Adventures in Line-Setting (and, the Keys to the Game)

Posted on Fri 03 January 2014 in Uncategorized by Alex Dewey

Hello, readers! There are 16 games on Thursday and Friday. I'll probably watch some of them. Before you and I partake in the ritual entertainment provided tonight, though, I'd like to show you something I've been thinking about. It's probably obvious, it's probably trivial, and yet I don't think I've ever made a working example.

One thing that's always bugged me about basketball broadcasts (mostly because I'm such a junkie for the sport) are those "Keys to the Game" bullets you see before broadcasts, where analysts will try to pinpoint the most important few things that both teams can do to maximize their respective chances. Offensive rebounding, "get out in transition," or "get off to a good early start". Hit your open shots, Serge Ibaka! After all, you can turn the ball over a hundred times -- if turnovers aren't one of your Keys to the Game and you handle your keys to the game, you guys are gonna win!

I mock it, but it's a neat and quick little feature that works even for the most pedestrian of broadcasts. We can laugh, but yes, there are "Keys to the Game" in every game. Call them what you will: Leverage points, facets of special interest, stochastic weights that -- pulled or pushed -- favor one or the other team. Call them what you will, but recognize them. After all, they're little things, and mostly trivial. But much like pills in a pharmacy, the most powerful of these little things can be fatal or can save a patient from death. Enter "Keys to the Game".

But here's the thing, even if your team does put all its effort into offensive rebounding, and even if it is all "effort": If you're a bottom-10 offensive and defensive rebounding team going against a top-10 team in those categories, you're probably not going to out-rebound that other team. Oh, sure, your team can, because one game is one game, but at this point in the season? 31 games in, after both of the teams have been trying their best, your stats somewhat reflect your personnel that night, usually. If you have shot creators, your offense will be good. If you have good rebounders, you'll have those rebounding stats. If you have elite rim protectors? You'll probably have a good defense. And... if you're an NBA team in the bottom ten of both rebounding categories, and your opponent is in the top ten of both categories? Then your opponent will probably out-rebound you. And there's nothing wrong with that.

See, there are a lot of ways to win an NBA game. And sometimes that means giving up the offensive rebounding battle because the other team has Love and Pekovic... and because your team has something else going for it, too. You don't necessarily need to be ashamed that you only have two offensive rebounds and they have seven. They have Kevin Love. That's what he does! What's more.... __If you_ lose an individual battle by less than you'd expected, that can be a win, too. If you're outrebounded by the Wolves, and you're not a good rebounding team, but you held the battle close? That swings the game in your favor. Holding the rebounding differential to a minimum (i.e. losing by less_ as opposed to winning the battle) is somewhat of an anathema to how we typically think about sports, but we've all heard the sentence "You'll live with Dirk scoring 30" at some point in our lives, often accompanied by alcohol to deal with the pain. And it's almost always right. You'll live with Dirk scoring 30 so long as Jason Terry doesn't score 30, too. I'll live with the Wolves out-rebounding me if we make them pay in transition. I'll live with the Rockets out-shooting me if they're throwing the ball every which way before shots. I'll live with the Pacers out-defending me if they can't enter it into the post. I'll live with the Spurs out...-not-fouling me (???) if Tim Duncan never sees the ball go through the net.

We'll live with our disadvantages if we can also march forth with our advantages and let the ledger judge the better at the end.

• • •

Very quickly, I did a little empirical stuff for this post, like, with data and such. I haven't figured out how to present quite all the results, but for now, I'll leave you with a few simple projections for last night's games, based only on the average efficiency (and pace) of what we've seen so far, adjusted for the strength of their respective opponent.

