[$Ball]Quantifying Next Year's Success(or Failure)

Yes, it seems that one of the flaws would be that you're making a direct correlation between salary and the players output/results which isn't necessarily so, because as DD mentions, some players outperform their contracts while others may be overpaid for what they do for the team, for example: Tmac.

 

I think the Rockets are better suited than most teams to win without their big-money players. We've emphasized the importance of role players that play smart, play hard, and know how to make winning plays. And we've got a really good coaching staff that knows how to maximize the talent they've got. We're probably as well suited as any other team at competing with 40 million inactive.

But it's hard to come up with a figure for how many wins we should get. I don't think there's a simple formula for it, because when teams put together a roster they aren't just taking current needs into account. Sometimes teams may spend a little extra on players they think will pan out for them in the future, while on another team they'll spend extra for players that should win now. Maybe the team decides not to play to win, but rather to develop young players. So depending on where the franchise is at and what their long term plan is, one team with an X payroll may be expected to win 50 games, and another team may be expected to win 30 games.

 

Great points, and to the point in your first paragraph, I don't really think are role players are valued fairly by the amounts of their contracts. I think we have exceptional/special role players that either outperform what they make or are worth every dollar of it, and like you said, they are used very effectively.

I expect this team to be good, because they are.

 

Cool, it's something interesting to think about, though, thanks for the OP.

 

Why don't you go back and read the post that you had originally responded to, which prompted my response? Your methodology clearly assumed a linear relationship, someone pointed out that you were incorrect, and you replied to them with a post that made it clear that you didn't even understand the substance of why that poster needed to correct you. In fact, it's obvious you still don't understand. I don't even understand why you're still posting here. It wouldn't take you that long to go and LEARN a bit about stat so that you could have an intelligent conversation instead of trying to smile and nod so you can save face.

That aside, even if this was set up in the correct way by someone who understood what they were doing, it wouldn't tell us very much. Given that the salary cap results in most payrolls being similar and falling within a narrow range, the predictive value is already pretty limited. Then throw in rookie contracts and albatross contracts and the correlation coefficient becomes very low, especially given the already small sample size. Your "model" can't do anything more accurate than tell us something we already know: the Rockets have a lot of role-players on bargain contracts. Thanks for that earth-shattering breakthrough, chief.

 

here's a thought on how to pretest your formula:

create a par for every team. shouldn't be too tedious since you already know the team salaries and it's a simple division formula.

add up the "pars" and see if they equal the 1260 wins that the nba has every year. if it comes up short, your formula set the bar too low, and vice versa.

 

We'll win 22 games before chrismas.

 

I think this is also much too simplified view of the determining wins/loss.

Mainly Win Shares is determined by one's, Offensive Rating & Defensive Rating which are based on one's stats and one's efficiency in producing those stats. Given the assumption that a star player (in our scenario Ron & Yao) generally makes his teammates better, the other players' overall numbers would increase, but most importantly their efficiency would increase. As such, were one were to replay the year removing those star players from the equation, you cannot only remove their total win share numbers, you would have to create some adjustment to their teammates win shares as well. In all, it causes one to overestimate the end prediction.

In reference to the original post, I like the idea, but I differ on my interpretation of moneyball which is fundamental to your premise.

Moneyball is an attempt to develop a true value of a player which is and their effect on the game beyond what can be determined by use of the current box score. In essence, you are attempting to get a player who value is greater than the 10 pt, 3 asts, 4 rebs, and 1 steal their stats score, since there are a number of players who average those number, but not all of them do so efficiently, within the game, while improving teammates, etc.

There is an assumption within Moneyball that GM's & teams pay for box score values, as in the box score sets the price of the player. So you could pay $5 million for the two players who produce the same box score, but one may have more value not seen in the box score. As many have already pointed out, that assumption breaks down when discussing rookie salaries, vets wanting to win a ring, and new oversea imports.

In essence, I believe you have chosen the wrong independent variable, salary, since is a roughly raw variable in itself dependent on box scores and statistical production. The variable used to make any win-loss prediction would have be the original independent variable developed from the box scores.

How to find that variable that? Well, I figure Morey has been working on that, and all those stats that he and his staff have made are an attempt to make that relationship. And given his recent success, I'd wager he's developed a pretty good independent variable.

 

Wow, you're so clueless that I don't even know where to begin. Here's a hint: your model assigns the same dollar value to each win. If you "readily admit" that the relationship between wins and money might not be linear, you're also readily admitting that you didn't have a clue what you were doing when you tried to make your so-called model.

No, it would completely debunk your conclusions because your whole formula is premised on the idea that you can calculate a dollar value per win, multiply that value by the number of wins, and compare that product to payroll to determine whether or not the team is performing well for the money spent. If, as is ACTUALLY the case, each win is worth a different amount, you need to create a moderately complicated function to compare wins and money spent.

Just quit. Seriously. You're humiliating yourself further by trying to save face. Here's a hint: A team can only put five players on the floor at a time. Production and salary do not scale linearly. A 20/10 player will cost more than twice as much as a 10/5 player. The arguments you're trying to make actually disprove the position you're advocating, and you don't realize this because you don't even understand your own position. You're just trying to make claims contrary to mine because you're upset that you were disproved so brutally.

 

What I am saying is that to truly see if we have succeeded in Moneyball you see if the team outperforms their box scores not their salary. When Morey then determines if a player does exceed their box scores and to a certain extent, he would have to use an entirely different algorithm by which to offer a salary to a player. In essence, the team using Moneyball as a basis to put a salary value on their players will always be an exception to the assumption that the box scores determine the normal salary that another team would offer.


Also, by substituting salary as your independent variable your trying to determine the value of a dependent variable (wins or outperforming whatnot) from another variable (salary) which is dependent on the original independent variable (box scores). Not only this, but your attempting to do this, without first establishing an effective relationship between the original independent variable (box scores) and its primary dependent variable (salary) no matter what that relationship may be, direct/indirect/linear/ bell curve/what have you. That method is going to cause your end result to have wide margins of error (low validity) and hard to reproduce consistently (low reliability) which would, as equation to evaluate success make it invalid.

I felt you were trying to determine if "Moneyball succeeded" and you then hypothesized that succeeding in Moneyball equated to getting more Win's from the salary than salary would alone suggest. The only way to determine if you have succeed under that hypothesis is to make a win-loss prediction, or at least a win-loss constant to measure your success or failure from.