I've been banned from the Wages of Wins Journal. What can I say? I don't like to follow the rules. I post under different names, mock people, and generally do things I wouldn't do in person. Gotta make it a note to work on my internet etiquette. I'm not a total douchebag. Just half of one.
What do I have against the Wages of Wins? Nothing really, I suppose I just don't like the way some of its adherents treat the system as if it were The Gospel. There are many reasons I don't particularly like any of the "statistical" basketball models, but the short story is that I put "statistical" in quotes for a reason. I have no hatred of statistics. If you've paid close attention, I love them, as I graduated with a degree in Mathematics (which does not make me an expert in most every sense of the word "expert"). But the "statistics" that the models gather aren't the data that reflect the game. Points? Rebounds? Assists? Plus/Minus? If I were to create a system, the data I would gather would be a lot more specific than that. Junk in, junk out.
I divagate. The purpose of this post to respond to Wages of Wins devotees who take a gander at the projections based on that system and say "look how close they are"! I decided to do my own investigation, and compared the projections of the Wages of Wins, taken from its own pages, and compare them to what eventually happened, as well as what would have happened if I decided to use it and bet in Vegas.
To begin, I took the projections of Wages of Wins from this page on its site. All the explanations for the different types of projections are on that page, so I won't repeat them here. Do some work on your own, lazy butt!
Then I looked for the first page that had NBA regular season Team Total Wins odds. Voila! Google is friggin' awesome. And how serendipitous, the date of the odds is 10/17/2008, and the Wages of Wins post is dated 10/28/2008, giving WoW a full 11 days informational advantage! Wah wah wee wah.
Then I compared them all to the actual results of the 2008-09 NBA Regular Season.
Here are the results (I hope I can figure out how to make these tables as pretty as me wife):
|2008-09||07-08||Berri Rookie||3yr avg, BR||Vegas|
|Team||Actual Wins||Wins Produced||Pelton Minutes||PM||PPP||10/17/2008|
|Golden State Warriors||29||46.2||37.0||38.9||42.3||37.5|
|Los Angeles Clippers||19||23.5||42.7||32.9||32.1||34.5|
|Los Angeles Lakers||65||59.0||67.1||55.0||60.0||54.5|
|New Jersey Nets||34||30.1||33.8||32.1||23.7||27.5|
|New Orleans Hornets||49||53.3||56.4||54.0||45.1||51.5|
|New York Knickerbockers||32||25.9||24.5||28.2||36.5||32.5|
|Oklahoma City Thunder||23||20.2||16.5||18.9||21.2||25.5|
|Portland Trail Blazers||54||39.6||40.1||39.3||54.2||44.5|
|San Antonio Spurs||54||52.4||53.2||52.2||44.8||48.5|
What does all this mean?
Here's the upshot:
Using the 07-08 wins produced (no rookies, same minutes), the model went 16-14 against Vegas' marks.
The second method, implementing rookie projected production and some guy's projection of 2009 minutes, the model went 16-14.
The third method used rookie projections, projection of minutes, and 3 year averages of production. That model went 15-15 against Vegas.
The last method was some conglomeration of methods (look it up, lazy butt! wait, I guess that makes me lazy for not explaining it) that one went 16-14.
I guess it's not that bad, but you have to think that you could do better than that. Of course, you'd be wrong - Vegas sets these odds pretty well as you see. If I were to bet 100 bucks on each of the teams using one of the 16-14 models, and assuming a 10% vig (I cough up an extra $10 on losers), I'd still come out up $60. If I did the same and went 15-15, I'd be down $150. Ouch. It's a thin line between success and suckess.
I want to make clear, I'm not "debunking" anything here, as you would need several years of these comparisons if you even want to begin making conclusions about the accuracy of WoW's projections. The results shown here are for entertainment purposes only, like your calls to Jamaican fortune tellers, except not as expensive.
I do remember learning some elementary statistics in my math courses, so I decided to take the variances of each of the models and see which was the closest in all their predictions of reality. I did the same with the Vegas marks.
Here are the standard deviations of each type:
As you can see, Vegas was the clear winner here. I'm sure someone else is willing to do more fancy statistical analysis, but that's as far as I go. Nerds.
A large part of this discussion is assuming that Vegas is setting the mark at where it believes it should really be, not the halfway point of total expectation of bettors. But even if it does represent the point where half of the bettors would be one way, and half the other, then it would still show that popular expectation was a better indicator than the WoW methods, at least for this one season.
More fascinating stuff on the 2008-09 season projections:
The teams Vegas was most wrong about (you can look at the table to see how far off they were): The Wizards, Cavaliers, Clippers, Raptors, Nuggets, Pistons, Magic, Kings, Hawks, and Lakers.
The teams the best performing model of Wow (the third one) was most off in predicting actual win total: Nuggets, Cavs, Wiz, Kings, Blazers, Pistons, Hawks, Clippers, Raptors.
Basically everyone was caught off guard about how good the Cavs and Hawks would be, the Nuggets made a big trade, and none of them could foresee the suckitude of the Clippers and Raptors.
The teams Vegas was closest to the mark (all within a half a game, the closest possible): Pacers, Suns, Knicks, Bulls, Rockets.
None of WoW's third method projections were within a half a game, but here's who they predicted the best: Celtics, Bobcats, Spurs, Grizzlies, Rockets. Timberwolves, Bulls.
But of the teams that Wow 3 said were screaming "over" bets (they projected more than 5 games higher than Vegas), they were three of three on the Mavericks, Celtics and Pacers over bets.
On screaming "under" bets (they projected more than 5 games fewer than Vegas), they were only 1-2 on the Nuggets, Thunder, and Blazer under bets.
In conclusion, I have no conclusion! There was more ideas for breakdowns, but I just am pooped now. Take a look at the results, and decide for yourself if you think the system is any better than your best guess. Or don't decide. Just read and enjoy!