D-Lee MVP Odds

[treebird]

Guess Pujols won the MVP battle for today.

 

:(

 

he won it a long time ago, when the cubs decided it would be fun to have a horrible team.

[TigerRules]

Dusty gave Lee a hell of a chance to win the thing with his early season top of the line up configuration. :?

 

How many times did Derrick bat with nobody on base this year?

[1wizard]

Dusty gave Lee a hell of a chance to win the thing with his early season top of the line up configuration.

 

How many times did Derrick bat with nobody on base this year?

 

I cannot answer your question, but I saw a stat that had the number of RBI chances for D. Lee, Pujols, and AJ. Jones had far more opportunities than both Lee and Albert. Albert only had about 6 or 8 more ABs with RISP than D. Lee. I have been unable to find it.

 

Does anyone have this stat?

[1wizard]

Jones is hitting .225 with RISP. That is hardly the type of performance I'd expect from someone "carrying his team". He's had 160 ABs with RISP to Lee's 112 and AP's 118.

 

This is a post from the Post Dispatch. Many poster on that site are irritated by ESPN's handling of the MVP. Most of there talking heads (ESPN) are saying that AJ is the leagues MVP. Looking at the numbers, I have a hard time believing that anyone besides Albert or Lee get the award. Jones has a very low batting average and average with RISP, yet he may benefit from the exposure of a closer race in his division.

[K-Town]

Win Shares are now dead even for Pujols/Lee, and VORP is extremely close (Lee with a slight edge over Pujols). If Lee wins the MVP, it won't be a crime, but I think you have to favor the guy that's leading his team to the playoffs (Pujols). Lee is deserving, but Andruw Jones shouldn't even be in the conversation.

[JC]

Blutarsky's GPA.

[JGalt73]

BP's Joe Sheehan weighs in (registration required):

 

WARP (wins above replacement, includes defense):

 

WARP

Derrek Lee 11.2

Albert Pujols 9.6

Roger Clemens 8.9

Jim Edmonds 8.9

Morgan Ensberg 8.8

Dontrelle Willis 8.8

Jason Bay 8.7

Miguel Cabrera 8.3

Marcus Giles 8.3

Todd Helton 8.1

Andruw Jones 8.1

 

"Derrek Lee has been the most valuable player in the National League... It's not going to be, though.

 

...voters may be inclined to treat this ballot as something of a lifetime achievement award for Pujols, who has been eating Bonds' dust for four remarkable seasons.

 

Pujols getting the MVP Award would be wrong, but not silly, as he's clearly the second-best player in the league and at least within shouting distance of #1."

[CardsFanInChiTown]

BP's Joe Sheehan weighs in (registration required):

 

WARP (wins above replacement, includes defense):

 

WARP

Derrek Lee 11.2

Albert Pujols 9.6

Roger Clemens 8.9

Jim Edmonds 8.9

Morgan Ensberg 8.8

Dontrelle Willis 8.8

Jason Bay 8.7

Miguel Cabrera 8.3

Marcus Giles 8.3

Todd Helton 8.1

Andruw Jones 8.1

 

"Derrek Lee has been the most valuable player in the National League... It's not going to be, though.

 

...voters may be inclined to treat this ballot as something of a lifetime achievement award for Pujols, who has been eating Bonds' dust for four remarkable seasons.

 

Pujols getting the MVP Award would be wrong, but not silly, as he's clearly the second-best player in the league and at least within shouting distance of #1."

 

 

AP leads in win shares now:

 

AP: 34

DLee: 33

 

http://www.hardballtimes.com/winshares/index.php

 

 

Lots of stats can be used. For this argument I'm going to say win shares is a better stat. :wink:

[AlwaysaCub]

You wanna know why Cards fan? Because win shares are adjusted for team wins, and the cards have overachieved their pythagorean record, so the adjustment between runs and wins works in his favor. Meanwhile, DLee's team is underachieving relative to the pythagorean record, which win shares work off of. Thus, the system has created a mirage based on how the team is compared to the runs ratio. At least I think that's true, correct me if I'm wrong.

[vaball]

You wanna know why Cards fan? Because win shares are adjusted for team wins, and the cards have overachieved their pythagorean record, so the adjustment between runs and wins works in his favor. Meanwhile, DLee's team is underachieving relative to the pythagorean record, which win shares work off of. Thus, the system has created a mirage based on how the team is compared to the runs ratio. At least I think that's true, correct me if I'm wrong.

 

So if win shares are predicated on pythagorean record, how do we know that the pythagorean projections are correct? I mean no stat is perfect right? There are always some variables at work that any given stat can't take into account.

 

This isn't me just being argumentative, I just wonder sometimes about some of these sabermetrics.

[CardsFanInChiTown]

You wanna know why Cards fan? Because win shares are adjusted for team wins, and the cards have overachieved their pythagorean record, so the adjustment between runs and wins works in his favor. Meanwhile, DLee's team is underachieving relative to the pythagorean record, which win shares work off of. Thus, the system has created a mirage based on how the team is compared to the runs ratio. At least I think that's true, correct me if I'm wrong.

 

So if win shares are predicated on pythagorean record, how do we know that the pythagorean projections are correct? I mean no stat is perfect right? There are always some variables at work that any given stat can't take into account.

 

This isn't me just being argumentative, I just wonder sometimes about some of these sabermetrics.

 

and for that matter why didn't Bill James use the pythagorean record?

[ConstableRabbit]

You wanna know why Cards fan? Because win shares are adjusted for team wins, and the cards have overachieved their pythagorean record, so the adjustment between runs and wins works in his favor. Meanwhile, DLee's team is underachieving relative to the pythagorean record, which win shares work off of. Thus, the system has created a mirage based on how the team is compared to the runs ratio. At least I think that's true, correct me if I'm wrong.

 

So if win shares are predicated on pythagorean record, how do we know that the pythagorean projections are correct? I mean no stat is perfect right? There are always some variables at work that any given stat can't take into account.

 

This isn't me just being argumentative, I just wonder sometimes about some of these sabermetrics.

 

and for that matter why didn't Bill James use the pythagorean record?

 

You know, I DID mathematically state and prove this 2 pages ago or so....

 

"You know, win shares are great most of the time, but many use them as a failproof way of determining player value, when, in fact, the stat does have a large flaw which most overlook:

 

Players who play for teams that win more games than expected, (found by using the Pythagorean expectation), will receive more win shares than players whose team wins fewer games than expected. Beacuse a team going over or falling short of its Pythagorean expectation comes to such a conclusion by chance (according to Bill James anyway), you cant give out credit based on wins. (and no, I'm not just making this up, you can read into it)

 

Sure enough, when calculating the Pythagorean expectation for Stl, the winning % is .629, lower than their actual % of .634. They are winning more (albeit a bit more) than expected due to chance.

 

but that's not all...

 

The Cubs have a much greater difference from expectation to reality, and of course, they've got the short end of the stick.

 

Pythagorean expectation for the Cubs: .491

Actual winning % for the Cubs : .474

The Cubs are losing more than expected, due to chance, and are doing so by a much larger margin than the Cardinals are winning due to chance.

 

So...

Pujols is getting more win shares than he deserves, and Lee is getting less than he deserves, due to chance, and not value.

I wouldn't have mentioned it if everyone wasn't making such a big deal over winshares and how the two were tied. Winshares are usually great, but in this case, they're flawed, and in reality, Lee has the advantage. "

-ConstableRabbit, Sept 2 2005

[vaball]

Constable I'm not disputing your analysis of win shares. I'm not really disputing anything. I'm asking about Pythagorean projections.

 

How do they work? Until I understand them better I'm kind of dubious about them.

[ConstableRabbit]

Constable I'm not disputing your analysis of win shares. I'm not really disputing anything. I'm asking about Pythagorean projections.

 

How do they work? Until I understand them better I'm kind of dubious about them.

 

Haha, I know you're not.

 

Ok, it's like this:

 

A winshare is a third of a win, so if a team wins 80 games, there are 240 to go around. Obviously, the point of the stat is to find out how many runs a certain player is directly responsible for.

Now,

I dont know if youve read Moneyball or not, but anyone in baseball knows that there is an element of luck involved in the game. The thing is, math proves that some teams and players are luckier than others, which is proven by the Pythagorean projection (or expectation), the formula of which is as follows:

 

Winning pct= runs scored squared/[(runs scored squared) + (runs allowed squared)]

 

Now, for whatever reason, this formula was the one which was derived and it works most of the time. Here's the deal, sometimes it doesnt work out..

 

When a team wins more than the projection, they are said to have won those games due to luck. Likewise, when a team loses more than the projection, they are said to have lost those games due to luck.

 

In the case study at hand, the Cardinals have won more than their projection (albeit by just a little bit) and the Cubs have lost more than their projection (by a larger margin than the Cards won due to luck)

 

This means that the Cardinals have more win shares to go around and the Cubs have less to go around, so naturally, Pujols will have an unfair winshare advantage to Lee, at no fault to either player, giving Lee the overall advantage.

 

I hope I explained that well...

 

:wink:

[ConstableRabbit]

Oh yeah, one more thing. The difference between the projected and actual winning percentages aren't 100% luck... a quality bullpen can also sway the verdict for or against a team (which would make a lot of sense in this discussion, haha)

 

Also, I crunched those numbers on the 2nd, and while they're probably not much different, they have probably changed a bit since then.

[vaball]

So basically the more wins a team has the more potential win shares for any given player. That seems pretty straight forward. Pujols may end up with more total win shares because his team has more wins. Is that right.

 

I don't get Pythagorean projections though. Doesn't that formula only take into account total runs scored? How does it take into account the distribution of a team's runs scored? A team might score 14 runs one day and win, then 1 the next day and lose. It just seems like an overly simplistic formula to me, but then I'm not a statistician.

 

Is there any data on how accurate the pythagorean percentages are? If the formula is good shouldn't the actual percentages conform to the predicted ones more often than not? At what threshold do we say that in fact the formula is inadequate, not that a given team is getting lucky?

[ConstableRabbit]

So basically the more wins a team has the more potential win shares for any given player. That seems pretty straight forward. Pujols may end up with more total win shares because his team has more wins. Is that right.

 

I don't get Pythagorean projections though. Doesn't that formula only take into account total runs scored? How does it take into account the distribution of a team's runs scored? A team might score 14 runs one day and win, then 1 the next day and lose. It just seems like an overly simplistic formula to me, but then I'm not a statistician.

 

Is there any data on how accurate the pythagorean percentages are? If the formula is good shouldn't the actual percentages conform to the predicted ones more often than not? At what threshold do we say that in fact the formula is inadequate, not that a given team is getting lucky?

 

Yes, if there are 2 players with identical stats on 2 different teams, the player on the team with more wins gets more winshares, and because Pujols' and Lee are so close statistically, any additional wins out of the pythagorean expectation for him and any less than expected for Lee sway the winshare advantage to Pujols. Even though they are neck and neck, Lee should have more and Pujols should have less.

 

I'm not a statistician either, the thing is, a math (and baseball) geek, if not genius, Bill James, sat down and found that that formula is ridiculously close to what has happened to every team in baseball history. In my 5 minutes of trying to find out how close it actually is to perfect, I did find that there have been efforts to get as close as possible, with something called "the pythagenport formula" (invented by Clay Davenport), which is:

1.5log[(runs+runs allowed)/games]+.45

 

Either way, it's a Bill James creation that was the result of trial and error with the history of baseball. There's no specific reason as to why it works, but it has and does. I'm not sure of the actual numbers pertaining to how accurate it is (anyone want to volunteer to crunch all of those numbers? :shock: ), but like you suggested (I think), more often than not, it's very close.

With regard to your 14 run win, 1 run loss argument, runs allowed are equated, and like I said, a good bullpen will screw with the numbers.

[ConstableRabbit]

I'm getting my wisdom teeth out at 9:30 and as of now, winshares and the pythagorean expectation will be the last things on my mind before they turn the gas on... :D

[vaball]

I'm getting my wisdom teeth out at 9:30 and as of now, winshares and the pythagorean expectation will be the last things on my mind before they turn the gas on... :D

 

Well good luck with that. You'll likely spend the next couple of days lying in bed on pain killers. It'll give you plenty of time to crunch those numbers and report back on the accuracy of pythagorean projections. :wink:

[sonofsamiam]

I've read the Win Shares book, and I'm pretty darn sure that a team's wins, in the end, has very little total effect on a player's Win Shares. At the most, it might knock down 1 point or add 1 point, which are pretty minimal. (James states that the difference of 2 or 3 shares isn't much difference at all.)

 

The reason Win Shares work is you can work out each player's numbers using the method first (using the "simple" method that James explains), and then extrapolate them to the team. A little rounding up or down is usually needed, but almost always the total Win Shares before adjusting for a team's wins match up pretty closely with the actuals. That's why it's such a cool stat.

 

I looked it up, and yes, Pythagorean records can have some bearing, but I doubt it would affect a player's Win Shares by more than 1. A great record over Pythagorean still has to be spread around all the players on that team, and vice versa.

[CardsFanInChiTown]

You wanna know why Cards fan? Because win shares are adjusted for team wins, and the cards have overachieved their pythagorean record, so the adjustment between runs and wins works in his favor. Meanwhile, DLee's team is underachieving relative to the pythagorean record, which win shares work off of. Thus, the system has created a mirage based on how the team is compared to the runs ratio. At least I think that's true, correct me if I'm wrong.

 

So if win shares are predicated on pythagorean record, how do we know that the pythagorean projections are correct? I mean no stat is perfect right? There are always some variables at work that any given stat can't take into account.

 

This isn't me just being argumentative, I just wonder sometimes about some of these sabermetrics.

 

and for that matter why didn't Bill James use the pythagorean record?

 

You know, I DID mathematically state and prove this 2 pages ago or so....

 

 

I know, I was asking the why not the how. Why didn't he use the Pythagorean expectation rather then real wins. Just wondering.

[cubbieinexile]

Um. . . maybe I missed it but where does Bill say Win Shares is based off of pyth expectations? Win Shares is based on actual wins. If you have 80 wins then you have 240 win shares. The division between hitting and defense is fixed so the pyth isn't used their either. The division of win shares for individual players is based on runs created, but the runs created is tied to actual runs so it isn't used their either.

[cubbieinexile]

Just because a mathematical formula tells you one thing does it mean that the Cards have more wins to go around by luck. It simply means that the mathematical formula does not adequately model the way the team played. IT wasn't luck it wasn't random it just means the model isn't accurate for this team. R^2/R^2+RA^2 is simple formula and it will get you close to actual wins but it is not and exact and precise formula.

[sonofsamiam]

Um. . . maybe I missed it but where does Bill say Win Shares is based off of pyth expectations? Win Shares is based on actual wins. If you have 80 wins then you have 240 win shares. The division between hitting and defense is fixed so the pyth isn't used their either. The division of win shares for individual players is based on runs created, but the runs created is tied to actual runs so it isn't used their either.

 

This link addresses the Pythagorean question:

 

http://www.hardballtimes.com/main/article/2004-win-shares-have-arrived

 

See the 6th bullet point.

 

I still feel this discrepancy hardly makes a large difference in a player's eventual Win Shares.

[cubbieinexile]

Okay then explain why that is so. Are they saying that if we based win shares on pyth he would get more or less.

 

I just don't see it. Win Shares has nothing to do with pythagorean expectations.

 

On the same site they explain how they calculate Win Shares and they make no mention of pyth expectations. I have the Win Shares book and I have never read anything about pythagorean and how it is used to calculate win shares

 

 

 

Basically what I think the author is saying is that if we went by pyth wins and losses (which to statheads they feel is more important) then a player would get more or less win shares depending on the difference. But that is using Win Shares incorrectly. Win Shares is trying to explain how a team won its games not the supposed value of events in baseball. Metrics like VORP and WARP do that