How much is separating the Atlanta Hawks from the big 3?

You've read all the projections, and all the predictions, so I won't bother listing them all here.  There's three true beasts of the East-Cleveland is considered the top team, with Orlando and Boston basically a toss-up for 2/3.  And almost everyone agrees that the Hawks are the fourth best team.  But the general consensus is that Atlanta is a far-cry from the top 3, and are much closer to the rest of the field.

I was unhappy with some of the projection systems and how they reported the standings, so I did a little bit of work with Dean Oliver's four factors.  I've always been a bit skeptical of methods like win shares, or PER, or most other single factors that attempt to isolate a player's contribution to team wins independent of team context.  It's for that reason that you'll see most projections, like these here, are very highly regressed.  Of course, I found out that I'm not capable of designing a full wins projection system based on the four factors.  I wasn't able to figure out how to piece together some very fractured data in a workable model that was overall consistent.  Maybe given a few more weeks, or someone much more experience in statistical analysis than I, I would have worked some of the details out.

Still, I did come up with some interesting data.  After the jump, a bit of explanation, followed by a few conclusions.

For those of you who don't know, Dean Oliver found that there are four different factors correlated with winning that cover most of what happens on the floor.  These are, in short, shooting (eFG%), turnovers (TO%), Offensive rebounding (OR%), and getting to the line and making free throws (shown as FT/FGA).  And each of the four factors has its opposing counterpart-ie, how well your opponent works with the four factors.  If you're not interested in some dry statistical explanations, please skip a few paragraphs to reach my conclusions.

I started with the assumption that each of these four factors can be combined with its counterpart.  From a common sense standpoint, it makes sense-how well you shoot is just as important as how well your opponent shoots, and how often you turn it over is just as important as creating turnovers, etc (There may be some question as whether this is actually true).

Then I took the four factors data over the last four years from basketball-reference.com, and subtracted each team's individual four factors from their opponents', which served to zero out each of them.  At this point, it should be assumed that if a team had 0 difference in each of the four factors, that team would play exactly .500 basketball.  At that point, I made a separate regression for each of the four factors to winning %, and then found the correlation of each.  The strongest correlation is with EFG%, with TO% and ORB% are fairly close and are middling, and FT/FGA being the weakest (just as Dean described).  I found the weighted mean of the total which gave me an approximation for four factors adjusted win%.

This gave me a result that was correlated with winning % pretty strongly-.932-though I found that my sample was very heavily regressed to the mean.  Every team over the past four seasons ended up clustered from 30-55 wins.  In an attempt to line the data more up with what had actually happened, I found one more regression which basically served to correct the standard deviation.  It somewhat affects my results, but it does nothing whatsoever to affect the win correlation; it's still .932, a very strong correlation.

Now I can look at last year and draw a couple of conclusions.  I'm mostly interested in last year, since we're looking ahead to the next NBA season.  Here's a quick table for our top four teams in the East.

Looking at the four factors, you'd expect Cleveland to be perhaps closer to 64 wins than 66, but it's clear that they did everything quite well.  They're better than average (and in the case of EFG% ratio, far above average) across each of the four differentials.  One big aspect of that EFG% number is the fact that their opponents only made 33.3% of their three point attempts, best in the NBA.  Partly that can be attributed to strong perimeter defense, but it also is partially luck.

Boston does extremely well in two areas; shooting better than their opponents, and owning the boards, but they're prone to turnovers, and don't get as many points at the line as their opposition.  A big part of their strength is 2 point FG% defense, where they held their opponents to 45.4%.  Having Kendrick Perkins and Kevin Garnett are both largely to thank for this.

Orlando is fairly dominating in getting points at the line (a team with Dwight Howard would be) but surprisingly, they don't own the boards like you'd expect.  This is partly due to the lack of a secondary interior presence outside of Dwight-he can't dominate every team on the boards by himself, but he's good enough to break even; Rashard Lewis spends a lot of time on the perimeter.

And with Atlanta, we have a big shocker-their weakness is rebounding!  24th in DRB% and 19th in ORB% adds up to a rebounding deficit.  The reasons for this are plenty and bountiful, and have been discussed plenty elsewhere.   But it's interesting to note that they were 10th in EFG% and and 11th in EFG%.  They're above average in both respects, but despite some great shooting, they're closer to being worst shooting team in the league than they are being best.

