The 2010 Fan Projections!
For those of you who have been hanging around FanGraphs since at least last season, you’ll know that we carry various projections in the off-season, which are for the most part generated by computer programs.
This off-season, in addition to carrying the various computer generated projections, we’ve teamed up with Tangotiger of insidethebook.com to give you, the fans, a chance to generate your own projection line for each major league player. Hopefully our collective brains will be able to pinpoint things that computer systems don’t.
With that said, let me give you all a quick tour of a projection ballot:
Before you can project any players, you’ll have to select the team you follow most closely towards the top of the screen. If you really don’t follow a team, just pick one. You’ll only have to do this once.
After you’ve selected a team, there are 8 categories for pitchers and 10 categories for position players. Pick the values in the drop-down boxes closest to what you think the player will do in 2010, hit the submit button and you’re done! If you made a mistake, you can always go back and change your selection at any time.
That’s really all there is to it. You can filter players by team, or if you go to the player pages, you can project players individually. If you want to see all the players you’ve projected, you can click on the “My Rankings” button which will show you only what you specifically projected a player to do.
As with all new features, we hope everything is bug free and we test things as much as possible, but if you do notice any issues, please let us know.
14
This seems like a promising feature. Thanks for setting it up.
How do you want us to handle catcher defense? Should we estimate a value, or just treat it like UZR and ignore it?
You can definitely go ahead and project catcher defense if you like, but it’s not required.
Having trouble getting some of the projections to submit, and the majority of the time the “loading” screen sticks (though the projections usually go thru).
I figure its just the initial kinks in a new system.
Love the idea, I would love to see how the “Fan” system fares against the other systems at the end of the season.
Should we figue manager styles into our lineup projections?
The answer should probably be yes. The idea is that while projection systems do not have information about things like lineup spots, platoons, and playing time in general, fans have that available and thus offer a unique perspective.
Great stuff David. Thanks for this. I can’t wait to do this for my team. Will we only be allowed to vote for one team?
You can vote for as many teams as you like.
Yeah, there’s no way that this can end up a clusterfuck.
Nice idea though.
For every crazy optimist about a player, there’s a crazy pessimist.
Big sample, it’ll work pretty decently.
And since it uses a lot of basic stats that most every fan can wrap their brain around quickly, it could provide fangraphs some attention as well.
Oh I like the idea, unfortunately it relies on the intelligence, and more important the maturity and objectivity, of the readers of this site. I like to keep projections pessimistic, so maybe a good idea is that to sort projections that are only done by the fans of rival teams.
I mean, really though I’ll accept anything that’s better than Bill James’ projections, though I’m sure the internet will disappoint me.
You could view this as an authority or a hypothesis subject to empirical test. Viewed in the latter way, this is a very interesting project; viewed in the former, it probably does about as well as the other projection systems.
BTW, I’m pretty sure that this data will get a statistical massage after it’s in there. I’m not sure how stuff like that works exactly, but the basic idea would be find the averages predicted by fans and compare that to the averages in recent baseball years, and then systematically (i.e. in the same way for every player) adjust their fan projections so that the average fan projection matches reasonable expectations of league averages.
Great idea. First try took 10 minutes to load and didn’t go through. Bummer.
I changed a few things on the back end, hopefully it should be running a bit faster now.
Great concept – I hope it works out well. I have been advocating for an online engine for this concept for awhile; glad someone(s) took the time to hammer something out. I am very interested to see how the community projections compare to the pundit projections that are posted.
Will these projections just serve as a personal thing where we can see at the end of the season how right (or wrong) we were – or will all the fan projections eventually be aggregated and averaged out and then appear alongside Bill James, CHONE, etc. on the player pages?
They’ll eventually appear on all the player pages. Once the minimum number of required ballots are cast for a player, it starts showing up on the pages.
Very cool, thanks for this.
Dave,
I’m having an issue seeing where I select a team. I can only see individual players right now.
Its only a problem when I click my projections, not when I follow your link FYI.
