For those of you that want the details, behind the IPS, here is a brief explanation.
The fundamental concept is simple enough. Let’s take Akron-Penn State in Week 1 as an example. Akron averages 30.0 points a game while their opponents normally give up 27.0 points a game. So, their Offense Impact is 3.0 – that is they will typically score 3 more points than their opponent give up. This week that is Penn State who is only giving up 14.4 points a game, so Akron should score 17.4.
Of course, you also have to factor in Penn State’s defensive impact. They are holding team to -11.8 less than they normally score. So, they should hold Akron to 18.2 points. Average those two to get the predicted score for Arkon. Then, do the same to calculate the score for Penn State.
Actually, that produces the predicted score for a neutral site. So, you need to adjust for home field. A typical method is to simply adjust by 3.5 points across the board, but I’m of the opinion that not all HF advantages are equal. So, I adjust between 1-5 based on the team’s ranking (CBS 120 poll) based the idea that the higher your ranking, the more supportive your fans. Top 25 ranked Penn State gets a full 5 points, so I add 2.5 to Penn States expected score and subtract the same from Akron.
Next, you have to consider the relative strengths of schedules. The IPS approach takes some of that into account. But, It really needs to go deeper – to the strength of your competition’s competition. I could come up with some recursive formula, but that’s too much work. So, I simplify my assumption. No doubt some conferences are tougher than others, making some schedules fundamentally more tough. I could try and rank them all, but that’s too subjective, so thusfar, I just bucket them as BCS or non-BCS. I’ve done some modeling and on average BCS is worth +10 over non-BCS.
Finally, you have to figure out what to do about Div. II opponents. Do you include them and then have to worry about calculating all of their IPS statistics? Do you exclude them and then have to back those scores out of their Div 1 numbers? In the end, I did what was easiest. The results of the game figure into your scoring average, but not your opponent’s average. Think of it as getting half-credit.
OK, that’s it you should be able to run the numbers for yourself. Or, just take my word for it.
Posted by Jack Barber 
