Tuesday, April 30, 2013

The 2013 Season Predictions Review

As the 2012-13 season has closed, it's time to look back on season with reverie -- and critical on our perception of what we thought would happen.

In my Denver preview, I discussed at length how important and underrated Nene was, and how McGee was a basketcase who wouldn't help a team despite his stats. Well, I was mostly correct because Washingtion went on a great run with Nene to end the year, but Denver stormed the league in the second half as well. I think that has to do with their other players improving, like Koufos, especially since McGee came off the bench.

My Lakers preview? Well, you can imagine how that looks now. I mentioned how putting a bunch of high usage players together increases their efficiency. Er, whoops. But to my credit there's this sentence: "Injuries are also a concern, and those issues could affect this team more than almost any other one because their bench is, simply put, deplorable." Injuries destroyed the team, but in hindsight the team just didn't fit together. For my Warriors preview, even though my win prediction was inaccurate I was actually pretty spot on: the article was about how, due to Bogut and Curry's injury uncertainties, the team had the highest variance of what we'd expect. Curry had a healthy year, playing a huge amount of minutes, while Bogut at least saw some action, while the team put in a good defensive system to essentially replace his value. As a result, they were an above average team and are now making noise against the Nuggets in the playoffs.

Random predictions

In previewing the season, I listed a random prediction for each team. It was all in good fun, and some weren't serious. But read today? It's pretty hilarious how wrong I was. I'll go through them briefly:


Minnesota Timberwolves: Kirilenko had a fine year as I suggested, but Love had no opportunity to flirt with a 50/40/90 season with 15 rebounds for a month. But someone else had a 50/40/90 season....

Utah Jazz: "Fans chant for Kanter to replace Jefferson, but it's the Millsap-Favors lineup that kills." Well, that didn't exactly happen -- looking at their lineup data, they really didn't play in the frontcourt together. Oops.

Oklahoma City Thunder: "Thabeet will once again fail but not after showing a couple promising games. ... Ibaka will put up heavy DPOTY consideration again, even though the subtraction of Howard decimates the Magic's defense." Betting against Thabeet is like taking candy from a baby, and Ibaka indeed garnered well in the defensive player of the year voting, inexplicably.

Denver Nuggets: JaVale McGee has a respectable season with some maturity? Hm, I was almost correct. He's been better, and on a better team his strengths are showcased.

Portland Blazers: "Lillard will be a weak starter at best." ...In this case, I'm glad to be wrong.

Golden State Warriors: "Curry will have another injury-plauged year, but a new player will emerge from out of nowhere and impress in typical Warriors fashion." Curry was healthy enough to break the three-point record. It's hard to be more wrong than that. Breakout player? Draymond Green?

Sacramento Kings: "Thomas Robinson will end up as runner-up rookie of the year with the still mercurial Cousins as one of the most intriguing young frontcourts." I didn't even get the team correct.

LA Clippers: "Chris Paul will shoot 90% from the line, but the team will still finish in the bottom three in free-throw shooting." I actually wasn't far off -- Paul was at 88.5% yet the Clippers were 27th in the league. So close!

Phoenix Suns: "Despite Nash's departure, Gortat has a career year averaging 17 and 10 on healthy percentages. And everyone forgets about the team." The latter point? Completely right. But Gortat averaged 11 and 8.5 and threw his team under the bus.

LA Lakers: "The Nash-Howard duo is an unguardable combination, but the year will be marred by discussion of how often Kobe shoots. (Hint: too much.)" Uh, yeah, well, the year was marred by a discussion of many things the Lakers did wrong, and one of them was Kobe's volume, so
....

Houston Rockets: "The Rockets surprise people again through their gluttony of forwards, Asik's defense ... a record that keeps them out of the playoffs but far from the top of the lottery." At least I knew they were going to be good and that Asik wouldn't be overmatched in a starting role.

San Antonio Spurs: "After people catch on the Spurs won't go away, they actually don't finish with the best record in the west; they go second after Oklahoma City. Also, Parker misses too many games." I was completely right for once, even down to Parker missing games.

Dallas: "Eddy Curry plays 12 games." Well, I was only off by ten games -- he played two. Strange as it sounds, but at the beginning of the season there was some actual discussion about his signing with the Mavs.

New Orleans: "Anthony Davis will narrowly miss the all-star game, and many will discuss why he should have been on it." Davis kept getting injured and his minutes were too low. Otherwise he might have had a shot ... at least in the east.

Memphis: "Tony Allen and Tony Wroten combine to form arguably the most formidable backcourt defending duo in short time." It was a formidable defense, but uh Allen-Wroten only played eight minutes together.

Toronto Raptors: "Lowry and Jonas push the Raptors toward the direction of the playoffs, narrowly missing the extra season as everyone forgets Calderon." In a way, this was quite accurate, as they traded Calderon and didn't seem to miss him.

Philadelphia 76ers: "Bynum will miss at least 20 games and the 76ers will nearly miss the playoffs." They missed the playoffs, and I was technically correct he missed at "least" 20 games ... in another more accurate assessment, I was wrong about his missed games.

New York Knicks: "The Knicks perform well with Amare injured, but when he comes back and Carmelo plays small forward more their record dives and they refuse to adjust." This sorta happened, but fortunately Amare got injured again. Unfortunately to their bank account, he has a huge contract.

Brooklyn Nets: "Brook Lopez ups his rebounding to ... 6.2 a game. Their wings of Deron Williams, Joe Johnson, and Gerald Wallace often outrebound the frontline, hearkening back to the Kidd-Carter days." Lopez averaged 6.9 rebounds, but his minutes were pretty low. It was a nice season for Lopez, who rebounded much better. But the rebounding story this year is that Reggie Evans rebounded like prime Dennis Rodman, posting the second greatest rebound rate in league history.

