I get a couple of these "well, this doesn't make sense!" comments every season--heck, I even find at least a few myself, but upon checking out the numbers, the rating being questioned makes sense.
I also usually ignore these, but when the Dean of Class AA reviews and the Dean of Class A reviews posts in the same thread, I take a look. And what I found startled me. Call it a glitch, an outlier, a blip, a "Well, now I know PageStat sucks"
, whatever term you wish to use here. It's none of those. It's simply PageStat doing exactly what it's programed to do, but PageStat has never encountered numbers like these before.
Centennial. Holy Angels. La Crescent. Those three teams may not mean anything but play a part in this explanation.
My first look was at the Falls SOS ratings I posted last week at
https://leepagen.com/strength.htm and I found something unprecedented:
I Falls SOS (gross); 7.19
I Falls SOS (net) 18.44
I've never found a disparity as great as this (11.25 pts difference)
The net is the number PageStat uses. Gross, of course, is the SOS of all their opponents. Net is what you have left after certain games are ignored. Why ignore a game? Because, in this case, when the PageStat rating of I Falls and their opponent exceed the Goal Cap number (currently set at 6), then regardless of the game's result, I Falls rating will decrease.
An example; Team A has a PageStat rating of 20 and plays Team B which has a rating of 10. PageStat would predict a 10 goal win by Team A. What if they win by that exact amount? The game score for Team A is:
(20+10)/2+(6/2)=18
This is the basic formula PageStat uses to calculate every game score. Each team's overall rating is simply the average of all their game scores, adjusted for what I call "momentum factor", which means that more recent games are weighted heavier than earlier games.
The "6" in the the formula I showed represents the margin of victory the goal cap; when the margin of victory is higher than 6, then the goal cap is implemented and 6 is used as the margin of victory.
The goal cap exists because a) teams like Centennial and Holy Angels used to run up huge wins mostly due to being in a very weak conference; way back before the Internet, they were overrated and their early exit from section tournament play proved that. It puts the brakes on teams running up the score in certain games (Moorhead vs. Monticello, anyone?)
Why is the goal cap 6? Mitch Hawker and I used to have discussions about this the goal cap he used, as I recall, was either 5 or 6; I settled on 6 after doing some heavy research on section tournament games since 1998 and learning that 6 is the "sweet spot", sweet spot defined as the goal cap which results in PageStat predicted the highest number of section tournament game winners. I've research a goal cap anywhere from 1 to 8, and also what happened when you remove the goal cap. This research, by the way, is redone every two or three seasons.
By the way, if I remove the goal cap from this season's results, I Falls rating drops from 8th in Class A to 26th.
You might say, and I would agree, that not having a goal cap would result in a more realistic rating for I Falls. But there is something else which happens here other than what I already said about predicting section tournament games. It's called the Law Of Unintended Consequence. One quick example has to do with Holy Angels moving from 12th overall to 4th as a result. Opinions may vary, but I don't see that as being realistic. But examples of this are everywhere.
Back to the Falls; I looked at their opponents and found that, with every Minnesota team they've played, they all had low PageStat ratings, low enough that PageStat ignored every one of those games! The only game their rating comes from is the Fort Frances game. It's not shown in my ratings, but Fort Frances' rating is, guess what? The same as I Falls! Or rather, I Falls rating is the same as Fort Frances at 18.44.
Keeping in mind that we're still early in the season and a lot more games have to be played, the Falls rating will very likely drop. They don't have many tough opponents coming up but they do have a game scheduled with Hibbing. If they lose that one by a substantial margin, this will drop their rating enough that the goal cap will be removed from some of their other previously played games. This will settle their ranking to a more reasonable number.
I played with the numbers a bit, removing their Fort Frances game and found out...it's Fort Frances fault! (the Blame Canada tune from the South Park movie comes to mind
). Removing that game drops I Falls rating from 18.44 to 12.75! and their Class A ranking from 8th to 46th.
So, you might ask, why even include games with out-of-state (and in this case, out-of-country) opponents? Enter La Crescent. Again, back when I had long brown hair, I didn't include those out-of-state games. La Crescent played a overwhelmingly out of state schedule. In 1997-98, here were their Minnesota opponents; North Branch (which had a horrible season), Mpls North/Henry (likewise), St Paul Academy, St Paul Central and Fairmont. Their SOS for Minnesota opponents that year was 7.62. I decided to include out-of-state games. (As I recall, I had an email discussion with JP Piche, who was a La Crescent coach at the time...nice guy) This made their PageStat rating (along with other teams playing a number of out-of-state opponents) more realistic.
The solution? There is no good one. As I said, this was a unique situation. I'm not rooting for them to lose just to make PageStat look better, but if they play Hibbing close, their rating will probably actually go up. Fort Frances losing a couple of games to Minnesota opponents will also resolve this weirdness somewhat.
Karl and Class A guy, thanks for your input! We'll see what happens going forward.
(This response will lengthy due to the influence of some great Hazelnut coffee
)
Lee
