re: Week 8 Power Ratings and SEC slate

So you weight blowouts against terrible teams the same as close wins against good teams? Lmao

quote:

---

So you weight blowouts against terrible teams the same as close wins against good teams? Lmao

---

No. You just made this up. I never said this nor implied it.

But you also sound like one of those casuals who thinks there are 10 good teams and everyone else is "terrible".

quote:

---

Are you weighing turnovers or havoc rates?

---

Let me think about this, actually. I'll need to run some tests. I only add things like this if it improves the MAE (mean average error). Most of the time, that's not the case.

Regardless, it's not going to produce the outcome that most people on here would want. It would hurt Kentucky, ND and help LSU, Oklahoma. But if it's more accurate for predictions, then it's more accurate for predictions and should be included.

quote:

---

This would hurt Kentucky.. Jesus Christ, you people are morons. Kentucky has BENEFITED from turnovers. The model would see Kentucky has been very lucky in a few games and it would dock their rating. It would actually help teams like LSU and Oklahoma who turn the ball over.

---

Then your model is stupid if it doesn’t account for a defense creating turnovers. And instead punishes them for it.

You aren't looking at this objectively and you're cherry-picking games. You have to look at the entire body of work.

If Florida didn't have two very bad losses - blowout losses, they would be a good bit higher. Those were potent losses.

Outperforming expectations means exceeding your projected margin of victory against another team. This may be a simpler way of viewing it:

LSU and Florida against expectations (spread) (+ = overperformed, - = underperformed)
LSU:
v Clemson = +31
v LA Tech = +1.5
v Ole Miss = -13
v A&M = +17
v Kentucky = -11.5
v Miss St = +19

total: +44 (LSU has overperformed by 44 points)

Florida:
v FAU = +18.5
v Auburn = +1.5
v Ole Miss = +17
v Missouri = -39.5
v SC = -30
v Texas = +15

total: -17.5 (Florida has underperformed by 17.5 points)


Florida was favored against Missouri by 11.5 and SC by 11. If Florida would have lost those games by say 3 points in both, their total would be +23.5. See the difference? The model doesn't work exactly this way, but this is just a simpler way of putting it. The model looks at offense and defense separately and it doesn't weight every game the same, whereas, just now I only looked at the total and I weighted every game the same. In reality, LSU's total is little lower and Florida's total is a little higher. But I hope you at least understand the point being made.

Keep in mind, the model is trying to PREDICT SCORES. That's it's only purpose. When LSU blows out A&M (#30 team in the ratings), it matters. And it matters more than LSU blowing out LA Tech (#75 team in the ratings). You may think A&M is a "terrible" team, but they aren't. They are better than 75% of the teams in the country.

Again, it's a gradient over 138 teams - there aren't 10 good teams and the rest "terrible".

quote:

---

You aren't looking at this objectively

---

We are objectively calling you a retard.