In my previous post, I mentioned pushing the 2007 NBA Tournament results live last Friday. Well, I sorta did …

Ok, so I started anyway. I didn’t finish (in fact I conservatively estimate I have 2% of all the content live). So what else is new? But, really, I made significant progress! The hardest part was solving the problem of having each dynamically driven page (players, teams, etc.) contain Wordpress’ commenting system. It took me two solid days (being a 1-man blogger/programmer can take time) but I overcame this obstacle … however it still remains to be seen how well this will work out.

Unfortunately, this means that every entry in my database, whether it be a player, a team, a series, a game, everything needs a separate Wordpress page. Thus, you can imagine how long it will take (there are 300 players, alone). I’m thinking I can get this done by the time I finish the next NBA Sim milestone: September 20 I should be done with the 2008 Legend Counterparting process. I’m actually just a shade over half-way done with that task so I’m right on schedule with it.

So, sorry about the delay but having the ability to comment on every page was well worth the 2 extra weeks in my mind. Later!

I apologize for being away for so long but, believe it or not, I’ve been extremely busy. I finally moved into a Manhattan apartment (I was out in dirty Jersey) and I’ve been feverishly working on NBA Sim junk. Well, I have some crazy news for you all (are any of you out there?)

I don’t know why I’ve been sleeping on this information (it’s been about a week) but the official team rosters for the 2008 NBA Sim Tournament have been set! I’m currently in the process of setting up the players’ attributes and tendencies. This will be so much hotter than the 2007 Tournament as I learned a lot and have made some significant upgrades to the whole process.

Here’s the tentative schedule going forward from here:

» Sep 06 - Push 2007 NBA Sim Tournament results live
» Sep 20 - Finish 2008 NBA Sim Players Counterparting
» Sep 27 - Finish 2008 NBA Sim Players Attributes and Tendencies
» Oct 01 - BEGIN 2008 NBA SIM TOURNAMENT

The old adage is “defense wins championships” yet most of the NBA Sim teams are full of offensive Legends. This is due to the fact that blocks and steals (the only defensive stats recorded) weren’t recorded until the 1974 season. Unfortunately for NBA historians and NBA Sim fans (are there any?) this means we’ll never truly know how great defensive specialists (like Bill Russell) really were between 1947 (the NBA’s inception) and 1973.

Well, just because professional basketball ignored these stats for almost three decades doesn’t mean I can have the same cavalier attitude! For the 2007 NBA Sim Tournament I decided to fake steals and blocks the easiest way possible. First, I would set a shape for each player. NBA 2k7 had three shapes for each position, each with different starting steal and block values:

Once I had a default steals and block value based on a player’s shape, I next tweaked this value using a points pool based on the Legend’s TENDEX value (discussed in detail in my old, Air Craft: Simming the Silver Legend post). Essentially, players with higher TENDEX values had more points with which to tweak their attributes higher. This system felt so artificial however. I wanted something that seemed more organic and based on the Legend’s actual season stats rather than how many TENDEX points they had left in their pool.

So I turned to my other stats-obsessed basketball nutcases at APBR. I found a gentleman by the name of Mike Goodman who said he did some experiments into estimating steals for players pre-1974. His original formulas (in Excel speak) were:

Est. Steals = SQRT(Assists Per Game) - 0.4

Well, I put this formula to the test. I applied the formula to all players from the 1974 season on (sample size of 15481) and came up with estimated steals and blocks. Then I compared this to the actual blocks and steals, determined the difference and then averaged the total sum of these differences. Using Goodman’s formula, the average estimate was off by 16.9 steals (not great … but not terrible either, considering). Tweaking the formula a bit I came up with something slightly more accurate (only off by an average of 15.7 steals).

Est. Steals = SQRT(Assists Per Game) - 0.53

Goodman’s formula for blocks (which came out to be off by an average 20.7 blocks) was:

Est. Blocks = SQRT(Rebounds Per Game) - 1.6

The best I could tweak this to (off by an average of 19.6 blocks) was:

Est. Blocks = SQRT(Rebounds Per Game) - 1.46

Just for good measure I did some regression testing on steals (using assists) and blocks (using rebounds). The steals regression came out to be less accurate an estimation device (it was off by an average of 16.9 steals). The blocks regression, however, was more accurate a predictor than the tweaked Goodman’s formula (off by 14.5 blocks).

Est. Blocks = (Total Rebounds * 0.134) - 3.741

I am eagerly looking forward to using this new tool in the 2008 NBA Sim Tournament (tentatively scheduled to begin mid- to late-August).

Earlier in the week my girlfriend made the claim that “no one really likes” baseball and that it is lame. Now, of the three most popular American sports, I would agree that baseball is my least favorite, but is it true that no one else likes it? Of course not … but it got my stats-obsessed juices flowing.

In fact, after a little research I discovered that baseball has the highest total attendance between it, basketball and football by a factor of four!

