Hollinger underrated their defense prior to the year. And if the stars play during blowout fourth quarters the point differential would be even bigger. 82games.com is much better for individual defense I'd say, especially for perimeter players. Counterpart PER is VERY useful, it would show you for example that Chris Paul is not an elite individual defender. In past seasons Varejao also qualifies for mediocre D.
It isn't a percentage per se (despite the name), it is Estimated points per possession. Over the course of a season, not individual game, .44 is a good coefficient.
Mike, you're killin' me. That stuff makes my head hurt. In case you haven't figured this out yet, there's a very self-serving reason why I'm targeting this stuff to casual fans...I'm a better writer than a statistician. And this applies even if you think my writing sucks. As for calculating PER, I remember a poster named sdeezy who posted a link to a sort of down and dirty way for mere mortals to calculate a PER facsimile. I tested it and it was pretty damn close. On my Excel worksheet, I named it "lwts PER," but can't remember what it stood for (Logical Weights PER?). I've begun drafting the PER piece. As I did with the TS% article, I'll put it up here first so everyone can make fun of it.
Thanks! I really like the changes made to the final as compared to the draft. Very solid and thought-provoking such as the AND1.
They won a close one today with the full roster. 3-10 (or 3-9 if you believe the graphic they showed today). :>
Do you think there's a possibility that Hollinger messes with his formulas as the season progresses? I watch his team rankings on a daily basis and there are always weird things happening. Some of this may be attributable to old games coming off the books in his recent game input; but other times, there are giant fluctuations for teams that don't play on a given night, and teams that come off big wins see their number drop, not improve. I understand that there should be a good deal of noise, but it seems to me that his numbers are often arbitrary and counter-intuitive. I never feel that way with Nate Silver's numbers, which seem to bring clarity to my non-numerical perception of what's taking place.
Are you talking about a Hollinger power ranking formula or PER? I actually don't remember anything about his power rankings. PER might occasionally see shifts because ultimately it's a relative comparison. PER is basically a composite of individual player stat rates, per 48 minutes, compared to league averages for those rates. I've never tracked it, but I can see how that would fluctuate a lot early in a year and then start to "congeal" to a more solid set of numbers as the year goes on.
Counterpart PER isn't as useful as you think based on the way it is generated. It's completely automated and no corrections are made for cross-matchups on defense. So the numbers aren't as reliable as you'd think. So it's sort of like APM in that you need a huge sample size to get a stat that gets close to the truth.
I don't think Hollinger tweaks mid-season. His PER has hardly changed at all, if any, over the years. He has experimented with some different defensive formulas. I think the Power Rankings have some rolling 10-day or 10-game component. Maybe this is why it is noisy.
It's much more specific than defensive rating, which automatically credits/penalizes you for what 4 other guys do. Unless you're a Big or some elite perimeter defender it probably means little to nothing. Over the course of a career? It makes no difference who "cross-matchups", 90% of the time you're on a guy at about your size. And even undersized Centers like Chuck Hayes still get credit for defending PFs/Centers on 82games. Also APM probably has the goofiest results out of all advanced stats, I'd rather look at SPM, but of course the best thing to do is look at various things.
The individual defensive ratings are not nearly as good as PER. PER has the occasional outlier the individual defense ratings are consistently noisy. I think the fact that D is so dependent on team scheme is a big factor.
So they have won 2 close ones in a row...The interesting thing is Hollinger is trying to predict the future. If you want to analyze play to date, heck yes, W-L would be critical. But for teams that consistently have a good point differential and are poor in close games, perhaps they are statistically likely to figure out how to win close games in the future due to the dominance suggested by the point differential. It's certainly something that the coaching staff would be working on. Now, on the other hand, you could be on to something... I did say at the start something about pure point differential leads me to believe the Heat may be overrated.
Basketball-reference did a post on this. In the first 17-19 games or so, what were the odds that the Heat were just "in a slump"? As in why were their stars playing weird, was it just a slump? Statistically they found that to be impossible, they were feeling themselves out I guess but they weren't just in a statistical slump. Basically chemistry quantified. Now after that was taken care off, they have certainly improved. So they did technically grow as a team, and when they were 9-8 (based on point differential and adjusting for schedule, so it isn't something to just ignore) they were still ranked as the second or third best team in the league. It turns out that was a pretty good ranking, as losing close games comes down to luck if the point differential is small enough. Miss a replay here, or bank in a crazy shot there. They fell into that universe where they got a bit unlucky, but at the start of the year they were also a different team and trying the wrong kinds of things. That's true, it is a tough world where taking into account Defense isn't perfect. But over the course of a few seasons, maybe not half a season, those defensive ratings can mean something on 82games. Personally I think it affects mediocre defenders the most, making them look good/poor based on what their team does. A Dwight howard or someone like that is probably going to look good with almost any defensive metric.
