s values are accurate.
Using the updated “average power rating,” the lazindex will
actually “replay” every single game for every single team. Each replay will
result in a new average power rating. These replays are often referred to as “iterations”
or “loop backs” and are performed by the computer until the average power
rating value no longer changes. This process is done using a simple macro and can
sometimes take over 3000 iterations before the numbers stop moving or are so
infinite that the computer stops changing them. For instance,
The next step which is run simultaneously with the power rating process is to determine a Winning Edge Factor. This is done in the exact same fashion as the power rating with the exception that the result always has the victorious Team winning by (+1) and the defeated team losing by (-1).
With the Winning Edge Factor, the scale is different because this is calculation is predicated on
wins and losses and opponent wins and losses. Not margin of victory:
For Example:
Team A opening WEF = .75
Team B opening WEF = -.25
.50
.50/2
Game WEF Value =
.25
Team A opening WEF = .75
Team B opening WEF = -.25
Delta =1.00
100-1 = 99
99/100 = .99
Game Significance = .99
Maximum Possible = 10
Game significance
=x.99
Maximum range for
Team A vs. Team B =
.99 (plus or minus)
Final Margin = 1
Game Significance
x .99
Adjusted Margin = .99
Winning Team -
Adjusted Margin Assignment .99 / 2 = .495
Game WEF = .25
+ .495
Team A WEF = .745
Losing Team -
Adjusted Margin Assignment -.99/ 2 = -.495
Game WEF = .25
-.495
Team B WEF = -.245
The next step is to “replay” the entire season up to that point just as we did with the power ratings.
Over and Over again until all the team values cease to adjust.
After the all the iterations have taken place and all the Power Rating and WEF values stop moving we have one more calculation which takes place before the rankings are complete. In order to come up with an overall Ranking Value for each team:
A) Multiply each team’s WEF by the total number of games they have played.
B) From that number, subtract the number of losses the team has.
C) Add the value of B to the Power Rating and divide by 100.
For Example:
Team A Record = 6 wins, 5 losses
Team A WEF = .75
Team A Power Rating = 80
A) 11 * .75 = 8.25
B) 8.25 – 5 = 3.25
C) 83.25/100= .8325
As a result, Team A has a Rank Value of .8325. This will determine where they are ranked.
The Power Rating value of 80 will be displayed so that predictions can be made.
In summary, there are several key elements that all tie together to complete the ranking process.
1) Connectivity
2) Replays (also known as Iterations or “loop backs”)
3) Establishing a Game Significance Factor
4)
5) Final Game Margin
6) Winning Edge Factor
To conclude, here are some
additional “connections” for
6A
Monsignor Pace vs. Belen Jesuit
Belen Jesuit vs. American Heritage (Delray)
American Heritage (Delray) vs. St. Edward's
St. Edward's vs. Warner Christian
Warner Christian vs.
5A State Champ
Osceola (
Lake Highland vs. Cardinal Mooney
Cardinal Mooney vs. The Villages
The Villages vs. Trinity Prep
Trinity Prep vs.
4A State Champ Nease
Nease vs. Armwood
Armwood vs. Washington (
Washington (
Monsignor Pace vs. Belen Jesuit
Belen Jesuit vs. American Heritage (Delray)
American Heritage (Delray) vs. St. Edward's
St. Edward's vs. Warner Christian
Warner Christian vs.
3A
Wakulla vs.
North Florida Christian vs.
Eagle's
Remember. EVERY GAME COUNTS. Insignificant as some may seem, they all mean so much to the rankings.