2012 Marcel Projections!
The 2012 Marcel Projections courtesy of Tangotiger are now available in the usual sortable format and in the player pages!
If you are unfamiliar with the Marcel the Monkey Forecasting System, please read the FAQ and all subsequent links.
David Appelman is the creator of FanGraphs.
# of guys forcasted to have more than 30 home runs:
Steamer: 14
ZiPS: 12
Marcel: 2
It also has no one receiving more than 600ABs.
Use rate stats.
And there’s a very good explanation for that. Look at the most durable position players from 2008-2010. 51 guys had 1800+ plate appearances (600+ PA’s a season). That group averaged 651 PA’s over those 3 seasons. However, those same 51 players averaged 558 plate appearances in 2011. That’s about a 15% drop in playing time. Blame Joe Mauer, Stephen Drew, Hanley Ramirez, David Wright, Chase Utley, Adam Dunn, etc.
The other forecasting systems will assume that everyone will continue to be healthy and get playing time. But they’re going to be very very wrong about a handful of guys who unexpectedly get injured or have an unforeseen decline.
Marcel just assumes that everyone has an equally small chance of having a fluke injury, and factors that into its playing time projections. Marcel is not saying that no one in baseball is going to get 600 AB’s. It’s just hedging its bets. And despite the fact that its run by a monkey, it’s right to do so.
Great job Yirmiyahu!
I don’t know how much Marcel forecasted for those guys, but it would have been somewhere between 550-600 PA.
Marcel is not forecasting that exactly 2 players will hit over 30 HR’s.
The system forecasts what you might call a “medium” performance for each player. Out of perhaps a few dozen players forecast to have 20-25 homers, maybe a dozen or so will hit over 30, a dozen or so less than 15.
Exactly. Its usually folly to make extreme predictions. Sure, maybe 20 players a season will have 30 HR’s or 100 RBI’s or 15 Wins or an AVG over .300 or an ERA under 3.20. But its very difficult to project which individuals will reach those milestones.
Over the last 2 seasons, ZiPS projected 28 guys to hit 30+ homers, but only 17 of those projected reached the milestone. On the other hand, Marcel only projected 16 guys to hit 30+ homers, but it was “correct” on 13 of them.
Great job Yirmiyahu !
Is it better to be right 13 on 16, or to be right 17 on 28? Or even say just 3 for 3 (as I presume someone could have made it really really conservative)?
Good point. 2011 was a pitcher-dominant year and yet 20 players got over 30 home runs.
ubaldo 12 W…. thats 1 less than david price.
strasburgh 79 inning pitch..
i think price goes 15-9 and stras goes 125 innings ;) a stint on the DL
These are the worst projections I’ve seen yet. Bob M said it right about the 30 HR projection, but even worse are these facts:
Ryan Braun HR : 24
Mike (Giancarlo) Staton HR: 28
JBats HR: 34 (MLB leader) – Pujols is the only other guy “projected” to hit more than 30 bombs (32).
Verlander wins: 17 (tops in MLB)
Kershaw Ks: 206 (tops) – Verlander the only other 200K pitcher.
Get these projections off this site ASAP – Credibility is getting thrown out the door with these jokers.
Marcel the monkey will be very hurt that you don’t like the projection system he worked so hard to perfect.
Yep. Eel, you should read up on Marcel and the philosophy behind it before you make yourself look more foolish.
Still trying to find your projection system… what site is it on, again?
That’s the same line I use to defend Glenstein’s Dartboard Projection System TM.
We project Jed Lowrie to hit 87,000 home runs.
Win Glenstein’s Money was my favorite show as a kid.
Exactly.
Have him make projections … and then compare to Marcel at the end of the year.
Guys a lot smarter than he have attempted that before. On a large scale, Marcel is pretty good.
Circle: right.
I’ve been doing this comparison thing for nine years now. As much as we think Marcel should NOT do well (and it was purposefully designed to do as little work as possible), STILL the other systems can barely beat it. And the beating only happens with rookies for the most part.
If you are unfamiliar with the Marcel the Monkey Forecasting System, please read the FAQ and all subsequent links.
I believe the OBP calculations you list under projections are slightly off. It appears you did not include Marcel projected Sacrifice Flies in the denominator. They are all a couple points on the high end because of it if I am calculating correctly at home.
It also appears that some of the ERA’s you list don’t match what is in the downloadable spreadsheet from Tango. Not trying to nickpick, just don’t understand why there would be a difference? Thanks.
Hmm… I will take a look at the OBP calculations. I know that the ERA calculations are going to be very slightly off from the spreadsheet because ERA is calculated as an average of mER and mBSR and I round the average mER and mBSR. I think this was the way Tango advised me to go about doing it, but if I’m wrong I’ll definitely change it.
