Wednesday Notebook Dump
Posted by Derek
The Eagles Almanac, a collaboration of nine Eagles fans and writers that previews the 2012 season, came out Monday. I contributed a few pieces, but didn't have time to write up everything I researched. So buy the book, and then if you want more, I'll be dumping out the notebook for a few days.
Back in the old days, today's post would have been a humdinger. There's a lot of data collection involved, reference to some work other people have done, and so many underlying tables that -- even now -- it's clear you readers haven't lost the taste for.
But we're in modern times, so apologies, I'll give you what I've got.
Full disclosure: I set out this offseason to "prove" Asante Samuel is overrated. Not that he's not good, he is, but that if you take away the interceptions, he's not otherwise a great cornerback.
Now, of course, that bolded part is ridiculous. Interceptions are hugely valuable and a guy who's merely "good" but also chips in 8 INTs is obviously someone you want on your team.
Asante is a complicated fellow, though, and his particular ... "style" of play means he has more negatives to set up against those interceptions than the average NFL cornerback. Missed tackles, for starters.
But we've talked about all those other issues for years and we're not learning anything new there. So I wanted to set aside all the extra stuff and just look at how he covers.
For years, I've been unsatisfied with the traditional CB yardage metrics. You'll see guys who have great YPA figures, but then you don't know if that's because they're really that good or because they play in a cover two system where they get to sit down on the flats all day. A true "cover" corner is going to have responsibilities extending much further down the field.
So what I took into my head to do was to look at every single pass thrown over the past few years, figure out what the average result was for each distance, then map the attempts faced by all our cornerbacks to figure out how much they under or overperformed the average.
This is the part where I'm going to skip a lot of the backstory. Know that the database I built contains 88,000+ attempts, average results are broadly similar across all three directions, a linear function is a decent approximation but works better if you take out INTs -- that's a column in itself -- and for every INT I assigned a result of -50 yards, a figure about which we can quibble, but should be reasonable.
I talked a lot in the Almanac about the number of attempts -- which I think is an important, if second order, metric -- so now let's look at what actually happened on all those attempts:
And the numbers settle down somewhat if you drop the (-50) for INTs, but the overall story doesn't change: