FSRU Rankings - Fantasy Baseball Cafe 2014 Fantasy Baseball Cafe
100% Deposit Bonus for Cafe Members!

Return to Baseball Leftovers

FSRU Rankings

Moderator: Baseball Moderators

Postby GotowarMissAgnes » Wed Jan 21, 2004 8:10 pm

wrveres wrote:Total up a couple of the teams 'runs'. Add up all the players on Baltimore, for example. If it shows more than 900 total runs, you got bad projections. No team should ever total more than a 1000 obvioussly. And I'd gamble you should not "project" any team to score more than 900 runs. Certainly not Baltimore. But go ahead and do it, any team, it doesn't matter. If the total is over 900 . you got bad projections.


Good point! Actually, it makes sense to do this for the entire league, since it can catch problems with your projections not being "normed" to the real world. In the Baseball Prospectus tests you see that for one system if you calculated the average ERA for its entire sample of pitchers, the overall league ERA was projected to be something like 3.50. It's been decades since the overall league ERA as that low, so the projection was normed inappropriately, and all the ERAs should probably be inflated by some constant factor.

That raises another point which I think someone hinted at...if you are using projections, it's important to understand the system. Some projections do not even try to estimate playing time or injury liklihood. That can really give you goofy results, even if the system is good for projecting rate stats like Average or Slugging.
GotowarMissAgnes
Hall of Fame Hero
Hall of Fame Hero

User avatar
Fantasy Expert
Posts: 5516
Joined: 12 Dec 2003
Home Cafe: Baseball
Location: Happy Valley

Postby Erboes » Wed Jan 21, 2004 10:28 pm

GotowarMissAgnes wrote:
wrveres wrote:Total up a couple of the teams 'runs'. Add up all the players on Baltimore, for example. If it shows more than 900 total runs, you got bad projections. No team should ever total more than a 1000 obvioussly. And I'd gamble you should not "project" any team to score more than 900 runs. Certainly not Baltimore. But go ahead and do it, any team, it doesn't matter. If the total is over 900 . you got bad projections.


Good point! Actually, it makes sense to do this for the entire league, since it can catch problems with your projections not being "normed" to the real world. In the Baseball Prospectus tests you see that for one system if you calculated the average ERA for its entire sample of pitchers, the overall league ERA was projected to be something like 3.50. It's been decades since the overall league ERA as that low, so the projection was normed inappropriately, and all the ERAs should probably be inflated by some constant factor.

That raises another point which I think someone hinted at...if you are using projections, it's important to understand the system. Some projections do not even try to estimate playing time or injury liklihood. That can really give you goofy results, even if the system is good for projecting rate stats like Average or Slugging.


You know, Agnes, every projection I've ever seen could barely beat this Marcel, and there lies the problem. The basic assumptions are incorrect, whether you choose to believe it or not.

I'll tell you another thing, there's nothing worse you can do for your projections than trying to balance out the runs scored and pitchers runs given up. So much of a team's ERA comes from scrubs called up from the minors or picked up off the streets due to injury, ineffectiveness, or any other thing that causes lost time. The point is, there's no way you can predict these things except with abnormal situations (like Pedro's yearly jaunt on the DL). KC used 29 different pitchers last season. 29! And most of those extra pitchers caused the team's ERA to over 5.00.

Since there's no way to predict totally who's going to be called up or picked up, you are forced to put about a 5.00 ERA in the original 12 or 13 guys you are pretty sure will start the season in the majors. So you have to give some guys who you'll think will be pretty good next season and give them a 4.50 ERA such as Anderson and May for KC, but what you are doing is tacking on the other 16 or 17 pitchers' miserable stats who were called up onto the Anderson, Mays, and any other KC pitcher who may start the season on the staff.

What you are doing is trying to fit the facts into an incorrect assumption. I have no clue what the numbers are, but I'll bet close to half the pitchers used this season will be called up from the minors at some point after the season starts. Do you think you can name them all? Half? And what do you think they're combined ERA will be? I'll bet you it is 5.00 minimum -- well enough to affect any team's ERA I got to say.

Baseball Notebook does this. I don't know if anyone here has used it before, but they pride themselves in the offense and pitching numbers being equal, and it leads to some ridiculous projections. With my projections even, I was looking at the Red Sox and realized that for me to fit their team under a realistic ERA for them of 3.85 I'd need to raise the Schilling and Pedro's ERA over a half a point, and this was after already raising Lowe's, Kim's, and Wakefield. This is ridiculous. You are just penalizing pitchers for the sins of pitchers who will be called up at some point in the minors.