OBP is corrected, and should show up on the player pages immediately on the projections pages in about 1 hour.
Wow, Fangraphs must be hitting the big time if it has multiple commentors who don’t understand why Marcel is named after a monkey.
seriously. I love the feigned outrage coupled with a lack of understanding…
The entertaining part to me is that links explaining Marcel and the process were included in the few sentences of the opening post.
They discovered how to download the spreadsheet and sort the columns. I suppose that’s half the battle.
ESPN points their mongos in this direction every day :(
tell me about it byron. the “buster posey injury opens door for colorado” article last year hit the flood gates for monkeys.
Nah, Giants fans have been like that since they won the world series. There were suddenly a million new fans and they were all internet savvy. Kind of the perfect storm of baseball ignorance, homerism, and internet forum douchebaggery.
For a good primer on the Marcel projection system, as linked above, this is helpful and good to read before looking at the stats if you’re not familiar with the system:
http://www.tangotiger.net/marcel/
The counting stats shouldn’t be taken at face value as noted in the link. Remember, no projection system gets it bang on, but the goal is to be in the ballpark and give the reader a general sense of what to expect. I know, I know. We all want to know exactly what to expect, but that’s not realistic. At first glance it’s alarming to see all the other projection systems a fair distance away from Marcel, but in the end, I have a feeling they’ll all come out close to the same in accuracy scores.
You people are so hilarious.
“These projections suck because player X on my fantasy baseball team is only projected to hit 17 home runs.”
whats so funny? nobody getting a full season worth of PAs is a legitimate complaint.
How projections are formed is an important thing to understand.
monkeys everywhere rejoice
Even on this day of Great Pirates Joy (TM), Marcel has found a way to make me sad about the offensive prowess of the Pirates infield. It’s surely not Marcel’s fault; I just get sad when my delusions are crushed.
even a blind monkey gets a nut…
Russel Martin didn’t even make the front page of catchers. Apparently he’s going to be outside the top 40 in value. Also, the league is about to lose all of its power. Teixiera, Braun, Miggy, Granderson, Fielder, Kemp all falling short of 30 home runs.
As you can probably tell, I’m a bit confused about this system. I am attempting to learn more about the in depth analysis provided by this site, but I just don’t understand what I am supposed to take away from this particular tool. You have a projection system, but I’m not really supposed to take the individual projections as actually attempting to be accurate?
Marcel is the standard by which all other projection systems are measured. As the FAQ says, it takes the last three years of data (where available), weighs the most recent year the heaviest, regresses to the mean, and adds an age factor. That’s it. The challenge for other projection systems is to demonstrate that the added tricks that they incorporate lead to significantly better performance than the system so simple a monkey could’ve designed it.
Marcel is the man. A monkey of a man.
I know many of us are “hilarious” or have a “complete lack of understanding” because we are missing whatever insight some of the commentators so obviously have. It is just not clear to me what value the Marcel system has when, for example, it doesn’t project anyone to have over 600 at bats when last year 20 guys in the MLB did. If you want to dismiss it by saying questions like this are dumb, that’s fine, but it would be more helpful if someone would offer an explanation. And I have read the Marcel site explaining the model but it still doesn’t explain why, for example, Ichiro, who had 639, 680 and 677 AB’s over the past three years is not projected to crack 600. I know he is dropping to #3 in the line-up but the system doesn’t appear to take that into account. That is just one of a number of questions.
I’m guessing Marcel just regresses everyone to the mean a lot more than a more advanced projection system. Probably why Marcel is better for projecting average players than the mighty beasts among mere men who can hit 30+ HRs.
Marcel projections are the most conservative of all projections. Everything is regressed heavily to the baseline. Everyone has less than 600 ABs because Marcel assumes EVERYONE will get hurt during the season.
They are not what you use for stuff like fantasy baseball, the projections simply provide a baseline to compare other projection systems against.
It’s not that everyone WILL get hurt, it’s that everyone COULD get hurt, but no one can get NEGATIVE HURT and play 30 extra games. So the mean projection gets pulled at least a little south of a full season without anything to pull it up.
1) Marcel isn’t very useful for fantasy.
2) Marcel’s best use as a comparison tool to see if the more advanced projection systems can do better than a stupid and simple one.
3) Like it or not, every year, Marcel’s projections are shown to be as accurate as, if not better than, the more complicated ones like ZiPS, Bill James, CAIRO, Oliver, and PECOTA.
Agree with everything you said.
I look at marcel with its intended purpose in mind. But for what I use projections for, I generally go with a weighted combination of guru, Zips, and THT (Oliver). I’ve included Steamer into this year’s mix.
I think to add to this stream, Marcel is probably most useful for projecting the total output of a group of players. Marcel, in a sense, projects the target mean, or perhaps median. By definition 5/10 players will outperform a good median projection. So when a team of 25 players is projected, all of the “overshoots” are more likely to balance out the “undershoots,” and a very good team projection is born.