Oh, the “My Projections” link at the top of the page is not part of the balloting, that is where you can customize a list of players and is part of the “My Team” feature, if that’s where you see the problem?
Dave,
I guess I’m having a problem with getting to the balloting outside of your link.
Unless of course the link is the only way you want us to get to the ballot.
There will be another link on the site later today. The only way you can currently get to the ballot is by following the link in this post, or if you go to any of the individual player pages and hit the “click here to project this player” button.
Thanks Dave. BTW- not that you don’t hear it enough – what your doing with this site is truly unbelievable.
Are there any plans to eliminate obvious bonus ballots?
Any chance that I can wait till February to do this? I love the idea, but I’d like to wait and see how the FA market plays out so I can take into account lineup effects
Yep, these will be open the entire off-season. You can update your projections whenever you want too.
I love how Pedroia is the first positional player w/ the sample size to be projected.
Apparently all of us nerds love having opinions on Pedroia (see 2008 MVP vote discussions).
I can’t find the projection on Pedroia’s page — where did you find it?
I raised the threshold for when projections will show up on a player page. He’s close, so it’ll be there again soon. We’re still a little touch and go on a couple things with these.
30 sample points, I like the Central Limit Theorem-ness.
Can we please have an option added to decide the amount of starts/appearances a pitcher is going to make? Right now I have Jiminez slotted for 210+ innings…I don’t think he is going to make 37 starts by any stretch of the imagination.
I’m adjusting this. Seriously though, if he throws 210+ innings, he’s probably going to make somewhere around 33-34 starts, so while 37 is too high, it’s not off enough or something I can’t adjust where we’re going to change the ballot to included redundant data.
Right, don’t worry about this stuff. When I did the Playing Time forecast last year, I didn’t ask for plate appearances, just games. And by games, I got to PA. I didn’t do a straight multiplier. But, a higher multiplier the more the games. That is, if you forecast someone with 150 games, I might give him 4.2 PA per G. But if you forecast someone with 50 games, I might give him 2.9 PA per G (something like that).
The same applies here. If you forecast someone with 200 IP, we don’t need you to ALSO tell us how many starts. Pretty much, if you are forecasted for 200 IP, we know how many starts you will make. For example, there were 290 pitchers in the last 10 years with 190-210 IP, and the average number of starts was 32. So, that’s what we’ll see. Would it really matter if someone meant to put 31 and someone else 33?
In my opinion, you can’t ask the fan to forecast TOO much, especially such dependent data.
The way I’m doing it now is just scaling PA by batting order and games for batters and scaling IP by 6.1 per start. I guess at the high ends I need to up the IP per start since the guys who throw 220+ innings will typically go longer.
It’s really easy to change the formula which calculates the projection lines and since the projection lines are not being calculated in realtime anymore because it was a real drain on the server, I have a program which just goes through and updates all of them with any changes every 10 minutes.
Right, no biggie to worry about it now. Batting order, for example, means each guy gets one-ninth PA more than the next slot. Games, as I said, has a sliding function. Same thing for IP. All these things are tweaks that will get handled in due course.
On this page:
http://www.fangraphs.com/projections.aspx?pos=all&stats=bat&type=fan
Can you list the projected fielding rating, too, please, David? (Position adjustment is always nice, too, so you can show people how much position matters.)
Would also love to see WAR/150 listed.
Sky, I can add Fielding, but probably not WAR/150, just because it isn’t anywhere else on the site, but maybe I can get WAR/150 added everywhere at some point.
This is great, guys. I will promote it big time.
I’m not crazy about WAR/150.
The entire point of WAR is that it includes playing time. If you are going to force the playing time component, you might as well do WAA (wins above average) / 150. After all, the difference between WAR/150 and WAA/150 is that you have to add 2 wins to WAA/150 to get to WAR/150.
WAA is fine, too, although it has the downside of being on a different scale.
Wouldn’t it be better to standardize the number based on Plate Appearances instead of games?
