PECOTA and FF (1 Viewer)

[Bri]

From Newsweek

http://blog.newsweek.com/blogs/stumper/arc...is-pitches.aspx

On May 6, expectations were high for Hillary Clinton. After all, the latest polls suggested the former First Lady had built up a 5-point cushion in Indiana and slashed Barack Obama's 20-point lead in North Carolina to 8. But over at FiveThirty Eight.com, an anonymous blogger (nom d'écran: "Poblano") wasn't convinced. Relying on demographic data from previous primaries and ignoring the usual mishmash of polls, the mysterious upstart projected that Clinton would win Indiana by 2 percent and lose North Carolina by 17—a far-less favorable outcome. When the results finally rolled in—1 in Indiana, 15 in North Carolina—Poblano had outperformed every established pollster. Clinton never recovered, but with the National Journal, the Guardian and the New York Post suddenly dissecting or demanding the secrets of his success, Poblano became an Internet sensation. "It was kind of amazing," he says.

It only gets better. For the man behind the blog, outpredicting the experts wasn't anything new—even if outpredicting political experts was. On May 30, Poblano finally revealed his offline name: Nate Silver. Doesn't ring a bell? Chances are you're not a baseball geek. Silver, 30, is already celebrated among ball fans for inventing something called PECOTA. Developed while the University of Chicago econ alum slogged through a post-collegiate consulting gig—"I'm used to not sleeping," he tells NEWSWEEK—PECOTA is now recognized as the most accurate system for forecasting how athletes and teams will perform in the future (down to the number of singles). In 2007, Silver's algorithm enraged at least half of Chicago when it said the White Sox—2005 champs—would post a 72–90 record. Turned out PECOTA was exactly right. For laypeople, the leap from the national pastime to national politics might seem like a stretch. But not for Silver (who posted his first political item on Daily Kos in October). "Baseball and politics are data-driven," he's written. "But a lot of the time, that data might be used badly. In baseball, that may mean looking at a statistic like batting average when things like on-base percentage and slugging percentage are far more correlated with winning ballgames. In politics, that might mean cherry-picking a certain polling result." In other words, different sport—same skill set.

From the start, Silver took pride in myth-busting the MSM, which has tended to reduce 2008's complex calculus—delegate distribution, demographic coalitions—into not-quite-true narratives. Obama has a problem with working-class whites? Actually, he has a problem with Appalachian working-class whites—and not their cousins in Oregon and Wisconsin. And so on. The response was ecstatic, and FiveThirtyEight's daily traffic increased 5,000 percent between March and June. But the main attraction was always Silver's primary predictions. Taking a page from PECOTA—a comprehensive historical database, it projects future performance by matching current players to comparable predecessors—Poblano predicted the results in, say, Pittsburgh by measuring how Clinton and Obama did in demographically similar congressional districts earlier on (once set, their coalitions were remarkably stable). Silver's score wasn't perfect—he underestimated Clinton in Kentucky and South Dakota. But ultimately, he came within 20 delegates of the final split on Super Tuesday (out of nearly 1,700) and 2.5 percent, on average, in the other six post-March primaries. "Nate's work is innovative," says Mark Blumenthal of Pollster.com.

Just a snip From Wikipedia http://en.wikipedia.org/wiki/PECOTA

PECOTA relies on fitting a given player's past performance statistics to the performance of "comparable" Major League ballplayers by means of similarity scores. As is described in the Baseball Prospectus website's glossary:[5]

PECOTA compares each player against a database of roughly 20,000 major league batter seasons since World War II. In addition, it also draws upon a database of roughly 15,000 translated minor league seasons (1997-2006) for players that spent most of their previous season in the minor leagues. . . . PECOTA considers four broad categories of attributes in determining a hitter's comparability:

1. Production metrics – such as batting average, isolated power, and unintentional walk rate for hitters, or strikeout rate and groundball rate for pitchers.

2. Usage metrics, including career length and plate appearances or innings pitched.

3. Phenotypic attributes, including handedness, height, weight, career length (for major leaguers), and minor league level (for prospects).

4. Fielding Position (for hitters) or starting/relief role (for pitchers). . . . In most cases, the database is large enough to provide a meaningfully large set of appropriate comparables. When it isn't, the program is designed to 'cheat' by expanding its tolerance for dissimilar players until a reasonable sample size is reached.

PECOTA uses nearest neighbor analysis to match the individual player with a set of other players who are most similar to him. Although drawing on the underlying concept of Bill James' similarity scores, PECOTA calculates these scores in a distinct way that leads to a very different set of "comparables" than James' method.[6] Furthermore, Silver describes the following distinct feature:

The PECOTA similarity scores are based primarily on looking at a three-year window of a pitcher’s performance. Thus, we might look at what a pitcher did from ages 35-37, and compare that against the most similar age 35-37 performances, after adjusting for parks, league effects, and a whole host of other things. This is different from the similarity scores you might see at baseball-reference.com or in other places, which attempt to evaluate the totality of a player’s career up to a given age."[7]

Once a set of "comparables" is determined for each player, his future performance forecast is based on the historical performance of his "comparables." For example, a 26 year-old's forecast performance in the coming season will be based on how the most comparable Major League 26 year-olds performed in their subsequent season.