In short, I can understand how someone reached that 3.50 for a league, although mine doesn't come anywhere close to that. They'll probably have every team scoring 5 runs a game too. Unless you know every player who'll get hurt or lose any time at all and who exactly will be called up to replace them, it is suicide to use this balanced approach. I do think 3.50 ERA to 5 runs a game scored is a bit of a stretch, though. Still, don't be so harsh on them, or at least for that.
Erboes
Major League Manager
Major League Manager

User avatar

Posts: 1433
Joined: 26 Mar 2003
Home Cafe: Baseball

Postby HOOTIE » Wed Jan 21, 2004 10:34 pm

Erboes wrote:I agree that the method you fellas will let you compete in most seasons, but competing is not my goal -- winning in a dominating fashion is.

and Blalock is not that good anyway



He's not? Can the 1st statement be achieved using the 2nd comment?
Smells Like Teen Spirit
HOOTIE
Hall of Fame Hero
Hall of Fame Hero

User avatar
CafeholicResponse TeamFantasy ExpertCafe Ranker
Posts: 15115
(Past Year: 297)
Joined: 12 Jan 2003
Home Cafe: Baseball
Location: Pearl Jam country, right next door to Nirvana, Soundgarden, and Alice in Chains.

Postby Erboes » Wed Jan 21, 2004 11:01 pm

HOOTIE wrote:
Erboes wrote:I agree that the method you fellas will let you compete in most seasons, but competing is not my goal -- winning in a dominating fashion is.

and Blalock is not that good anyway



He's not? Can the 1st statement be achieved using the 2nd comment?


Yes it can, Hootie. Blalock is over priced, and the first rule of fantasy baseball is (or should be) stay away from over priced players. He's being drafted under the assumption that he's going to be a .320-35-120 guy next season and I see nothing that even comes close to suggesting that. If you pick him in the fourth round, a slot where you should still be able to get $25-30 players, and he repeats what he did last season you'll end up losing value since he was only worth $24 last season. He's a bad value pick, Hootie, and those bad values are what keeps teams from the goal stated in the first statement. I'm really surprised you are getting caught in the hype. Even Ron agrees with me on this one (ugh, that's a bad sign).
Erboes
Major League Manager
Major League Manager

User avatar

Posts: 1433
Joined: 26 Mar 2003
Home Cafe: Baseball

Postby GotowarMissAgnes » Wed Jan 21, 2004 11:34 pm

Erboes wrote:I agree that the method you fellas will let you compete in most seasons, but competing is not my goal -- winning in a dominating fashion is.


FWIW, I won my league by almost 30 points last year using my method. No one was within 20 points of me the last 2 months of the season.
GotowarMissAgnes
Hall of Fame Hero
Hall of Fame Hero

User avatar
Fantasy Expert
Posts: 5516
Joined: 12 Dec 2003
Home Cafe: Baseball
Location: Happy Valley

Postby GotowarMissAgnes » Wed Jan 21, 2004 11:48 pm

Erboes wrote:
You know, Agnes, every projection I've ever seen could barely beat this Marcel, and there lies the problem. The basic assumptions are incorrect, whether you choose to believe it or not.

I'll tell you another thing, there's nothing worse you can do for your projections than trying to balance out the runs scored and pitchers runs given up. So much of a team's ERA comes from scrubs called up from the minors or picked up off the streets due to injury, ineffectiveness, or any other thing that causes lost time. The point is, there's no way you can predict these things except with abnormal situations (like Pedro's yearly jaunt on the DL). KC used 29 different pitchers last season. 29! And most of those extra pitchers caused the team's ERA to over 5.00.

Since there's no way to predict totally who's going to be called up or picked up, you are forced to put about a 5.00 ERA in the original 12 or 13 guys you are pretty sure will start the season in the majors. So you have to give some guys who you'll think will be pretty good next season and give them a 4.50 ERA such as Anderson and May for KC, but what you are doing is tacking on the other 16 or 17 pitchers' miserable stats who were called up onto the Anderson, Mays, and any other KC pitcher who may start the season on the staff.

What you are doing is trying to fit the facts into an incorrect assumption. I have no clue what the numbers are, but I'll bet close to half the pitchers used this season will be called up from the minors at some point after the season starts. Do you think you can name them all? Half? And what do you think they're combined ERA will be? I'll bet you it is 5.00 minimum -- well enough to affect any team's ERA I got to say.

Baseball Notebook does this. I don't know if anyone here has used it before, but they pride themselves in the offense and pitching numbers being equal, and it leads to some ridiculous projections. With my projections even, I was looking at the Red Sox and realized that for me to fit their team under a realistic ERA for them of 3.85 I'd need to raise the Schilling and Pedro's ERA over a half a point, and this was after already raising Lowe's, Kim's, and Wakefield. This is ridiculous. You are just penalizing pitchers for the sins of pitchers who will be called up at some point in the minors.

