re: CFB Advanced Metrics and Computer Polls, Week 12

quote:

---

It involves math and sports. It’s a secret beyond that.

---

That makes a lot of sense haha, I’m guessing you’re probably pretty good at it

quote:

---

Basically, I look at LSU's recruiting rankings on average over 3 years. I look at who they beat and how those team's recruiting has ranked.

---

Recruiting is already one a number of variables that some of the metrics use as to create an apriori distribution to initially rate teams since it has some predictive utility. But that distribution gradually loses its influence as real game data and the posterior are added to the model until it’s eventually out of the model altogether.

And those ranking are included with a number of other variables and they are included and weighted based on empirical-evidence, not just some arbitrary designation based on a whim. And those early models clearly have some major misses and flaws, and if those initial variables were still useful then there wouldn’t be replaced from the model so easily.

So the value of recruiting is already reflected in the data by the end of the year, but arbitrarily adding a fixed weighting of anything without any empirical basis would partially ruin that true value and the value of any other variable in any model, and would probably make every model worse.

quote:

---

Is Wisconsin better than Illinois? Is UGA better than South Carolina? If you answer yes to either of those questions then you can understand why a model that is predictive based on a variety of factors can spit out teams higher/lower ranked than their results indicate. To you, because South Carolina beat UGA @UGA then they are for certain a better team and should be ranked ahead of them.

---

You are purposefully being dense at this point. Your comparison isn't even apples and oranges at this point. It's not based on "one game" like the OSU numb nuts like to point out.

Specifically with UGA and South Carolina there is a vast difference in record and resume. Similarly the same can be said in regards to LSU and Bama but LSU also has the result on the field.

quote:

---

guess what dude. this why xiv been in my top 3 for months now

---

quote:

---

Bottom 3 posters: 1. xiv, 2. Jbeam 3. chalmetteowl;

---

quote:

---

I just had a hypothesis about recruiting averages and was trying to see how it tested.

---

If you’re testing something to determine its influence and predictive utility of some outcome model, then you already have the outcome model and don’t need to add the predictor, which will invariably be less reliable and valid than than thing it’s predicting.

It’s like trying estimating how smoking contributes/predicts a lung cancer diagnosis than diagnosing people based on their smoking habits not the actual measure of cancer itself.

quote:

---

guess what dude. this why xiv been in my top 3 for months now

---

Right back at ya, you handsome genius.

quote:

---

You are purposefully being dense at this point. Your comparison isn't even apples and oranges at this point. It's not based on "one game" like the OSU numb nuts like to point out.

---

No. Being dense is choosing a desired outcome, THEN arbitrarily creating and setting the rules to get that desired outcome, then dismissing any example where a strict application of those rules leads to an undesired outcome.

There is nothing wrong with having rules and heuristics to form one’s views. I too would rate LSU above Alabama because they deserve it based on the outcome of the game, despite their similar level of quality as a team.

But to say any model that is intended to objectively and empirical estimate the quality of the teams is wrong because it doesn’t conform to my own subjective ranking based on these rules alone, then they better either always hold true and/or you need to adjust them when they don’t. Just outright dismissing them is a pretty good measure of one’s “density.”

Why doesn't Mike Trout hit a home run on every at bat? He just hit a 450 foot bomb last time up, then strikes out.

WTF?


Variance.

There's also film study and adjustments, among other factors.