Has anyone ever tried to take this PECOTA system to FF?

 

[gheemony]

From Newsweek

http://blog.newsweek.com/blogs/stumper/arc...is-pitches.aspx

On May 6, expectations were high for Hillary Clinton. After all, the latest polls suggested the former First Lady had built up a 5-point cushion in Indiana and slashed Barack Obama's 20-point lead in North Carolina to 8. But over at FiveThirty Eight.com, an anonymous blogger (nom d'écran: "Poblano") wasn't convinced. Relying on demographic data from previous primaries and ignoring the usual mishmash of polls, the mysterious upstart projected that Clinton would win Indiana by 2 percent and lose North Carolina by 17—a far-less favorable outcome. When the results finally rolled in—1 in Indiana, 15 in North Carolina—Poblano had outperformed every established pollster. Clinton never recovered, but with the National Journal, the Guardian and the New York Post suddenly dissecting or demanding the secrets of his success, Poblano became an Internet sensation. "It was kind of amazing," he says.

It only gets better. For the man behind the blog, outpredicting the experts wasn't anything new—even if outpredicting political experts was. On May 30, Poblano finally revealed his offline name: Nate Silver. Doesn't ring a bell? Chances are you're not a baseball geek. Silver, 30, is already celebrated among ball fans for inventing something called PECOTA. Developed while the University of Chicago econ alum slogged through a post-collegiate consulting gig—"I'm used to not sleeping," he tells NEWSWEEK—PECOTA is now recognized as the most accurate system for forecasting how athletes and teams will perform in the future (down to the number of singles). In 2007, Silver's algorithm enraged at least half of Chicago when it said the White Sox—2005 champs—would post a 72–90 record. Turned out PECOTA was exactly right. For laypeople, the leap from the national pastime to national politics might seem like a stretch. But not for Silver (who posted his first political item on Daily Kos in October). "Baseball and politics are data-driven," he's written. "But a lot of the time, that data might be used badly. In baseball, that may mean looking at a statistic like batting average when things like on-base percentage and slugging percentage are far more correlated with winning ballgames. In politics, that might mean cherry-picking a certain polling result." In other words, different sport—same skill set.

From the start, Silver took pride in myth-busting the MSM, which has tended to reduce 2008's complex calculus—delegate distribution, demographic coalitions—into not-quite-true narratives. Obama has a problem with working-class whites? Actually, he has a problem with Appalachian working-class whites—and not their cousins in Oregon and Wisconsin. And so on. The response was ecstatic, and FiveThirtyEight's daily traffic increased 5,000 percent between March and June. But the main attraction was always Silver's primary predictions. Taking a page from PECOTA—a comprehensive historical database, it projects future performance by matching current players to comparable predecessors—Poblano predicted the results in, say, Pittsburgh by measuring how Clinton and Obama did in demographically similar congressional districts earlier on (once set, their coalitions were remarkably stable). Silver's score wasn't perfect—he underestimated Clinton in Kentucky and South Dakota. But ultimately, he came within 20 delegates of the final split on Super Tuesday (out of nearly 1,700) and 2.5 percent, on average, in the other six post-March primaries. "Nate's work is innovative," says Mark Blumenthal of Pollster.com.

Just a snip From Wikipedia http://en.wikipedia.org/wiki/PECOTA

PECOTA relies on fitting a given player's past performance statistics to the performance of "comparable" Major League ballplayers by means of similarity scores. As is described in the Baseball Prospectus website's glossary:[5]

PECOTA compares each player against a database of roughly 20,000 major league batter seasons since World War II. In addition, it also draws upon a database of roughly 15,000 translated minor league seasons (1997-2006) for players that spent most of their previous season in the minor leagues. . . . PECOTA considers four broad categories of attributes in determining a hitter's comparability:

1. Production metrics – such as batting average, isolated power, and unintentional walk rate for hitters, or strikeout rate and groundball rate for pitchers.

2. Usage metrics, including career length and plate appearances or innings pitched.

3. Phenotypic attributes, including handedness, height, weight, career length (for major leaguers), and minor league level (for prospects).

4. Fielding Position (for hitters) or starting/relief role (for pitchers). . . . In most cases, the database is large enough to provide a meaningfully large set of appropriate comparables. When it isn't, the program is designed to 'cheat' by expanding its tolerance for dissimilar players until a reasonable sample size is reached.

PECOTA uses nearest neighbor analysis to match the individual player with a set of other players who are most similar to him. Although drawing on the underlying concept of Bill James' similarity scores, PECOTA calculates these scores in a distinct way that leads to a very different set of "comparables" than James' method.[6] Furthermore, Silver describes the following distinct feature:

The PECOTA similarity scores are based primarily on looking at a three-year window of a pitcher’s performance. Thus, we might look at what a pitcher did from ages 35-37, and compare that against the most similar age 35-37 performances, after adjusting for parks, league effects, and a whole host of other things. This is different from the similarity scores you might see at baseball-reference.com or in other places, which attempt to evaluate the totality of a player’s career up to a given age."[7]

Once a set of "comparables" is determined for each player, his future performance forecast is based on the historical performance of his "comparables." For example, a 26 year-old's forecast performance in the coming season will be based on how the most comparable Major League 26 year-olds performed in their subsequent season.

Has anyone ever tried to take this PECOTA system to FF?

Go to FootballOutsiders.com and pay for their KUBIAK system. It's based on the same premise as PECOTA. FO.com is a sister site of BP.com, so you probably can't get any closer. The KUBIAK projections are also available in Pro Football Prospectus 2008, which is on sale now. Better use of your $$ to buy PFP and the KUBIAK projections (along with your FBG subscription) than buying a useless magazine.

 

[Jason Wood]

I'm wondering what basis the article has for calling PECOTA "the most accurate system for forecasting how athletes and teams will perform in the future." Anecdotally, I haven't found PECOTA to be anywhere close to that claim.

 

[Bri]

Go to FootballOutsiders.com and pay for their KUBIAK system. It's based on the same premise as PECOTA. FO.com is a sister site of BP.com, so you probably can't get any closer. The KUBIAK projections are also available in Pro Football Prospectus 2008, which is on sale now. Better use of your $$ to buy PFP and the KUBIAK projections (along with your FBG subscription) than buying a useless magazine.

I figured Football Outsiders would have something halfway similar but I don't hear you guys rave about the accuracy of Kubiak projections as some do about this pecota. Their prospectus is highly regarded but I don't think they get the same feed back for their projections.Why not?The pecota guy went back to world war II, compares 10 different similar players......do they go into that much detail? I guess going back to when the NFL didn't pass too much wouldn't be useful. Similar players can be an opinion right? There's room for error there then. I guess what I'm asking is translating from baseball to football, do you feel they get it right? How could it be better etc

 

[AB in DC]

The real problem with KUBIAK is that the sample size of past football players is much smaller than in baseball, so there aren't enough "comparable" players in the past to make reliable predictions about the future. On top of that, the game has changed so much in the past few decades -- much more than baseball has. For example, the passing game in the 1970s and earlier looks nothing like today, so looking at "comparable" WRs from more than 30 years ago doesn't really tell you very much about today's players. Player skill sets are a lot different now.

That said, KUBIAK factors in some big things (such as player aging) that many FFers miss. Probably the best strategy is to use KUBIAK as a reality check against your own predictions -- if you project a RB to improve next year, while KUBIAK projects a 20% decline, you may want to rethink. But right now, KUBIAK isn't reliable enough to use as your primary source for player projections.

 

[Bri]

I'm wondering what basis the article has for calling PECOTA "the most accurate system for forecasting how athletes and teams will perform in the future." Anecdotally, I haven't found PECOTA to be anywhere close to that claim.

I don't follow baseball. We all know "the most" and "the best" in fantasy sports are useless terms due to a ridiculous amount of haphazard overuse. I'm not ready to agree and proclaim this guy Nostradamus but it is interesting nonetheless. I got that newsweek mag at least a couple weeks back and have been googling on it from time to time. That guy really does get a whole lot of praise. Naturally, I think it's pretty cool that he's a math geek applying a formula against a database of stats, that he uses history, compares to 10 other players, and that the results included especially high and low predictions. There's a good amount of detailed thought put in there that I've been enjoying reading about.Do you compare your projections for one player to 10 similar? Do you wind up with some extra high and extra low? I don't think so. I think yours are resulting in a single number. I also don't think you allow a computer to project results but instead put your own thought into your projections. So there's some significant differences to get to a similar result. What do you think would be the advantages/disadvantages to doing it his way vs your way? Before ya reply to that Q, let me word that another way-Suppose Doug had made a program that took into account all he did but for football- what would be the advantages/disadvantages to each? I use Doug in the example because if you don't trust the program it's useless. I think you would if Doug had made it.

 

[Doug Drinen]

The real problem with KUBIAK is that the sample size of past football players is much smaller than in baseball, so there aren't enough "comparable" players in the past to make reliable predictions about the future. On top of that, the game has changed so much in the past few decades -- much more than baseball has. For example, the passing game in the 1970s and earlier looks nothing like today, so looking at "comparable" WRs from more than 30 years ago doesn't really tell you very much about today's players. Player skill sets are a lot different now.

[DISCLAIMER: I don't know the ins-and-outs of PECOTA or KUBIAK, so I don't want to inadvertently short-change them. I'm just going to make a few remarks about similar-player projection methods.]This is a nice summary of why similar-players type projections methods are much more useful in baseball and football.

And I'd add a couple of more issues.

1. A baseball player's stat line tells you more about his style than a football player's does. For WRs, for example, you've just got catches, yards, and TDs. Two receivers can post fairly comparable numbers in all those areas without really being very similar. If you want to get serious, you can throw in things like percentage of team receptions. If you want to use targets (and therefore catch percentage), that might help, but you pay a stiff price because it limits your already-much-smaller-than-baseball database to just the last handful of seasons.

2. In baseball, the pitcher-hitter conflict is essentially coaching-independent. In some cases, you have to guess whether a certain young player will play every day or be platooned, but his per-at-bat stats have really don't depend in any meaningful way on how his manager chooses to use him. In football, the players are all at the mercy of what their coach chooses to do. With one twitch of his enormous brain, Bill Belichick could easily have decided for the Patriots to have about 10 more rushing TDs and 10 fewer passing TDs than they did last season. Finding players historically similar to Laurence Maroney wouldn't have helped you predict that.

I'll that said, I gave it a shot a few years ago, more just for fun than anything. Read about it here if you're interested.

 

[Chase Stuart]

The key with this system, as with any other system, is deciding and properly weighting what the important variables are. That's not very easy, and it's often going to be controversial.

 

[ConstruxBoy]

The real problem with KUBIAK is that the sample size of past football players is much smaller than in baseball, so there aren't enough "comparable" players in the past to make reliable predictions about the future. On top of that, the game has changed so much in the past few decades -- much more than baseball has. For example, the passing game in the 1970s and earlier looks nothing like today, so looking at "comparable" WRs from more than 30 years ago doesn't really tell you very much about today's players. Player skill sets are a lot different now.That said, KUBIAK factors in some big things (such as player aging) that many FFers miss. Probably the best strategy is to use KUBIAK as a reality check against your own predictions -- if you project a RB to improve next year, while KUBIAK projects a 20% decline, you may want to rethink. But right now, KUBIAK isn't reliable enough to use as your primary source for player projections.

I like them to compare with my projections to see where I may need to rethink things or more importantly, escape FBG Group Think. Having said that, the one league I did win last year was the one where I showed up at the draft and no one was using a computer so I just used the KUBIAK projections I had printed out.

 

[Wilbur Wood]

I'm wondering what basis the article has for calling PECOTA "the most accurate system for forecasting how athletes and teams will perform in the future." Anecdotally, I haven't found PECOTA to be anywhere close to that claim.

PECOTA pegged the 2008 Tampa Rays W/L record at 86-76, despite the fact that they had never won more than 71 games in a prior season.I spend more time on Fantasy Baseball than Football (largely because there is more to follow, including prospects) and PECOTA is clearly a cut above the rest for certain predictive results such as overall team performance.I don't put as much stock in PECOTA for individual player projections for the upcoming season, although I again give PECOTA a "best in class" grade for their analytic tool in assessing overall career player projections and the player comparable profiles.They use many factors beyond just traditional peformance metrics. They look at "player body type", "draft slot", "age related performance in minor leagues", etc. to shape a profile. They've broken new ground with this model in certain areas. For example, a player like Travis Hafner per PECOTA has a "body type" that correlates with premature erosion of production (big heavy, slow, 1B/DH type of skills, decline more quickly --- Boog Powell's production rate falls off much more quickly than a Dave Winfield/Andre Dawson body type). In fact Baseball Prospectus (the creator of PECOTA) is used and subscribed to by many MLB front offices -- a testomonial in and of itself that I do not belief can be claimed by any Fantasy Football tout services.I can't vouch for KUBIAK to any similar extent. I've bought and casually read their Football Prospectus publication, but I have not done any deep dives into it.

 

[Hoss_Cartwright]

I'm wondering what basis the article has for calling PECOTA "the most accurate system for forecasting how athletes and teams will perform in the future." Anecdotally, I haven't found PECOTA to be anywhere close to that claim.

PECOTA pegged the 2008 Tampa Rays W/L record at 86-76, despite the fact that they had never won more than 71 games in a prior season.I spend more time on Fantasy Baseball than Football (largely because there is more to follow, including prospects) and PECOTA is clearly a cut above the rest for certain predictive results such as overall team performance.

I don't put as much stock in PECOTA for individual player projections for the upcoming season, although I again give PECOTA a "best in class" grade for their analytic tool in assessing overall career player projections and the player comparable profiles.

They use many factors beyond just traditional peformance metrics. They look at "player body type", "draft slot", "age related performance in minor leagues", etc. to shape a profile. They've broken new ground with this model in certain areas. For example, a player like Travis Hafner per PECOTA has a "body type" that correlates with premature erosion of production (big heavy, slow, 1B/DH type of skills, decline more quickly --- Boog Powell's production rate falls off much more quickly than a Dave Winfield/Andre Dawson body type). In fact Baseball Prospectus (the creator of PECOTA) is used and subscribed to by many MLB front offices -- a testomonial in and of itself that I do not belief can be claimed by any Fantasy Football tout services.

I can't vouch for KUBIAK to any similar extent. I've bought and casually read their Football Prospectus publication, but I have not done any deep dives into it.

I don't know too many people that didn't see the Rays surge coming. At least from those who follow baseball. I don't think his prediction of the Devil Rays is all that spectacular. The Rays have drafted very well the last few years and have made some very good moves.

 

[tamales]

I'm wondering what basis the article has for calling PECOTA "the most accurate system for forecasting how athletes and teams will perform in the future." Anecdotally, I haven't found PECOTA to be anywhere close to that claim.

PECOTA pegged the 2008 Tampa Rays W/L record at 86-76, despite the fact that they had never won more than 71 games in a prior season.I spend more time on Fantasy Baseball than Football (largely because there is more to follow, including prospects) and PECOTA is clearly a cut above the rest for certain predictive results such as overall team performance.

I don't put as much stock in PECOTA for individual player projections for the upcoming season, although I again give PECOTA a "best in class" grade for their analytic tool in assessing overall career player projections and the player comparable profiles.

They use many factors beyond just traditional peformance metrics. They look at "player body type", "draft slot", "age related performance in minor leagues", etc. to shape a profile. They've broken new ground with this model in certain areas. For example, a player like Travis Hafner per PECOTA has a "body type" that correlates with premature erosion of production (big heavy, slow, 1B/DH type of skills, decline more quickly --- Boog Powell's production rate falls off much more quickly than a Dave Winfield/Andre Dawson body type). In fact Baseball Prospectus (the creator of PECOTA) is used and subscribed to by many MLB front offices -- a testomonial in and of itself that I do not belief can be claimed by any Fantasy Football tout services.

I can't vouch for KUBIAK to any similar extent. I've bought and casually read their Football Prospectus publication, but I have not done any deep dives into it.

I don't know too many people that didn't see the Rays surge coming. At least from those who follow baseball. I don't think his prediction of the Devil Rays is all that spectacular. The Rays have drafted very well the last few years and have made some very good moves.

Everyone knew the Rays improved but to predict 86 wins when theyve never won more than 71 was VERY radical. You may be the only genius who saw that surge coming. Check the baseball forum here as that PECOTA projection was discussed a ton and even Rays fans thought it was too optimistic.

 

[Jason Wood]

Just to be clear, I've been reading Baseball Prospectus since long before PECOTA was introduced, and continue to be a fan. Doug (not surprisingly) far more eloquently articulates why the methodology works in baseball MUCH more than it does in football. Of course, to be fair, the same could be said for any kind of forward projection. Taking simple 3-year averages of a baseball player is much more accurate than doing the same for an NFL player. Etc..etc...

Being in this business a long time, I understand the inequities of labeling the accuracy or reliability of a given claim. We all have to market our wares; that's part of the game. So maybe there really is some independent study that shows PECOTA to be the most accurate; I just haven't seen it (and I look every year, LOL). For example, I personally have found Ron Shandler's work at BaseballHQ to be, on average, a better tool for winning fantasy leagues than PECOTA's projections. But again, it's really comparing apples to oranges. PECOTA wasn't designed to forecast fantasy or factor in strategy.

Back to the marketing angle. I always cringe when I see things like, "PECOTA correctly predicted the Rays would be great in 2008" as though that's a validation of the mechanism unto itself. The Rays are one of 30 teams, and predicting a W-L record is only one aspect of the very complex process of figuring out the baseball season. So while showing PECOTA had the Rays figured out certainly does nothing to detract from the system, in and of itself it's meaningless. Did they get 50% of the records right so far? 70% 90% 10%

 

[Keys Myaths]

Back to the marketing angle. I always cringe when I see things like, "PECOTA correctly predicted the Rays would be great in 2008" as though that's a validation of the mechanism unto itself. The Rays are one of 30 teams, and predicting a W-L record is only one aspect of the very complex process of figuring out the baseball season. So while showing PECOTA had the Rays figured out certainly does nothing to detract from the system, in and of itself it's meaningless. Did they get 50% of the records right so far? 70% 90% 10%

I hope you know you're coming across as very arrogant here. You might want to tone it down a bit. I know you--and you're not meaning to, but you might be putting off some people with the tone.Anyway, here's the PECOTA projections for this year:

East 	W 	L 	RS 	RA 	AVG 	OBP 	SLGNew York Yankees 	96 	66 	881 	732 	.275 	.356 	.443Boston Red Sox 	91 	71 	841 	745 	.274 	.354 	.429Tampa Bay Rays 	90 	72 	804 	722 	.259 	.339 	.426Toronto Blue Jays 	77 	85 	753 	773 	.266 	.331 	.422Baltimore Orioles 	67 	95 	750 	881 	.262 	.329 	.414Central 	W 	L 	RS 	RA 	AVG 	OBP 	SLGCleveland Indians 	92 	70 	835 	742 	.267 	.343 	.438Detroit Tigers 	90 	72 	842 	762 	.276 	.342 	.439Chicago White Sox 	78 	84 	791 	814 	.263 	.333 	.439Kansas City Royals 	73 	89 	745 	823 	.271 	.332 	.408Minnesota Twins 	73 	89 	713 	782 	.265 	.325 	.405West 	W 	L 	RS 	RA 	AVG 	OBP 	SLGLos Angeles Angels 	85 	77 	809 	778 	.276 	.338 	.425Oakland Athletics 	79 	83 	726 	758 	.254 	.332 	.402Seattle Mariners 	76 	86 	692 	745 	.264 	.323 	.399Texas Rangers 	71 	91 	779 	878 	.267 	.334 	.427National League, ranked by projected 2008 recordEast 	W 	L 	RS 	RA 	AVG 	OBP 	SLGNew York Mets 	93 	69 	799 	687 	.265 	.338 	.424Atlanta Braves 	87 	75 	820 	759 	.274 	.343 	.434Philadelphia Phillies 	86 	76 	842 	788 	.264 	.341 	.450Washington Nationals 	73 	89 	768 	844 	.266 	.335 	.420Florida Marlins 	72 	90 	758 	835 	.259 	.331 	.429Central 	W 	L 	RS 	RA 	AVG 	OBP 	SLGChicago Cubs 	91 	71 	845 	762 	.273 	.343 	.452Milwaukee Brewers 	86 	76 	809 	768 	.261 	.335 	.445Cincinnati Reds 	81 	81 	768 	754 	.264 	.333 	.426St. Louis Cardinals 	75 	87 	710 	766 	.257 	.328 	.406Pittsburgh Pirates 	73 	89 	726 	806 	.263 	.328 	.417Houston Astros 	72 	90 	737 	821 	.262 	.329 	.414West 	W 	L 	RS 	RA 	AVG 	OBP 	SLGArizona Diamondbacks 	87 	75 	821 	763 	.265 	.338 	.448Los Angeles Dodgers 	87 	75 	785 	729 	.270 	.336 	.425Colorado Rockies 	81 	81 	874 	876 	.281 	.350 	.446San Diego Padres 	79 	83 	692 	708 	.252 	.324 	.408San Francisco Giants 	68 	94 	634 	747 	.256 	.313 	.384

Current standings:

American LeagueAmerican League EastTeam	W	L	Pct.	GB	Home	Road	East	Cent.	West	L10	StrkTampa Bay	55	35	.611	-	36-14	19-21	24-17	8-5	11-7	7-3	L 3Boston	55	39	.585	2	34-10	21-29	18-18	15-5	11-9	5-5	W 3N.Y. Yankees	49	42	.538	6½	27-22	22-20	19-17	11-15	9-3	5-5	W 4Baltimore	44	45	.494	10½	25-16	19-29	15-21	9-7	9-10	3-7	L 4Toronto	44	47	.484	11½	24-19	20-28	11-14	15-7	10-16	6-4	W 2American League CentralTeam	W	L	Pct.	GB	Home	Road	East	Cent.	West	L10	StrkChi. White Sox	53	37	.589	-	32-13	21-24	7-12	26-11	8-8	8-2	W 4Minnesota	50	41	.549	3½	32-18	18-23	7-12	24-16	5-9	6-4	L 3Detroit	46	44	.511	7	27-17	19-27	8-8	13-21	12-10	6-4	W 3Kansas City	40	52	.435	14	19-23	21-29	10-20	12-16	5-11	3-7	L 2Cleveland	37	53	.411	16	22-22	15-31	7-6	13-24	11-11	0-10	L 10American League WestTeam	W	L	Pct.	GB	Home	Road	East	Cent.	West	L10	StrkL.A. Angels	54	37	.593	-	26-20	28-17	12-9	16-9	16-11	6-4	L 2Oakland	49	42	.538	5	29-22	20-20	13-9	15-9	11-16	5-5	L 1Texas	48	44	.522	6½	23-19	25-25	10-14	12-11	16-11	7-3	W 2Seattle	36	55	.396	18	19-27	17-28	10-18	6-12	11-16	5-5	W 1 National LeagueNational League EastTeam	W	L	Pct.	GB	Home	Road	East	Cent.	West	L10	StrkPhiladelphia	49	43	.533	-	24-22	25-21	17-14	16-12	12-6	5-5	W 1Florida	47	44	.516	1½	26-20	21-24	19-14	11-10	12-10	5-5	W 1N.Y. Mets	47	44	.516	1½	24-18	23-26	18-16	6-8	14-14	7-3	W 5Atlanta	43	49	.467	6	30-18	13-31	16-19	9-16	10-7	3-7	L 1Washington	35	57	.380	14	19-27	16-30	14-21	10-17	3-9	3-7	W 1National League CentralTeam	W	L	Pct.	GB	Home	Road	East	Cent.	West	L10	StrkChi. Cubs	55	36	.604	-	35-10	20-26	7-4	22-16	20-7	6-4	W 3St. Louis	51	41	.554	4½	26-21	25-20	10-5	23-19	11-9	5-5	L 1Milwaukee	50	41	.549	5	29-15	21-26	10-11	22-14	11-8	6-4	L 1Cincinnati	43	49	.467	12½	26-19	17-30	16-12	11-20	7-11	6-4	L 2Pittsburgh	42	48	.467	12½	27-20	15-28	12-12	19-23	6-4	4-6	L 1Houston	42	50	.457	13½	22-21	20-29	8-8	15-20	12-11	3-7	W 1National League WestTeam	W	L	Pct.	GB	Home	Road	East	Cent.	West	L10	StrkL.A. Dodgers	45	46	.495	-	24-21	21-25	8-8	17-14	15-14	7-3	W 1Arizona	45	46	.495	-	27-19	18-27	11-14	7-13	21-10	4-6	L 1San Francisco	39	52	.429	6	17-28	22-24	9-9	8-16	16-15	4-6	L 3Colorado	39	53	.424	6½	25-21	14-32	9-10	10-12	13-23	7-3	W 1San Diego	36	56	.391	9½	22-28	14-28	9-10	8-12	16-19	4-6	L 1