Law of large numbers, or something…
Right. Imagine that the MEAN forecast of 10 players is exactly 28 HR each. Does that mean that all of them will hit 28 HR? Well, no. The MEAN forecast is the AVERAGE of all forecasts. So, if a guy is expected to hit, on average 28 HR, that means we should expect him to hit say 18-38 HR. Sometimes he’ll hit 18, sometimes 32, sometimes 38, etc. The average is 28.
So, if you have 10 guys, who on average is expected to get 28 HR, it’s easy to see that a couple are going to get 33 and 32 and a couple are going to get 18 and 21, etc.
Even if the average expectation is 28 HR each, it does NOT mean that ALL will hit 28, and none will hit more than 28 (or less than 28).
You can’t just say “Marcel expects no one to hit more than 28 HR”.
Fair point. One of the links from the Marcel page will direct you here, where TangoTiger explains why nobody has more than X whatevers. It’s ‘ancient’ in Internet time, but the philosophy still holds.
http://www.hardballtimes.com/main/article/forecasting-2006/
With regard to Ichiro, it’s not inconceivable that he accrues some DL time – he’ll be 39 by the time the World Series after all. And Marcel does bend the projections based on age. That said, he’s still forecasted for 634 PAs which is only about 6% less than all the other forecasting systems.
Most projection systems have no idea what to do with Ichiro. IIRC, ZIPS has been down on him for years.
If I am correct, I think Marcel is supposed to be oversimplified and should be considered at best the minimum numbers a player should be expected to put up. Given that information, it is useful when comparing to other systems. Or some shizz like that. Chill out, yo.
marcel is good he got adam dunn and alex rios right in 2011.
I figured everyone at Fangraphs would know about Marcel. I guess there’s alot less overlap with my site than I expected.
Perhaps what I should do is create a thread/mailbag here at Fangraphs and take your questions, so I can provide an answer. Clearly the FAQ at my site is no longer an FAQ.
I would urge you to read the link that Brent posted, which I’ll repeat here:
http://www.hardballtimes.com/main/article/forecasting-2006/
You should also read this:
http://www.insidethebook.com/ee/index.php/site/comments/forecasting_pujols_ab/
I urge you to spend at least 5 minutes reading both those articles.
Everyone at Fangraphs knew about Marcel at one point, but the ties to ESPN… changed the composition of people that frequent this site a bit…
Actually, one doesn’t need to spend all that time on even those short articles. Just ensure that these phrases of your own go along with, any time any projection system (Marcel, ZIPS, etc.) is mentioned:
We are trying to come up with reasonable over/unders, numbers that you could find equal reasons where the player will over-perform and under-perform. … So, what to do? Trust the forecast for a group of players, but don’t go betting on any one forecast. … There’s no book, there’s no program, there’s nothing to help you with any single forecast. That’s why we play the darn game, and that’s why we love its drama.
Please bear with here, but I am still confused. I am really putting an effort forth to better understand everything these analysis are trying to accomplish. I’ve never really paid much attention to projections before. Is this particular forecasting system simply an exercise meant to display the futility of trying to predict baseball?
To use an analogy, think of projection systems like the teams in a fantasy baseball league. Marcel would be that guy who didn’t show up and is stuck using Yahoo!’s (or whoever’s) standard rankings. Every other team in the league SHOULD be able to beat this team, but if you can’t do it consistently… well, your team/projection system needs some work.
Here’s a toy example that may help. Let’s say 1/10 players gets hurt and misses a full season. We don’t know which player it’s going to be, but we’re pretty confident one in every 10 goes down (again, over-simplified, but the point remains).
So we’d project 9 players to get 650 PAs and 1 player to get 0 PAs. This averages out to 585 PA per player. Since we don’t know which player is going to get injured, we just project them all for 585 plate appearances.
Thus the projection is likely to slightly under-predict 9 of every 10 players, and massively overpredict 1 player, but it hits the mean (or perhaps median, depending on the system) right on the head. In other words, it is best used for projecting groups of players, not single players.
Good job.
Comparisons between the predictors over a few years? I seem to remember that there’s always a ERA, FIP, xFIP comparison, and I think last year there was a comparison article too. . .
I don’t know about Dee Gordon slugging .390 but hell I’ll take it.
all the people on here talking about fangraphs and the marcel projections and how its going ‘mainstream’ remind me of hipsters who don’t want other people listening to the band they discovered
I can remember when “that band” was Metallica.
If any of you are interested in comparing the Marcels to what really happened and how well the Marcels forecast, check out this blog post. You can pick from many stats and see how they match up.
http://vizwiz.blogspot.com/2012/03/baseball-predictions-how-good-is-marcel.html