In short, I can understand how someone reached that 3.50 for a league, although mine doesn't come anywhere close to that. They'll probably have every team scoring 5 runs a game too. Unless you know every player who'll get hurt or lose any time at all and who exactly will be called up to replace them, it is suicide to use this balanced approach. I do think 3.50 ERA to 5 runs a game scored is a bit of a stretch, though. Still, don't be so harsh on them, or at least for that.


You assume, Erboes, that the basic assumptions are incorrect in other methods. It's possible that there is another explanation. That explanation would be that the difference between Marcel and what can be realistically achieved with projections is very small, because the rest of the variation in performance is simply random.

Second, you need to read what people write, Erboes. I'm not talking about "balancing out" runs scored. I'm talking about doing some basic face validity checks of your projection system. If your projection system projects that the overal league batting average is going to be .205, then you need to make some adjustments. While that sounds obvious, the recent test of systems on Baseball Prospectus clearly showed that some of these systems don't even pass that basic face validity test.

And FWIW, systems like DMB and Pecota do project the performance of players likely to get called up. Many of these systems are projecting the performance of 1,000-1,500 guys. DMB runs their simulation over and over again (I think 50 times is typical) to get an idea of how likely injuries and call-ups are. They then project playing time based on that. Systems like this build in basci validity tests, like league runs allowed have to equal runs scored.

And FWIW, last year 611 pitchers pitched last year. Assuming 12 pitchers per roster on opening day, that means 41% of pitchers were in the minors/on the DL, etc to start the season. They really don't impact ERAs much, because they pitch so few innings. It's the guys that are on the roster Opening Day who pitch 100 innings and give up 6 runs that crush those ERAs!
GotowarMissAgnes
Hall of Fame Hero
Hall of Fame Hero

User avatar
Fantasy Expert
Posts: 5516
Joined: 12 Dec 2003
Home Cafe: Baseball
Location: Happy Valley

Postby HOOTIE » Thu Jan 22, 2004 12:21 am

Erboes wrote:
HOOTIE wrote:
Erboes wrote:I agree that the method you fellas will let you compete in most seasons, but competing is not my goal -- winning in a dominating fashion is.

and Blalock is not that good anyway



He's not? Can the 1st statement be achieved using the 2nd comment?


Yes it can, Hootie. Blalock is over priced, and the first rule of fantasy baseball is (or should be) stay away from over priced players. He's being drafted under the assumption that he's going to be a .320-35-120 guy next season and I see nothing that even comes close to suggesting that. If you pick him in the fourth round, a slot where you should still be able to get $25-30 players, and he repeats what he did last season you'll end up losing value since he was only worth $24 last season. He's a bad value pick, Hootie, and those bad values are what keeps teams from the goal stated in the first statement. I'm really surprised you are getting caught in the hype. Even Ron agrees with me on this one (ugh, that's a bad sign).



You stated, he was not that good anyway, not what price he might go at, or what round he might go in. I doubt anyone expects those numbers you posted for this year. I don't care about hype. I care that a guy who at age 22, handled himself quite well. Guys at age 22 are in AA, not belting 29 hrs and hitting 300. I care that at age 20 in AA, he creamed the ball. Anybody can be overpriced. I see Arod in the Cafe priced at $50 quite often. Considering he's never had a $50 season, you get the point. Blalock had 6.54 rc/27. Chavez best year was 6.52 in 01 at age 23. Blalock will be a top 3 3b very soon.
Smells Like Teen Spirit
HOOTIE
Hall of Fame Hero
Hall of Fame Hero

User avatar
CafeholicResponse TeamFantasy ExpertCafe Ranker
Posts: 15115
(Past Year: 297)
Joined: 12 Jan 2003
Home Cafe: Baseball
Location: Pearl Jam country, right next door to Nirvana, Soundgarden, and Alice in Chains.

Postby ramble2 » Thu Jan 22, 2004 12:50 am

Great thread folks! Very interesting stuff.

GTWMA, thanks for the tip on Marcel. If that monkey's hitting .68, that's plenty good.
"The game has a cleanness. If you do a good job, the numbers say so. You don't have to ask anyone or play politics. You don't have to wait for the reviews." - Sandy Koufax
ramble2
General Manager
General Manager

User avatar
CafeholicFantasy ExpertCafe Ranker
Posts: 2952
(Past Year: 5)
Joined: 27 Feb 2003
Home Cafe: Baseball
Location: Salt Lake City

Postby Cornbread Maxwell » Thu Jan 22, 2004 3:33 am

GotowarMissAgnes wrote:Even more problematic for the roto projection systems (especially for hitters) is the information in the Baseball Primer thread I posted.