 

[Bri]

[DISCLAIMER: I don't know the ins-and-outs of PECOTA or KUBIAK, so I don't want to inadvertently short-change them. I'm just going to make a few remarks about similar-player projection methods.]

This is a nice summary of why similar-players type projections methods are much more useful in baseball and football.

And I'd add a couple of more issues.

1. A baseball player's stat line tells you more about his style than a football player's does. For WRs, for example, you've just got catches, yards, and TDs. Two receivers can post fairly comparable numbers in all those areas without really being very similar. If you want to get serious, you can throw in things like percentage of team receptions. If you want to use targets (and therefore catch percentage), that might help, but you pay a stiff price because it limits your already-much-smaller-than-baseball database to just the last handful of seasons.

2. In baseball, the pitcher-hitter conflict is essentially coaching-independent. In some cases, you have to guess whether a certain young player will play every day or be platooned, but his per-at-bat stats have really don't depend in any meaningful way on how his manager chooses to use him. In football, the players are all at the mercy of what their coach chooses to do. With one twitch of his enormous brain, Bill Belichick could easily have decided for the Patriots to have about 10 more rushing TDs and 10 fewer passing TDs than they did last season. Finding players historically similar to Laurence Maroney wouldn't have helped you predict that.

I'll that said, I gave it a shot a few years ago, more just for fun than anything. Read about it here if you're interested.

The linked article - at the charts everyone was coming up with negative result. Seemed like something was "off" there, no? You have some good points about baseball vs football.

 

[Bri]

Looking back, does (insert random staff member here) staff member X's projections generally fare better than the average of the staff's projections? or are the average #s more accurate?

 

[Jeff Pasquino]

The real problem with KUBIAK is that the sample size of past football players is much smaller than in baseball, so there aren't enough "comparable" players in the past to make reliable predictions about the future. On top of that, the game has changed so much in the past few decades -- much more than baseball has. For example, the passing game in the 1970s and earlier looks nothing like today, so looking at "comparable" WRs from more than 30 years ago doesn't really tell you very much about today's players. Player skill sets are a lot different now.That said, KUBIAK factors in some big things (such as player aging) that many FFers miss. Probably the best strategy is to use KUBIAK as a reality check against your own predictions -- if you project a RB to improve next year, while KUBIAK projects a 20% decline, you may want to rethink. But right now, KUBIAK isn't reliable enough to use as your primary source for player projections.

Baseball hasn't changed much over the years, AND it is the most individualized team sport there is, by far. Most of the stats that any fantasy league uses for baseball are batting related, and there's little team impact there (some situations do change things, but for the most part they balance out to offering little).

 

[ConstruxBoy]

Looking back, does (insert random staff member here) staff member X's projections generally fare better than the average of the staff's projections? or are the average #s more accurate?

Linky

 

[Jeff Pasquino]

Looking back, does (insert random staff member here) staff member X's projections generally fare better than the average of the staff's projections? or are the average #s more accurate?

There's a school of thought that a group outperforms an individual in nearly every forecasting or estimating exercise.It's similar to the "past performance is not indicative of future results" statement. You can take 10 experts and average them and on average they'll do a better job in forecasting than just one. Sure one can be better as a "one off" projection for some, but figuring out ahead of time just who that will be is nearly impossible.

 

[Bri]

Looking back, does (insert random staff member here) staff member X's projections generally fare better than the average of the staff's projections? or are the average #s more accurate?