 ORL  96.94   CLE  94.30
 GSW  99.75   MIA 102.35
 BOS  87.81   CHI  87.90
 BKN  93.83   OKC 107.35
 NYK  91.32   SAS 103.23
 MEM  95.21   PHO 100.86
 MIL  91.80   UTH  91.00
 CHA  93.90   POR 101.50
 PHI 104.79   SAC 109.08

This isn't adjusted for strength of schedule (or home court, in a huge and glaring omission. We're still ironing this out! Don't bet anything on this yet!). So, the Heat's offense is adjusted for Golden State's defense, but the disparity in schedule that helped to cause those offensive numbers (from being, like, almost the only good team in your conference) is not accounted for. So East-West match-ups are likely more lopsided towards West teams than they appear, so I'd probably nudge West teams up a couple points - i.e. I'm expecting the Jazz to win, the Kings to beat the Sixers by more than 5, the Heat-Warriors game to be awesome and probably closer than what you're seeing. And those Thunder-Nets, Spurs-Knicks, and Bobcats-Blazers games? Yeesh. Fuggedaboutit.

Update, Friday morning: Those projections above turned out to be unfathomably wrong.

  • First, let me note that this next part sounds reasonable. The away teams scored an average of 3.5 points better than I projected (home teams scored an average of 1 point better than I projected). So in terms of total points scored? My projections on average were 4.5 points lower than what we actually saw and home teams did about 2.5 points worse in terms of margin of victory than I'd projected.

  • But, see, my original projections never adjusted for home-court advantage. So even though I was assuming a neutral court, the road teams actually did 2.5 points better than that neutral-court assumption. If normal HCA is taken into account (call it, say, 3 points?), then I'm actually off by 5.5 points per game.

  • And that's not even counting the absolute margin of error here. Home teams were about 11.3 points off from my projection; road teams were more like 10 points (9.86). The margin (by which you'd probably choose your betting lines)? I was off by an average of 12.97 points. What's more, if I'd adjusted for home-court and strength-of-schedule, I likely would have underestimated even more the road teams.

  • By the way, by the stopped clock theorem, I actually got some things right. My total for Bulls-Celtics was .28 points too high and my Bucks-Jazz total was about .19 points too high. Not too shabby. But I also had three games (ORL-CLE [33.24 high], GSW-MIA [34.88 low], and CHA-POR [42.59 low]), where I was more than thirty points off the actual total. In terms of betting lines? I only had one game where I was less than 10 points off the margin of victory in regulation. Cavs-Magic (Cavs outperformed by 2.64), which might as well have been point-shaving the way regulation ended. Plus, I was more than 15 points too high in the point totals for both of those teams. Heh.

In short, and I don't want to belabor the point too much - I chose the worst night imaginable to start doing projections, and maybe my projections are also the worst. I was indefensibly wrong and I'm sorry. The only slight bit of fortune here is that you didn't see these predictions and use them, because you would have lost 50 dollars and held it against me forever.

Anyway, so I still have another day of projections to burn off, using the same model. Warning: The following is canon.

TOR     95.81    WAS     93.70
GSW    102.70    ATL    100.05
NOP     99.60    BOS     97.94
NYK     96.21    HOU    103.84
LAC    105.69    DAL    102.30
MEM     96.18    DEN     97.55
UTH     96.23    LAL    100.36

There you have it. But please don't use these numbers; they are the worst.

Also, if you do use them, remember that I haven't adjusted for home court, strength of schedule (especially East/West disparity), or anything else that isn't offensive efficiency, defensive efficiency, or pace. Those are the only three numbers I used.

Anyway, these projections are not to be trusted. Also, if you do put any remote faith in them (and don't, seriously), I'd bump the home team by a few points up in terms of the margin, give the Nuggets a big boost for the back-to-back, don't trust the Lakers, don't trust the Hawks, the Pellies will probably beat the Celtics, the Knicks just played a competent game so hell is freezing over, and I'd take the Warriors by more than 3 points. Also, I like Bradley Beal. The Wizards' offense is fine. I enjoy watching them. Wizards-Raptors feels like a basketball hellscape waiting to happen. If you gamble on that game, you will feel obligated to watch that game.