The correlation between your prediction and what really happened is one measure of how well your system works. A perfect correlation would be 1.0. With random error, you can never achieve a perfect correlation. So the "realistic" goal for correlation is less than 1.0

In the Primer thread they show two things:

A. A really naive formula achieves a correlation of .68
B. Given what we know about random error in baseball statistics, it is likely that the maximum correlation of the best prediction system (for hitters) would be .72

Now, while on Wall Street the difference between a correlation of .72 and .68 might earn you a few million or more, in roto ball it seems a little silly to try to invest a whole lot of time and energy on that (unless, like Erboes, you can provide Rotobabes as well!!)


Ah, I completely agree - and this is where value drafting (buying) comes in. The object is not to try and win with your entire predictions - thats the diversified portion. The goal is to identify and go after those specific players (stocks) that you feel have the strongest return for where they are priced. These are the "sleepers", and the value they generate comes from a misunderstanding of their value by the general masses. You are not going to win your league by taking Arod or Prior this yr. The winner will be the person who drafted wisely in the 10th-15th rds and ended up with some players who should have gone much earlier. This is the way it always has been - much the same on Wall Street.
Image
Cornbread Maxwell
Hall of Fame Hero
Hall of Fame Hero

User avatar
CafeholicFantasy ExpertPick 3 ChampionSweet 16 Survivor
Posts: 5694
Joined: 7 Jul 2003
Home Cafe: Football

Postby Erboes » Thu Jan 22, 2004 10:26 am

GotowarMissAgnes wrote:
Erboes wrote:
You assume, Erboes, that the basic assumptions are incorrect in other methods. It's possible that there is another explanation. That explanation would be that the difference between Marcel and what can be realistically achieved with projections is very small, because the rest of the variation in performance is simply random.

Second, you need to read what people write, Erboes. I'm not talking about "balancing out" runs scored. I'm talking about doing some basic face validity checks of your projection system. If your projection system projects that the overal league batting average is going to be .205, then you need to make some adjustments. While that sounds obvious, the recent test of systems on Baseball Prospectus clearly showed that some of these systems don't even pass that basic face validity test.

And FWIW, systems like DMB and Pecota do project the performance of players likely to get called up. Many of these systems are projecting the performance of 1,000-1,500 guys. DMB runs their simulation over and over again (I think 50 times is typical) to get an idea of how likely injuries and call-ups are. They then project playing time based on that. Systems like this build in basci validity tests, like league runs allowed have to equal runs scored.

And FWIW, last year 611 pitchers pitched last year. Assuming 12 pitchers per roster on opening day, that means 41% of pitchers were in the minors/on the DL, etc to start the season. They really don't impact ERAs much, because they pitch so few innings. It's the guys that are on the roster Opening Day who pitch 100 innings and give up 6 runs that crush those ERAs!


Agnes, you have well thought out arguments and I respect your opinions, but I have to disagree with you on some of them.

Every projection set I've ever seen is plain awful, be it in magazines or over the net. Most of what they are right about is explained by plain luck. When you're dealing with close to 1000 players you are bound to be right on some of them. I believe there's a better way, and I'm giving anyone a free chance to prove whether I am right or an idiot. Methods can only be proven or disproven by putting them to the test, so I don't think discussing them any further will be of much use.

"Validity checks" sounds reasonable. I've done the same myself and I'm satisfied with the results. The pitching and hitting is not equal, though. I don't think there's a computer in the world who can predict who can get hurt or how many pitchers a team will use like the Royals. 29 is a ton of pitchers,and I'm not even going to attempt to figure out how many they'll use next season. Even the Yankees used 23, many of whom were picked up in trades. Does this Pecota predict that? Can Pecota predict which 12 pitchers are going to make the opening day roster for Detroit out of a few dozen?

This is the first system I've ever heard of that attempts such an undertaking. Even HQ with all its arrogance doesn't claim to know who's going to get hurt and who'll be called up to replace them. I will look more closely at this system.

My projections, as they sit now, are incomplete because of all the FA's still out there and the final 25 man rosters not being set until right before opening day. Even then, most teams will be short on innings pitched and at bats because of the inevitable roster changes throughout the season. When new players arrive from the minors then I shall adjust those teams' projections to reflect that. If I were to guess, 20% of all innings pitched for a team on average is thrown by pitchers who are not on the opening day roster. I think this is significant. I am not about to attempt to look into the future to see which pitchers will be hurt and which will be called up to replace them beyond the obvious. Even if I were capable of this the benefits would be minimal at best and at worse could skew the projections.
Erboes
Major League Manager
Major League Manager

User avatar

Posts: 1433
Joined: 26 Mar 2003
Home Cafe: Baseball

PreviousNext

Return to Baseball Leftovers

Who is online

Users browsing this forum: No registered users and 9 guests

Forums Articles & Tips Sleepers Rankings Leagues


Today's Games
Saturday, Oct. 25
(All times are EST, weather icons show forecast for game time)

Kansas City at San Francisco
(8:07 pm)

  • Fantasy Baseball
  • Article Submissions
  • Privacy Statement
  • Site Survey 
  • Contact