Unless your leadoff hitter is Willy Taveras, a per 150 games number would inflate leadoff hitter numbers relative to middle of the order guys.
Great stuff! The My Rankings feature is no longer showing the players I projected. It has a random series of names that differs from the players with the minimum number of ballots. It worked fine earlier.
1. I’ve found a Bug in the My Rankings page (http://www.fangraphs.com/fanprojections.aspx?teamid=-1) I am getting someone else’s rankings mixed in with mine. I’ve never even heard of Jordan Brown let alone bothered to project him. I only did two projections last night and I already have 11 projections in My Rankings.
BTW someone’s really optimistic on Pedro Martinez – 10-4, 3.25, Nolasco – 3.25, CC – 2.75 and Halladay 2.75.
2. Is there a reason you are using dropboxes rather than letting us input more exact numbers? Personally I find the dropboxes less convenient as well as less accurate, particularly at the fringes. If you’re asking for fan’s opinions shouldn’t you trust them to be able to choose between 152 games played and 148?
I have the same problem right now. I was getting the guys I voted on.
I had Yuni Betancourt at -0.8 WAR and I wanted to enjoy my handiwork.
I did free-form text once. You wouldn’t believe the headache that caused me.
.763
763
76.3
Were all OPS forecasts I received. If you put user-validation on the front-end, you are going to p-ss off the user. If you put the validation after-the-fact, you have to worry about all this (and more!).
As for the fringes, ideally, if you think someone is going to get a .300 BA (and that’s the boundary) and are unsure about .270-.299 and .300-.329, believe me, there will be plenty of people just like you. It’ll all average out. Basically, the chances are close to nil that everyone will gravitate to the same upper boundary.
I could see such a problem with OPS but with the fields that this is asking for none of them would require a period since they are all whole numbers (even batting average would always be 3 digits (and only three digits) so the form could just place a period immediately in front of the form box, anything other than 3 digits would be rejected). If fans can’t figure out something simple like that is their opinion really worth anything?
Whoops, I was updating some code and accidentally made everyone look at my projections, should be fixed now.
The dropdown boxes are there just to make life easier. I don’t think it’s going to make much of a differences whether you project a guy for 152 games or 148 games, and even less when it comes to the actual community projection.
“I don’t think it’s going to make much of a differences whether you project a guy for 152 games or 148 games, and even less when it comes to the actual community projection.”
That’s exactly my point but since we have to choose between 150-62 or 120-49 that is a huge difference. Presumably you are taking the middle value so it’s the difference between 135 and 156. I think where this becomes an issue is for players that have hit 150~5 games for two or three years. Realistically you would think that a fair projection would be just under 150 games but that seems too unfair, so you might just choose 150-62.
Ultimately it is then the outliers who are making the decision on where the player’s stats end up rather then averaging out a the consensus number. I understand that overall this system works pretty well but I still think that at the fringes we’re going to get less consistent results than allowing a free form submission.
Can you at least consider smaller gradients? You could still use a dropbox to prevent incorrect submissions but allow many more options. Such a 1,2,3 games, etc and 16.7, 16.8, 16.9 K%. On most browsers nowadays you can just type the numbers/letters in a dropbox and they will come up in order.
Anyway, thanks for this toll it looks really awesome so far.
I really think that doing 16.7, 16.8, 16.9 K% is overkill. You’re talking about the differences between 1 strikeout. Even full percentage points is only a 5 strikeout difference if you’re projecting someone for about 500 AB.
The only drawback I really see for having slightly larger projection buckets is if you’re compiling your own personal list of un-aggregated projections.
We’ll see how this system works for this year, and if we need to add more selection options in the drop down boxes next year, then we’ll do that. I would honestly be more concerned about changing the ballot at this point and how that might effect the results since there have already been about 5000 projections entered so far.
Whoops, yeah, I meant 1% buckets, mind was somewhere else. I can see not wanting to mess with it now. Hope this gives us some great results. Thanks for all of the work that was put into this.
Thanks Dave.