Linky

Thanksto quote it for here

For fun, I included the last preseason ranking of 6 Staff members (David and Joe, Pasquino, Rudnicki, Tremblay, Wood and Yudkin). I also included the FBG Staff Average of all staff members as of September 10th. The FBG Average was a very good 202 points.

Bolt Backer was best at 165.See I think there's something to having an average. I like that part of the Pecota system

(I do notice it's rankings not stats but it's useful just the same)

 

[Bri]

Looking back, does (insert random staff member here) staff member X's projections generally fare better than the average of the staff's projections? or are the average #s more accurate?

There's a school of thought that a group outperforms an individual in nearly every forecasting or estimating exercise.It's similar to the "past performance is not indicative of future results" statement. You can take 10 experts and average them and on average they'll do a better job in forecasting than just one. Sure one can be better as a "one off" projection for some, but figuring out ahead of time just who that will be is nearly impossible.

(brainstorm)So how could that best be utilized then to come up with something like Pecota?

 

[Jeff Pasquino]

Looking back, does (insert random staff member here) staff member X's projections generally fare better than the average of the staff's projections? or are the average #s more accurate?

There's a school of thought that a group outperforms an individual in nearly every forecasting or estimating exercise.It's similar to the "past performance is not indicative of future results" statement. You can take 10 experts and average them and on average they'll do a better job in forecasting than just one. Sure one can be better as a "one off" projection for some, but figuring out ahead of time just who that will be is nearly impossible.

(brainstorm)So how could that best be utilized then to come up with something like Pecota?

Apples and oranges.....I don't see this "PECOTA" model working in football.As for group thought, there's a validity to using concensus answers as a result of the "group being better than the individual".

 

[gheemony]

I think KUBIAK and PECOTA are comparable tools, but KUBIAK probably won't ever be even close to as accurate as PECOTA because (1) limited sample size (16 games v. 162, fewer years of consistent history), (2) it is easier to separate individual performance in baseball than football, and (3) a lack of meaningful statistics (hard to find good football stats).

PFP and FO do a better job of explaining the limits of their projections. They readily acknowledge that their statistics won't catch up to baseball.

 

[ConstruxBoy]

Looking back, does (insert random staff member here) staff member X's projections generally fare better than the average of the staff's projections? or are the average #s more accurate?

Linky

Thanksto quote it for here

For fun, I included the last preseason ranking of 6 Staff members (David and Joe, Pasquino, Rudnicki, Tremblay, Wood and Yudkin). I also included the FBG Staff Average of all staff members as of September 10th. The FBG Average was a very good 202 points.

Bolt Backer was best at 165.See I think there's something to having an average. I like that part of the Pecota system

(I do notice it's rankings not stats but it's useful just the same)

That was just for the QB position. You're right that it's ranking and not stats. If I remember overall, Maurile did a good job and the staff average was better than the message board average and better than most individual guesses.

 

[Lash]

I'm wondering what basis the article has for calling PECOTA "the most accurate system for forecasting how athletes and teams will perform in the future." Anecdotally, I haven't found PECOTA to be anywhere close to that claim.

PECOTA pegged the 2008 Tampa Rays W/L record at 86-76, despite the fact that they had never won more than 71 games in a prior season.I spend more time on Fantasy Baseball than Football (largely because there is more to follow, including prospects) and PECOTA is clearly a cut above the rest for certain predictive results such as overall team performance.

I don't put as much stock in PECOTA for individual player projections for the upcoming season, although I again give PECOTA a "best in class" grade for their analytic tool in assessing overall career player projections and the player comparable profiles.

They use many factors beyond just traditional peformance metrics. They look at "player body type", "draft slot", "age related performance in minor leagues", etc. to shape a profile. They've broken new ground with this model in certain areas. For example, a player like Travis Hafner per PECOTA has a "body type" that correlates with premature erosion of production (big heavy, slow, 1B/DH type of skills, decline more quickly --- Boog Powell's production rate falls off much more quickly than a Dave Winfield/Andre Dawson body type). In fact Baseball Prospectus (the creator of PECOTA) is used and subscribed to by many MLB front offices -- a testomonial in and of itself that I do not belief can be claimed by any Fantasy Football tout services.

I can't vouch for KUBIAK to any similar extent. I've bought and casually read their Football Prospectus publication, but I have not done any deep dives into it.

I don't know too many people that didn't see the Rays surge coming. At least from those who follow baseball. I don't think his prediction of the Devil Rays is all that spectacular. The Rays have drafted very well the last few years and have made some very good moves.

to the above i have a neat tangential add-on: I had a baseball simulation PC game and just for kicks i took all the teams in 2005 and played 10 straight seasons just computer AI with no intervention on my part and in 2008 and 2010 TB won the world series so their system has been ready to surge for awhile (and to think they are missing out on Josh Hamilton just for the simple fact that he couldn't say no to drugs while in their system)now back on topic, another thing that makes football harder to futurecast much more than baseball is injury, a baseball player much more likely to play 90+ percent of his games over the period of time stats are predicted than a football player is