Boston Celtics: "Fab Melo proves to be useless, Milicic proves to be himself, and Jared Sullinger puts up some pretty numbers with some pretty bad defense." In hindsight, it's surprising people thought they'd be good.

Chicago Bulls: "Bulls struggle when Deng gets injured and Boozer keeps getting more minutes than Gibson. They finish sixth in the conference." Deng somehow didn't get injured, even though he was basically driven into the ground, Boozer got more minutes, and yeah they finished sixth in the conference. Not bad.

Indiana Pacers: "Paul George has a breakout year with 18 points a game." Everyone's breakout candidate, but I was pretty accurate with his points -- he averaged 17.4.

Detroit Pistons: "Together, Monroe-Drummond average 25 points, 18 rebounds, 3 blocks, 2 steals, and 3 assists." They actually averaged 23.9 points, 17.2 rebounds, 2.2 blocks, 2.3 steals, and 4 assists. That's not a bad center combo.

Cleveland Cavaliers: "Irving makes the all-star team averaging over 20 points a game." Yup. He averaged 22.5, comfortably over 20 but not destroying the mark. But to be fair this was an easy one.

Milwaukee Bucks: "At times, the Bucks' plan of Ellis and Jennings ball-hogging alongside long-armed defenders from Moute to John Henson to Udoh works, but at most other times it doesn't -- a bottom six offense and a bottom 13 defense." Bottom 9 offense and top 12 defense, thanks to Larry Sanders, who came out of nowhere.

Miami Heat: "James' MVP-campaign nearly comes to an end, but when Wade gets injured and the Heat maintain a big lead in the standings he wins it over Chris Paul." Er, Wade finally got healthy and they pulled off a 27 game win streak. Same thing!

Atlanta Hawks: "Even though their "star" left for greener pastures, the Hawks improve their win percentage with the further development of Teague, their outside shooting, and a healthy Horford." Yeah, they got worse....

Washington Wizards: "John Wall has a break-out year averaging near 9 assists and shooting a respectable percentage from the field for the first time." At 44% from the field and a TS% of 52, closer to the league average, he was finally respectable on offense, but his assists were at 7.6. However, he's started to capitalize on his defensive potential.

Orlando Magic: "Orlando will give up 130 points one game while losing by over 40. This is what happens when you replace your starting center, Dwight Howard, with Glen Davis." They were so close! And not for lacking of trying. One game against the Raptors of all teams they lost 123 to 88, and in another they lost 107 to 68. Glen Davis was indeed a terrible starting center, but he got injured and opened up a spot for rebounding savant Vucevic.

Charlotte Bobcats: "They increase their wins to an impressive ... 15. Veterans Sessions and Haywood have no point to participation in competition. By the way, they're paying Diop this much. Try to guess. It's fun!" Well, they had 21 wins, but tripling your win total isn't impressive when you started at 7 wins the year before. I had forgotten how much they were paying Diop this season. I guessed 5 million. Yikes....


Win totals

In reviewing my season prediction for games won, I wasn't Nostradamus but I wasn't terrible either. Truthfully, I used a very rudimentary system, and was curious about how it'd perform. To judge the predictions, there's root-mean squared error (take the sum of all the squared errors, divide by the number of predictions or teams, and then take the square root of the result.) Here's a handy link for a summary of numerous predictions made. How did I perform? Well, I actually ended up in between two Wages of Wins predictions:

My RMSE:
7.73

WOW1:
7.61

WOW2:
7.87

Wages of Wins made two sets of predictions. The second iteration exists because the author changed his minutes allocation. While the first version was better, I don't think it's fair to put out two projections in competition. I wish I had made two separation predictions so I could just flaunt the better one. In essence, I think it's fair to say my modest win projection beat Wages of Wins. However, there's another issue: instead of using integer win predictions, they used a decimal form, and I didn't. To put us on equal footing, I converted their numbers to integers: WOW1 goes to 7.71 and WOW2 to 7.91. Thus, my rough prediction was essentially tied with their first set of predictions, while it beat their real (final) prediction by a fair margin.

Breaking down my win predictions, I was of course destroyed by the Lakers, who suffered a calamitous season of outrageous fortune. I was 14 wins too high; however, I was 18 wins too low for the Golden State Warriors, who actually had a healthy season of Curry and also outperformed their point differential. For both New Orleans and Phoenix, I was 12 wins too high -- the frontcourt of Ryan Anderson and Anthony Davis didn't have much of an impact, and I thought it would, and Nash's departure destroyed the Suns' season. Likewise, Orlando imploded without Howard, even though they had some quality young guys. I was terrible in the west, where my RMSE for those teams was 8.90, while I was pretty respectable in the east at 6.35. But for nine teams I was within three wins, and one team I hit it point blank. Unfortunately for Sacramento, it was the Kings at 28 wins. And strangely enough, I underestimated both teams in the Harden trade -- Durant unleashed an offensive season for the ages, and while Harden and Asik performed about as well as I thought their motley cast of young players were hard to predict.

My takeaway from the predictions? Minutes projections are extremely important, and injuries are a last frontier in NBA analytics. Understanding Howard's injury, Bynum's knee, and Curry's ankles will save a season prediction from going sour. I now have an idea for a pretty interesting win projection model, but alas -- it won't understand Kobe's Achilles injury, and frankly modern medical science doesn't either.

No comments:

Post a Comment