Oops, Yuni at -1.1 WAR. In under 500 PA. I think I was generous about his fielding, too.
Seriously? ‘Optimistic’ on Halladay putting up a 2.75 ERA? When he’s posted 2.78 and 2.79 the last two years and the defense behind him isn’t likely to change much? Or am i missing something?
How are you coming up with the aggregate projections? When submitting a ballot, you don’t choose things like HBP or specifically doubles or triples…so, how do you come up with what percentage of 2B+3B are doubles? How do you come up with the HBP figure?
Also, how is it coming up with R and RBI?
Currently I’m looking at recent averages for individual players to determine things like HBP or 3B. Runs and RBIs generated as a function of a player’s wRC and Batting Order.
While I haven’t done any studies or anything yet, I’m quite pleased with the R and RBI numbers we’re spitting out, but there may be some changes from time to time on how we handle these extrapolated stats.
One thing that really glares is that in determining statistics between the chosen ERA and K/BB, stranded% is always near 70%, while BABIP varies widely as does H/9 and WHIP. The typical Roy Halladay projection looks something like this:
ERA: 2.75
FIP: 3.00
strand: 70%
WHIP: 0.78!
BABIP: .220!
H/9: 5.8!
Wouldn’t it make more sense to try and keep BABIP constant instead of strand, since it tends to vary less and is less reliant on a pitcher’s approach to regress?
Hits for pitchers do need to be adjusted, these are definitely coming in too low.
This feature just made my life, haha
or it at least makes my job a hell of a lot less boring.
thank you ever so much for this brilliant application, and i must say it has a very convenient set-up and user interface.
I cannot wait, as I am sure the writers on this site can’t wait either, to see the correlations between the Fans projections and those of BJ, Marcel, and Chone; and more importantly and excitedly to see the correlation between actual 2010 performances.
I can’t help but wonder how fantastic it would be to trump the correlations made by the various computerized systems.
This is phenomenal. Thank you for doing this.
Just curious as to how WHIP is calculated in the projections since there’s no H/9 slot. Great feature, at any rate.
Ok I have a question: After I submitted my projection, which projections am I seeing under “My Rankings”? Is it just my projections or the generell fan projections for these players?
Seriously, nobody can answer this?
Don’t worry about the buckets being too big. It simply won’t happen that there’s going to be so many people wondering “should I mark it at 149 or 150 G” that they will bias it toward one side too much.
Indeed, the buckets are identical to what I ran on my site last year:
http://www.tangotiger.net/survey/index5.php
In practical purposes, you are not going to see what you think might happen. It doesn’t happen.
And the drawback behind having too small buckets is that you are making the dropdown list pretty long, and now you are asking for precision, when all you really need is a quick answer. This is the point: you can ask someone to spend 30 seconds to forecast a player or you can ask him to spend 90 seconds, and those extra 60 seconds, while valuable if all you have is one forecaster goes out the window as wasted effort because once you have 20 or 30 fans, those extra 60 seconds each person is going to spend will be useless.
HOWEVER, if you really really really (has to be three reallys, not two or one) think this is an issue, you find me a “boundary player”, and maybe we can set up a test case and have people re-evaluate him, and see if we get anything that’s really different.
Eh, just ran into a bit of a problem with Adam Dunn. his UZR cannot be projected to be any worse than -20, but it’s almost a certainty that it will be well below that. I know numbers like that are uncommon enough to be considered outliers, but in a case like Dunn, where he is putting up that insanely bad number so consistenly, that option should be made available to get an accurate fan projection.
Any chance that our personal 2010 projection can be included with the 2007-2009 stats on the projection page, not just the My Rankings page? The instant feedback will make for better projections.
Great work all around everyone, not just useful but entertaining and engaging interface.
Mr. Tangotiger, this might not be the appropriate fourm and it’s obvious you spent a lot of time contributing to this tool — but any chance you could let us know when the 2010 MARCEL projections might be out? I believe 2009′s came out on November 9th last year.
Fan projections more than suffice in the meantime though. Great stuff.