Win Projection

[colbyjack]

I have a question for the stat-minded individuals, or anyone that can answer the question. Win projection is usually used moving and looking forward, trying to determine what a team may win with what players they have and/or may obtain. I see a lot of arguments that say something to the effect of "getting player X would only lead to a difference of Y wins." Has anyone ever run this exercise based on history and what has actually happened? Specifically, can you look at the individual statistics from the 2009 Reds, a 78-84 team (individual by individual), and accurately point to why they project to finish this year 92-70? Does the difference in stats equal the 14 win differential, or are there some collaborative results, including luck and/or defensive metrics, for reason as to why the team is playing significantly better as a whole?

I ask that looking forward to 2011, wondering if the Brewers really need certain acquisitions, and improvement from their own players, to exactly equal the win differential based on an improvement of statistics to become a playoff team? Is a complete overhaul needed to reach that goal, or is it realistic with the players they have right now (with some key additions and improvements from others of course)?

Thanks in advance.

[thebruce44]

I don't have time to type out a more complete post, but this article might point you in the right direction:

 

WAR: It Works

http://www.fangraphs.com/blogs/index.php/war-it-works/

[sveumrules]

Votto +2, Rolen +3.5, Bruce +2, Phillips +1, Hernandez +2, Cueto +1.5, Wood +2, Leake +1

 

That's 15 "wins" worth of WAR improvement from 2009 to 2010 from 8 players. Obviously some players (Dickerson, Harang) contributed less this year than last so that eats into the gains a little bit.

 

The other big factor at play here, LUCK, isn't quite so easy to quantify unfortunately.

[colbyjack]

Thanks thebruce for the link, and to sveumrules for the Reds breakdown. From the link provided:

 

One standard deviation of the difference between WAR and actual record is 6.4 wins, and every single team is within two standard deviations. Only four teams were more than 10 wins away from their projected total by WAR, with Tampa Bay ending up the furthest away from our expectation (96.6 projected wins, 84 actual wins), and 18 of the 30 teams were within six wins of their projected WAR total.

 

After reading this, along with several of the responses, I realized the projection isn't 100% accurate, not like any of them are, so there is some wiggle room to the overall vs. individual WAR projections. And again, I wasn't talking about projections as much as what actually happened. 12 of the 30 teams were off by more than 6 wins, with an average of 6.4 as noted, so obviously there is some significant give/take when it comes to the team's success in both directions. Basically, if the goal for next year's team is 92 wins, that doesn't necessarily mean that all of the individual players' WARs need to add up to 92. The closer the better obviously.

[rluzinski]

The minimum "wiggle room" any team has over the course of a season is huge. Even if god tells us the odds a team will have of winning each of their 162 games in a season, the absolute BEST we can do is say something to the effect of:

 

"Team A will finish with X wins, plus or minus 6 wins, 68% of the time."

 

Where did I come up with that? That's 1 standard deviation of 162 coin flips. That's all a season really is, afterall. Now, the odds of winning each game is going to vary but that is still a very good approximation. So you can take a team that projects to win 83 wins and 16% of the time ((100%-68%)/2), they are going to win 90 or more games (over 83+6).

 

In reality, there's a lot of other uncertainty (estimates of the true talent of each player is the big one), which gives a team that's projected to win 83 games even a better chance of winning 90+ games. The flip side is also true of course, in that they could substantially underperform as well.

 

Sports fans absolutely HATE the concept of luck but it's inescapable. You can't wish it away.

[colbyjack]

That helps a lot, thanks. It's my understanding based on your and others responses that the individual player WAR is pretty accurate at determining their individual worth, but an overall team's success isn't as easily determined simply by adding up all the WARs. I personally love the concept of luck, both good and bad, as it also provides hope.

[mothershipconnection]

Sports fans absolutely HATE the concept of luck but it's inescapable. You can't wish it away.

 

Hypothetically, you can imagine a reverse Nike Swoosh shaped curve plotting willingness to recognize luck along the continuum from the very gullible to the very skeptical. The gullible recognize it as magic while the scientific understand that outside of math, there's pretty much always going to be bouncing throughout a predicted range of effects of an IV upon a DV rather than a textbook linear relationship with all data accounted for by a straight line. Of course, luck is not an attribute or quality that an organism possesses, it's just several consecutive data points in the "good" direction away from the straight linear projection. 4

[LouisEly]

The are three big factors that statistics just can't quantify - motivation, psychology, and chemistry. And they really do have an impact on the performance of players.

 

Maybe someone is distracted by the birth of a first child, or the health/death of a family member, or financial problems. Maybe some is really motivated by being in a contract year, or trying to prove someone wrong who traded them or released them. Maybe someone finally finds a coach they "click" with or relate with. Maybe there are personalities in the locker room that just don't get along. There are many, many human elements that play a role in performance.

[rluzinski]

"There are many, many human elements that play a role in performance."

 

I think you won't find much argument with regard to whether it plays a role. I and others just feel it's generally a very minor role, compared to talent and luck. And because it's it's typically minor, you will never find any real statistical evidence of it.