Random SABR NERD rant

[Sven Draconian]

1. It is an oversimplification to say SABR stance is "Steals are bad don't steal." The official SABR stance is do not sacrifice outs. The break even point for stealing being a worthwhile risk is around 70%. Stealing is bad if you don't succeed enough times to justify the attempts.

2. The fractions for OBP and SLG are using two different denominators. One is divided by plate appearances and one for official at bats.

Isn't the break even point in the "current offensive environment" only about 65% or so?

 

[Warner2BruceTD]

I just made a NERD post on another site. See if you can guess the player being discussed:

A little early for this, but here goes.

His K% is down, and his BB% is a little up. That's good.

But just like last year at this time, his BABIP (.375) is way over his career baseline. If it stays that high, he's going to have a damn good year, because his power hasn't really come yet. If it normalizes, he's probably looking at a similar slash line as usual, because the walk rate isnt up enough to make a huge difference, and I expect the HR's to come.

If you are looking for a reason to think that BABIP will sustain, his line drive rate is way up (6% higher than his career rate), which suggest that his BABIP isn't up due to extreme luck. He's hitting the ball more frequently than ever, and harder.

EDIT - fwiw, his hot start last year was for two months. we're not even four weeks into the season yet.

 

[Senser81]

What do you mean by this?

It means that the writers for FootballOutsiders are incredible douchebags, and while some of their statistical breakdowns are interesting, most of it is self-aggrandizing hochum. Not everything needs to be thoughtlessly multiplied by 0.8 or divided by 5.

 

[Sven Draconian]

It means that the writers for FootballOutsiders are incredible douchebags, and while some of their statistical breakdowns are interesting, most of it is self-aggrandizing hochum. Not everything needs to be thoughtlessly multiplied by 0.8 or divided by 5.

Gotcha. I wasn't sure if that is what you were saying, or if you were putting them on the same level as Law/Sheehan ect in terms of accuracy.

Assuming they are the same outfit that produced SackSEER and the QB analysis they pegged:
JJ Watt and Jason Pierre Paul as massive busts with few of their better picks actually coming through (but hey, the pegged Von Miller....)

They also had Mark Sanchez as vastly superior to Matt Stafford. Stafford has been a little disappointing (then again, he wasn't every projected on the Andrew Luck level, or even the Matt Ryan level), but he certainly isn't fighting for his job and he's a more or less consensus mid-tier QB.

Of course, most of Football Outsiders shit is just made up. Bill James and the first wave SABR stuff was all scaled to fit actual results. The runs created formula is accurate because it actually predicts runs created. It's not arbitrary.

FO has put out some ok work. Their research on collegiate rushing attempts is relatively accurate. Running the football has little correlation to winning (other than winning teams run out the clock in meaningless time). That stuff is all pretty amateur, and there's been much better work posted other places dealing with the same topics.

Now, the guys at ProFootballFocus have some really awesome stuff....

 

[Senser81]

Of course, most of Football Outsiders shit is just made up. Bill James and the first wave SABR stuff was all scaled to fit actual results. The runs created formula is accurate because it actually predicts runs created. It's not arbitrary.

Yeah, thats my main issue with the FootballOutsiders group. You ask them why they are multiplying rushing attempts by 0.5, or if gaining two yards on 3rd-and-1 gets the same "success points" as gaining 42 yards on 3rd-and-1, and you don't get an actual answer. Most of their equations are kind of made up so that the final number ends up being whatever they wanted it to be from the start.

 

[Hasselbeck]

Yeah, thats my main issue with the FootballOutsiders group. You ask them why they are multiplying rushing attempts by 0.5, or if gaining two yards on 3rd-and-1 gets the same "success points" as gaining 42 yards on 3rd-and-1, and you don't get an actual answer. Most of their equations are kind of made up so that the final number ends up being whatever they wanted it to be from the start.

:golfclap:

I absolutely hate Football Outsiders. Baseball at least has a large enough sample where these Good Will Hunting type equations work, football is not a mathematical game by any stretch.

 

[Warner2BruceTD]

Gotcha. I wasn't sure if that is what you were saying, or if you were putting them on the same level as Law/Sheehan ect in terms of accuracy.

Assuming they are the same outfit that produced SackSEER and the QB analysis they pegged:
JJ Watt and Jason Pierre Paul as massive busts with few of their better picks actually coming through (but hey, the pegged Von Miller....)

They also had Mark Sanchez as vastly superior to Matt Stafford. Stafford has been a little disappointing (then again, he wasn't every projected on the Andrew Luck level, or even the Matt Ryan level), but he certainly isn't fighting for his job and he's a more or less consensus mid-tier QB.

Of course, most of Football Outsiders shit is just made up. Bill James and the first wave SABR stuff was all scaled to fit actual results. The runs created formula is accurate because it actually predicts runs created. It's not arbitrary.

FO has put out some ok work. Their research on collegiate rushing attempts is relatively accurate. Running the football has little correlation to winning (other than winning teams run out the clock in meaningless time). That stuff is all pretty amateur, and there's been much better work posted other places dealing with the same topics.

Now, the guys at ProFootballFocus have some really awesome stuff....

Good post.

The major problem with what FO tries to do a lot of the times, is football just isn't cut out for this type of analysis. Maybe on a team level, but on an individual level player performance is tied far too strongly to the performance of the other ten players on the field.

Also, Senser is 100% right with the random fractions and formulas. You see some of this in baseball metrics, but in general most can at least be explained enough that you can see where they are coming from, even if you ultimately don't like the idea. But some of these wacky FO formulas are seemingly pulled from thin air.

 

[FedEx227]

Football and Basketball analytics cannot be done with formulas. Just too much dependency on your team, where the rest of your team is, where they stand, what they do, who they are, how they play, etc.

Baseball can be broken down into 1 on 1 match-ups, basketball and football cannot.