clock menu more-arrow no yes mobile

Filed under:

Staturday: Predicting the Pelicans first month

The Pelicans' first month will be a hard one.

Brad Mills-USA TODAY Sports

Without a doubt, this year will be a pivotal year of sorts for the Pelicans. Anthony Davis, the half human, half dinosaur hybrid we have as a cornerstone, is coming into his own. He broke out last year with a splendid, historical individual season. He has the highest PER among players 20 years or younger. A 20, 10, 1, 2.5 line is no joke.

What I will do here is to try to predict how the Pelicans' record will look like after the month of October and November (15 games). I created a very complex but simple model (yes, sounds paradoxical) to actually predict the outcome in the Pelicans' first 15 games.

Prediction Model

Prediction always needs history. "History/Experience is the best teacher" and "history repeats itself" supports this notion. I mean, how can you predict the future if you don't even understand the past? So, this prediction model is based on the 2013-14 season.

Unlike the predictive model I used to analyze the probability of New Orleans keeping the pick (and whether it was worth tanking for), this new model will take into account match-ups. In a vacuum, Memphis (50-32) will win more than half of their games against New Orleans (34-48). And yet, we were 3-1 against them winning by an average of 5.5 points and basically just owning that matchup. So matchup matters.

To account for this, I created a 30x30 matrix that stores the point differential per game (PD) of Team A and Team B. Note also that the rows are the away team and columns are the home team, a positive number indicates that the home team won (column team) and a negative number indicates that the away team won.

Here’s how the 2013-14 season will look like:

201314 PD Matrix

After that, we convert the PD into a W/L% by assuming the linear regression of PD to Total Wins. It varies from year to year, but it’s somewhere around W = 41 + 2.5*PD. So a team that has a PD of 6 (a championship worthy team), will have a record of 56-26 (Spurs had a 62-20 record, the Heatles had 54-28 record). Once we get the total wins in an 82-game season, it’s easy enough to get a W/L%. Take note, that if the W/L% is more than 100 percent (which can happen e.g. a PD of 17 will result into 83.5 wins), we just assume it’s 100 percent. A PD of 17 or above in a particular matchup and home/away setup can happen, especially in an East/West matchup since they only meet twice – once in each others’ arena. The same can be said for when the W/L% is a negative (-17 PD or lower), we just assume it is 0 percent.

201314 WL Matrix

But teams change. The Cleveland Cavaliers of 2013-14 is drastically different (and drastic is an understatement) from their 2014-15 counterparts. To account for this, I included a "Change Factor" to the W/L%. In essence, I asked myself, "how different will this team be from the other, on a purely W/L% basis, from their previous iterations?"

201415 Change Factor

However, something isn’t right – no team, no matter how lopsided the matchup is, can be totally sure they’ll win. In order to correct this, I set some limits – W/L% can only go as high as 95 percent and as low as 5 percent. So if the adj W/L% (i.e. with Change Factor) is 95 percent or above, we set it at 95 percent. If the raw W/L% is 5 percent or below, we set it at 5 percent. This way, no team is almost assured/assured a win or loss (unless its Philadelphia, hehehe!).

For example:

The Cleveland Cavaliers finished last season with a 33-49 record. That's not happening, not with LeBron and Love on board and practically last season's entire team intact. To me, this Cleveland Cavaliers team will finish with a 54 win season (conservatively of course, considering it is their first year together and questions about their defense). Thus, they'll be ~25 percent better than their 2013-14 counterpart. This is what will be added (or subtracted) to the team's raw W/L%.

Now, we have an Adjusted W/L% (adjusting for both matchup, probability of winning and changes in team). Like before, I simply divided the W/L% by the sum of both teams' W/L%. The Pels will play the Orlando Magic at home for their opening game. Last year, in one home game (1/26/2014), the Pels won 100-92 (8 PD). An 8 point differential equates to a 74.4 percent W/L. With AD coming into his own, Asik being brought to the fold and with Holiday and Anderson returning from injury, I think this NOP will be 15 percent better than last year (i.e. I predict they'll win 46 games). For the Magic, it's all about minor improvements all around -- I predicted they'll be 5 percent better (or about 4 wins better than last year's iteration).


Pelicans Adjusted W/L = 74.4 + 15 = 89.4

Magic Adjusted W/L = 25.6 + 5 = 30.6

Pelicans Probability of Winning that 10/28/14 game against ORL = 89.4 / (30.6 + 89.4) = 74.5

Just like before, I did 10,000 simulations for the month of October and November. Unlike before, when I did an additional 10,000 simulations on that 10,000 simulation model, since the numbers are pretty stable, I decided to just simplify the results (tried to remove as much decimal places as I can).

The Pelican's Fate

Here is the probability of the Pels winning X number of games in Oct/Nov (max of 15 games which is the total number of Oct/Nov games for the Pels):

Wins Prob
1 0.0%
2 0.0%
3 0.1%
4 0.5%
5 2.5%
6 8.0%
7 16.0%
8 23.5%
9 23.5%
10 15.5%
11 7.5%
12 2.5%
13 0.4%
14 0.0%
15 0.0%

As you can see, the Pels are hovering around that 7~10 win range. In fact, the EV (expected value) of wins is 8.5 -- almost at 0.500.

Some match-ups of note:

1. November 4, 2014 (vs CHA) - We owned them last year at home. If memory serves me right, that was when AD just straight up went berserk against Bismack and Cody. I don't believe at all that this game is a sure win at home (unlike what the model says) because I think CHA has improved a lot this offseason.

2. November 10, 2014 (@ CLE) - Although I've already added a change factor -- largest actually of any team, it still wasn't enough to account for some things. That away game versus Cleveland will not be anywhere close to a favorable matchup for us (a ~70-30 game for us). In fact, I think it's quite the opposite. We'll be the clear underdogs, healthy or not, momentum on our side or not. LeBron is LeBron and Love is Love.

3. November 25, 2014 (vs SAC) - Although the game is at home, the model says this is a ~75-25 probability against us (we're the underdogs). And what reason do we have of not believing that? They beat us twice at home last year, by a total of 22 points. However, Anderson and Holiday were out for both games. I still think SAC has our number -- we don't matchup well with them but I don't think it's as lopsided as a 75-25 split. More like a 60-40 split.

Add these 3 points up and I think a win total somewhere south of 8.5 is more likely. Say maybe 6 or 7 wins?

As I said, it's potentially going to be a rough start for the Pellies. :(