Los Angeles, Dallas, Chicago and Atlanta — no, it’s not a listing of the busiest airports in the US; it’s the second round playoff games due to start May 2nd. I’m predicting who will win each series with the exact number of games (a la the stat geek smackdown) as well as the playoff game system I created, where you can bet up to 15 points for the winner of a second round series (you can bet 0 if you think it’s a toss-up.)
(1) Chicago and (5) Atlanta
Congrats, Atlanta on your impressive defeat of Orlando! Now prepare to be destroyed by the Bulls.
Given Atlanta’s dismantling of Orlando, one may think they have a chance at making the series interesting. Looking closer at the series, this logic gets crushed by quality evidence.
The first piece of evidence is the scoring differential of each team during the season. The Bulls had a differential of +7.32, and the Hawks, perplexingly, had one at -0.82. How did they manage to win 44 games? Over the last quarter of the season, this dropped to -0.59. How they beat Orlando is a function of the Magic’s terrible bout of shooting, partly as a result of single-teaming Howard.
While Atlanta had a plan to counter Orlando’s best weapon, controlling Rose is virtually impossible now that Hinrich is likely miss the entire season. Rose will be guarded by some combination of Jeff Teague, Joe Johnson, and prayers. Boozer has turf toe, but he’ll likely play and the Bulls amassed their season success with a long stretch without the Alaskan. He’s less important than people think — the Bulls are arguably better with Gibson.
So this series comes down to how much will the Bulls destroy the Hawks. Atlanta could steal one at home, but home court advantage is worth only 3 points for the average team. (Atlanta is not known for its fans, and that 3 points may actually be lower.) Given that their season differentials are roughly 7 points apart, a 3 point gain in differential means they are nearly two standard deviations apart (assuming a 2.7 point difference in opponents standard deviation, which simply means the spread or variation.) Without going into the statistical numbers, I can safely presume the most likely result is that Atlanta loses every game.
I’m not going to blindly follow the (basic) analytics here, but conventional wisdom is in agreement.
Prediction: 4 games, Chicago
15 to Chicago
(2) Los Angeles and Dallas (3)
This is a really interesting series, for more reasons than Kobe-Dirk, who will both likely eclipse 30,000 career points. Will Los Angeles find its groove? Is Dallas a real contender? Will Bynum stay healthy? How did Barea get such a hot girlfriend?
To get a handle on who will win the series in however many games, I created a couple similar measures to assess a playoff team’s scoring differential versus an “average” team. First I had to predict how many minutes each player and player combination was going to have in the series. This is a very difficult task, one that I partly based off the first round.
The first measure used basketballvalue.com’s adjusted +/- stats for each player, weighing this season’s by 2/3rds and the season before as 1/3rds (I used the season before to eliminate some uncertainty, since +/- can be very noisy.) Then I multiplied by the player’s expected minutes per game, and added up the result. This should be the team’s expected scoring differential against an average team.
For the other measure, I used 82games.com’s five man floor units point differential’s divided by the minutes that unit played. Then the result was simply multiplied by the unit’s expected number of minutes played. The problem, however, is that Dallas has a vastly different line-up than what they used during the season. Butler is injured, Roddy is back, and Stojakovic was acquired midseason and is finally healthy. This makes the +/- measures difficult to pin-down, and the line-ups with Stojakovic and Roddy have very limited minutes. The season long differential does not account for the team they’re playing now, but the individual +/-‘s have large variations because these players have not been playing much. Consequently, the X-factor in this series is Roddy Beaubois, and possibly Stojakovic.
The results are listed below where the first measure if “individual projection” and the second is “player unit projection.” The 7 game row is the projected differential of a seven-game series with the adjustment for LA’s home-court advantage (assumed to be +3 points for one extra home game.)
Individual Player unit Season
projection projection differential
Los Angeles +2.25 +11.6 +6.11
Dallas +3.01 +5.73 +4.23
7 game -0.33 +5.86 +2.31
While the individual project seems off, I just developed this method and I’ll probably change it drastically. But it can be used to compare the two teams, and the results are interesting. Over a seven game series, the average differential between the three is +2.75 for LA (+2.33 without homecourt advantage.) On LA’s home court they expect to win using a Z-distribution 85% of the time, and it drops to just 46% in Dallas. This suggests the most likely outcome is LA in 5 games where the probably of this happening is 33%. I know this is only a model, and there are lots of flaws and assumptions, but I thought I’d test it out because I had problems trying to figure out if LA was going to win in 7, 6 or 5 games. Note that my model has Dallas winning at only 7% of the time.
Prediction: Los Angeles, 5 games
15 to Los Angeles