Now, these attendance figures are based upon countless, different variables - variables that are completely different between the three sports. First, I isolated what I thought were the most important factors, and ones that could easily be qunatifiable: number of stadiums, number of games in a season, total seats in all stadiums, and ticket price. Here are the figures from the 2007 NFL, 2007 MLB and 2007 NBA seasons:

In order to fairly compare the attendance figures I next had to make sure they were all translated into the same language. Essentially, what I was trying to do with my math was answer the question “with the same amount of games, the same amount of teams, the same size of stadiums, and for the same price, which of the three sports would be more popular?” Then, for each variable I decided whether a lower or higher number was more impressive and added that to the end of the row (+ = higher number more impressive, - = lower number more impressive). I decided to translate all of the attendance figures into “football numbers” since it is undoubtedly the most popular American sport and it produced the smallest figures for the next step.

Formula: (Sport X’s Figure / Football’s Figure)
Example: (Baseball Teams / Football Teams) = (30/32) = 0.938

It didn’t matter, ultimately, which sport I chose to use for the base. Sure, this would affect the figures calculated above, but the end results would still be the same (I checked to make sure). Now, I multiplied the original attendance figures by these newly formulated coefficients. Thus, here are the attendance figures of Football, Baseball and Basketball if all three had the same amount of teams, games, seats, and prices that Football has in a regular season (obviously, Football’s attendance remains the same):

And, finally, to answer my question, I calculated the Popularity Strength. This was just basically comparing all of the attendance figures to the highest one.

Now, of course, I realize this conclusion is only based on attendance. Many fans of these sports participate by watching them on tv. I considered getting the nielsen ratings for each, but I found that too cumbersome and just plain difficult to track down. With that said, however, for all you baseball fans out there, I hate to say it but: my girlfriend may be right.

Total Stadium Seats thanks to Wikipedia
Ticket prices thanks to Team Marketing Report

Well, I goofed.

I don’t know how I let it slide past me but I actually managed to completely switch winners in a sim series from my 2007 NBA Sim. In the second round, the All-Time Sacramento Kings played the All-Time San Antonio Spurs. According to my detailed records, the Spurs clearly won the series 4-1 and yet I advanced the Kings to the next round. Whoops!

I am re-simming that next series (instead of the Golden State Warriors versus the Kings I’m now calculating the Warriors versus the Spurs) and, hopefully, this won’t screw up any further series. However, if it does, I shall dutifully re-sim these as well. I’ll keep you all posted. Also, look for the 2007 NBA Sim results to finally be presented very soon. And it will be very hot!

UPDATE: I just finished simming this series and the Spurs succumbed to the powerful Warriors, ensuring that all the rest of the series I originally simmed would remain the same. A close one, I know!

I really am a bit embarrassed about this one, folks. I haven’t come up with any formula that I’m even sort of fond of to encapsulate defensive performance. So, in the end, I half-assed something called DefeNDEX and used that to determine who should be the defensive player of the year. But don’t put too much weight on it. Yes, once again, Marcus Camby of the Denver Nuggets should win based on his stellar performance, but he won’t. Voters don’t like to echo themselves year after year and KG won’t be winning MVP for his historical turnaround in Boston.

Real-world prediction: Kevin Garnett
Real-world winner: Kevin Garnett

And here’s the All-Defensive Team selections based on my lame DefeNDEX values.

^ - Player was on multiple teams for season. I averaged the PaNDEX values for each team, with each value weighed by how many games they played for the team. For example, if he played 1 game with a 99.9 Pace team (A) and 4 games with a 90.1 Pace team (B) the formula would be: ((PaNDEXteamA * 1) + (PaNDEXteamB * 4)) / (1+4)
Pace - estimate of the number of possessions per 48 minutes by a team (thanks to Basketball-reference.com)
PaNDEX - a formula of my own design (TENDEX / Pace) to remove team play-style bias in stats
DefeNDEX - a half-assed formula of my own design (((Def.REB + (1.25*STL) + BLK - (FLS*0.5)/ Games Played) / Pace) * 100

Stats thanks to Basketball-reference.com

I think I’m officially putting myself into the camp that believes the NBA Most Valuable Player award should go to the best player in the game, regardless of their team’s situation. If a team is really good, they’ll win the championship, but this is an individual player’s award. Actually, with an entire site dedicated to simulating NBA games based solely on historical stats, maybe my decision about this is less finally being made and more finally being admitted to myself.

Yes, Chris Paul surprised everyone and brought his Hornets to the (near) top of the Western Conference. Yes, Kobe is the deadliest scorer in the game. But no one does it like LeBron. I remember him turning down a dunk contest invitation by saying he didn’t want to define himself as just a dunker, but rather as a complete player. 30 points per game. 7.9 rebounds per game. 7.2 assists per game. Almost 2 steals and over 1 block per game. Nearly 50% of his shots go down and his 3-point shot must be at least respected (32%). Yeah, I’d say he’s pretty complete. Complete and very, very valuable.

Real-world prediction: Kobe Bryant
Real-world winner: Kobe Bryant

Here are the All-NBA Teams (based on PaNDEX). The only rule is they must have competed in at least 50 games (or 60% of the season).