One more interesting fact, not shown here-it's possible the Lakers are closer to a 55 win team than a 65 win team.  With Andrew Bynum back, that should change, but it's something worth keeping an eye on as the season proceeds.  At the very least, I'll be interested to watch.

Now, what to expect in 2009-10?  If I had been more successful at creating projections, here's where I'd show them.  Instead, some rather simple analysis, with numbers mostly taken from 82games.com.

The biggest addition for Cleveland is obviously Shaq.  He figures to play at least half of their minutes at center, more if he can stay healthy all year, taking mostly Ben Wallace's vacated minutes, as well as some of Zydrunas Ilgaukus' and probably Anderson Varejao's.  Compared to them, he's easily a positive effect on EFG% differential and FT/FGA, especially over Wallace.  However, at this point in his career, he's only essentially even in REB%, and he's something of a liability in terms of TO%.  But that basically amounts to a solid positive overall effect for the Cavs.  They also added Anthony Parker, who'll replace some minutes played by Wally Szczerbiak and some others played by Pavlovic.  That's actually a net loss because Szczerbiak was a quality defender for Cleveland, while Parker is an inefficient scorer who can add some rebounding.  All in all, Cleveland improved, and we should see them closer to 67 wins.  They also added a good role player in Jamario Moon.

The biggest addition for Orlando is Vince Carter, and he's essentially replacing Hedo Turkoglu.  They also expect to have Jameer Nelson for more than 42 games, but that's not something I can easily project.  But Vince Carter for Hedo Turkoglu-that actually seems to be a net loss for Orlando.  Vince is only average in terms of EFG%, and he's by no means a lockdown defender.  Turkoglu, over the last few seasons, has been a plus defender (yes, he has), even though he also is fairly average in terms of shooting %.  Vince does take better care of the ball than Hedo, but Turk generally provides a positive rebounding impact.

Boston is much tougher to gauge.  They added Rasheed Wallace, but if KG is healthy, it's tough to determine just how the minutes will be distributed where he's playing at PF/C.  But I will say this, every minute that he takes from Kendrick Perkins is probably bad for the Celtics.  Perkins is an efficient scorer, when the opportunities come.  He rebounds better than Rasheed.  He doesn't foul as often, either, so his opponents don't get to the line as much.  The only area that benefits Rasheed is in terms of Turnovers, where Perkins has always had troubles hanging on to the ball.  However, he's a much better option than Brian Scalabrine, Glen Davis, and Mikki Moore for all the back-up PF/C minutes.

Even if Orlando and Boston regress slightly, that's still a lot of ground for the Hawks to make up.  According to four factors data, they're not 12 wins behind Orlando, they're 16, and if Orlando loses 3 wins because of the difference Hedo and VC, that's still 13 wins to make up.

Which brings us to Jamal Crawford, the Hawks' big addition this year.  It's been said plenty of places, but you can basically expect him to replace the year Flip Murray had last year, as Flip's career year is pretty well in line with Jamal Crawford's averages.  However, it will be tough to expect Jamal to replicate the positive effect that Flip had on EFG% on both offense and defense.  Flip was fairly average while being a plus defender last year.  Jamal has always been somewhat of a defensive liability, while not a particularly effective shooter.  Both are/were deficient in rebounding, about to an even extent.  Jamal actually gets to the line at a better rate, which is not what you'd expect from a near exclusive jumpshooter.  And he generally takes care of the ball more than Flip, at least compared to the large number of shots he takes.  Basically, this evens out.  I haven't got any kind of numbers to make any reasonable projections for Jeff Teague, unfortunately.

Joe Smith is a great improvement over Solomon Jones.  Unfortunately, I don't expect him to be limited to just the 675 minutes that Solomon played last year.  Any minutes he takes that could go to Zaza, Horford, or Josh Smith are probably negative minutes.  I'll call his impact fairly even overall.

Here's my rough, marginally stat-based predictions for 2009-10:

Cleveland: 69-13
Boston:  65-17
Orlando: 60-22
Atlanta:  49-33

Many thanks to Bret from Hoopinion for agreeing to look my numbers over and make sure they made sense-I also owe a shout-out to the APBRmetrics forum for providing me a few helpful links to look over.  If anyone has a sick interest in raw data, send me an e-mail and I'll shoot you an excel spreadsheet with charts and stuff.