^ - Player was on multiple teams for season. I averaged the PaNDEX values for each team, with each value weighed by how many games they played for the team. For example, if he played 1 game with a 99.9 Pace team (A) and 4 games with a 90.1 Pace team (B) the formula would be: ((PaNDEXteamA * 1) + (PaNDEXteamB * 4)) / (1+4)
Pace - estimate of the number of possessions per 48 minutes by a team (thanks to Basketball-reference.com)
PaNDEX - a formula of my own design (TENDEX / Pace) to remove team play-style bias in stats

Stats thanks to Basketball-reference.com

Unlike the Sixth Man of the Year award, stats are easily applicable towards the Most Improved Player. It’s a pretty convincing argument to show that one player is more deserving of the award than another by showing a larger increase in his stats. Sure, there are small gray areas like playing time, previous injury, and trades to consider but this is probably the easiest conversation to “win” using straight up math.

The odds on favorite to win this award is Hedo Turkoglu. He went from being a role-player to a clutch closer this year. But looking at PaNDEX change (explained below), he’s only 14th on the list. Andrew Bynum actually showed the most dramatic change this year, but he only played in 43% of the games this season. I’ve always liked a cut-off of 50 games (or 60% of the year) as this would decrease the impact of outlier games and be a stronger indication of a trend, so I’ve eliminated him from contention.

According to straight up change in PaNDEX, Chris Kaman is clearly the Most Improved Player by over 1 point (the meaning of which I have yet to define … or understand, even) than second place, Ronnie Brewer. Kaman was the best big man on the Clippers as Elton Brand was out for most of the season. But, this really doesn’t explain his increase in numbers. He started beside Elton Brand in both the 2007 and 2006 seasons as well. One year he was decent (17.12 PaNDEX in 2006) and one year he wasn’t (13.39 in 2007). So it’s not like he was just getting some boards and shots that would normally have fallen to Brand. There was something seriously in the way of Kaman’s improvement last year and it wasn’t Brand. The following is an extract from his Wikipedia article:

In January of 2008, Kaman revealed that he was misdiagnosed with Attention-deficit hyperactivity disorder.[3] Kaman spent much of his childhood on a farm and as a child he used to tear shingles off of neighbors’ rooftops and misbehave in school. His apparent ADHD affected his play in high school. He took Ritalin to treat the supposed condition, but the drug killed his appetite. Kaman became very skinny as a result. Kaman, diagnosed with attention-deficit hyperactivity disorder at 2½, found out the classification was wrong this summer. Instead, his brain was in overdrive, working too fast. Tim Royer, the neurosurgeon who discovered the misdiagnosis in July, worked on a daily training program with Kaman to slow down his thought process throughout the summer.

Real-world prediction: Hedo Turkoglu
Real-world winner: Hedo Turkoglu

So where do all the other, popular Most Improved Player nominees stack up?

* - Didn’t play enough games to qualify (I like at least 50 games played)
Pace - estimate of the number of possessions per 48 minutes by a team (thanks to Basketball-reference.com)
PaNDEX - a formula of my own design (TENDEX / Pace) to remove team play-style bias in stats

Stats thanks to Basketball-reference.com

Greg Oden and Kevin Durant went 1-2 in last year’s draft and they were also expected to go 1-2 (in one order or another) for this year’s Rookie of the Year honor. As we all know, however, Oden was out of the contest even before the season started and Durant was only decent, if inconsistent. Lucky #3 pick, Al Horford, took advantage of the opportunity, playing his way into Atlanta’s starting lineup and narrowly ousting Durant for my ROY vote (if I had one).

Real-world prediction: Kevin Durant
Real-world winner: Kevin Durant

* - Didn’t play enough games to qualify (I like at least 50 games played)
TENDEX - formula for TENDEX and explanation why I use it
Pace - estimate of the number of possessions per 48 minutes by a team (thanks to Basketball-reference.com)
PaNDEX - a formula of my own design (TENDEX / Pace) to remove team play-style bias in stats

Stats thanks to Basketball-reference.com

I always found the Sixth Man of the Year award to be a strange one. It’s usually easy to tell who the top 2 favorites are, but it’s harder to tell who even qualifies as being a sixth man on any given team. Sure, many teams announce at the beginning of the year who the first person coming off their bench is going to be - but that is rarely the same person at the end of the year. With so many trades, injuries, and jockeying for playing time, it seems more of an art than a science.

Still, I’ve applied my science to figuring out who 1) qualifies as being a sixth man and 2) is the best sixth man in the league. To figure out each team’s nominees, I removed the first 5 people on each team based on Games Started. From those remaining on the roster I picked out the player with the highest amount of minutes played. This actually came out pretty close to what I “felt” was right for each team. Oh, and by the way, as if I needed to go into any sort of detail at all, Manu is the clear favorite and should be receiving a unanimous decision.

Real-world prediction: Manu Ginobili
Real-world winner: Manu Ginobili

TENDEX - formula for TENDEX and explanation why I use it
Pace - estimate of the number of possessions per 48 minutes by a team (thanks to Basketball-reference.com)
PaNDEX - a formula of my own design (TENDEX / Pace) to remove team play-style bias in stats

Stats thanks to Basketball-reference.com