Confused Says What?… Getting to Know FanGraphs Stats
There are alot of questions in various threads in the forums and in the blog entries over the past few months as to what all these stats mean, especially those in which I’ve played a role. And David has a great series of “getting to know”, and he posts references, etc.
The intent of this thread is for me to capture all those questions, and provide a more complete and nuanced set of responses.
In this thread, no question is too simple or too complex. The question itself doesn’t even have to make sense. The only criteria to posting here is that you are confused.
Think of me as “Dear Abby”.
Fire away, and I’ll answer as I can…
Update: Just to let you know I have more answers starting here. You can make more comments in this thread.
Print This Post
Hey, Tango. Thanks for taking questions. I notice that you feel BP’s WARP set’s replacement level too low so you use WAR instead. WAR, however does not seem to be listed along with FanGraph’s other stats. In fact it seems to be calculated anew whenever you talk about a particular player. Is there a place where I can look up a player’s WAR? Thanks.
Tango,
Right now, not all of the stats on the boards are listed in the glossary. Can we get definitions for all the win probability stats? (I’m specifically looking at REW, which I assume is “Run Expectancy Wins”, but I’m not sure how that would differ from WPA).
Sorry if I’m missing something really basic here.
-Geoff
Thanks Tango,
I have seen mentioned often recently the +/- for positional adjustment, +/- runs for offense and defense and then a translation of these values into WAR. Would you be kind enough to run through that for me, especially passing on the +/- values for positional adjustments.
Cheers.
Tango, thank you so much, for some time I’ve wondered how 10.5 runs = a win. I searched and googled everything and I couldn’t find an easy answer. I’m sure I’m looking in all the wrong places, but that happens. :)
I’m just curious on how that came about.
What is FIP?
WAR
BP’s replacement-level in WARP is too low (there’s no question about this) and they stated that they are going to correct this.
As for WAR, the biggest hurdle for calculating it was the lack of fielding, and now that UZR is here, I would imagine it’s just a matter of time before you see it on Fangraphs. I’m not up on the day-to-day activities here, so I’m not the best person to answer this question.
For the benefit of others: WAR is Wins Above Replacement. It’s the theoretical number of wins generated above a theoretical baseline, based on actual performance. The average nonpitcher in the NL is about 2 wins per 162 games , and in the AL it’s 2.5. (These past few years anyway.) The replacement level pitcher in the NL is .420 and .400 in the AL. (Split between starters and relievers, the replacement pitcher as a starter is .030 lower and the reliever is .060 higher than those win%.)
10 runs is a win: why?
The basic idea is that if you look at all teams in baseball history that have scored 1 more run than they allowed, per game (+/- 0.1, to increase the sample size), you will find that they have a .600 win%. And similarly if they allowed one more run than they score, they will have a .400 win%. That means each additional run leads to 0.100 additional wins, above the .500 mark. And 1 divided by .1 is 10.
The lower the run environment, the more impact each run has. And the higher the run environment, the less impact each run has. So, the 10/1 ratio is not fixed, but dependent on the run environment.
Here is a chart that shows the various win%, for various run environments, at various run differentials:
http://www.tangotiger.net/winactuals.html
BRAA and REW
BRAA and REW are *directly* related. One is denominated in runs (BRAA) and the other in wins (REW).
BRAA which I personally think David should rename as RE24 to make clear it’s runs based on the 24 base/out matrix, doesn’t care about the inning/score, and is sensitive to the base/out state. Is a solo HR the “same” as a bases loaded one? If you think it is, then this is not for you.
WPA/LI
WPA/LI is the most sensitive to the context. A bases loaded, bottom of the 9th, tie game HR is IDENTICAL in reality to a walk, and WPA/LI reflects that. The other measures don’t. Is a HR and walk the same in this illustration? If so, then this is not for you.
wOBA and wRAA
wRAA seems to be based on wOBA, which is purely gameState-agnostic and only cares about the run environment. It doesn’t care about the inning, score, base or out situation. The lower the run environment, the less runs the single generates, and also the less costly the out. If you don’t care about the above, then this is the one you want.
This measure is on par with Pete Palmer’s Linear Weights.
***
It’s all based on what you want. They are all different measures based on considering different contexts.
Any chance of us also getting REW/boLI?
Hey TangoTiger,
Thanks for taking questions. Could you explain ultimate zone rating in layman’s terms? I’ve read the definition that fangraphs posted the other day, so I understand basically how its calculated. But I’d like a more in depth explanation.
Thanks, I appreciate it.
How can I get my mother to stop mothering me. I’m a grown man! I’m almost 40!
But seriously, folks, I’ll be here all week….
Here we go:
Of the various offensive stats, which is the most comprehensive and precise for getting a retrospective look at a players offensive “total value” a la linear weights: wOBARAA (wRAA)? WPA/LI? REW? BRAA? Another one? I.E. which one includes the “most” (not just hits, walks, etc, but also steals, groundouts, perhaps baserunning, and so on)?
If that question isn’t clear, I can try again. Which one would be the best for looking forward?
Side question (may go beyond Fangraphs, feel free to ignore): if none of them include non-steals baserunning, which freely-available baserunning stat is the best out there (are there comparable alternatives to BP’s?)?
Defense:
(a bit outside “Fangraphs, feel free to ignore): Are the Outfield Arms ratings at THT “compatible” with bUZR in the sense that one can add those ratings to bUZR in a total player evaluation? Any other publicly-available arm ratings that you recommend?
Positional Adjustments
At it’s most basic, the question is: “How would Willie Bloomquist field at each position, relative to the average player already at that position?”
It presupposes that Willie Bloomquist is not predisposed to have a special skill that can be leveraged more by one position than the other. That is, he has average speed, average strength, average everything. It doesn’t have to be literally Willie, but any composite will do. Willie fits the definition so well.
Anyway, if Willie played SS, then the average SS would save 7.5 runs with the glove more than Willie would. If Willie played 1B, then the average 1B would save 12.5 runs less than Willie would.
The problem that we are trying to rectify is finding some common baseline to compare each of our fielder to. Since obviously the average SS is a far better fielder than the average 1B, we need to find a player who plays both positions, and is not a “natural” at either position. That’s Willie. I sometimes call this “Wins over Willie”.
With hitting, we don’t have that problem, because we know EXACTLY what the average hitter will do.
The adjustments that I use are:
+12.5 C
+7.5 SS
+2.5 2B/3B/CF
-7.5 LF/RF
-12.5 1B
-22.5 DH
DH get a bonus of +5 runs because it’s harder to DH. So, it’s really -17.5, but I wanted to highlight it as two different steps. PH should get an extra +10 runs to account for how hard it is to pinch hit.
(You will note it doesn’t add up to zero. Eh. I like the symmetry. If this bothers you, add 0.5 runs across the board and it’ll work out closer.)
On my site, I have many threads and many discussions on this issue. You can go here, and I recommend, among others, the most recent one from Nov 19, 2008:
http://www.insidethebook.com/ee/index.php/weblog/category/Talent_Distribution/
So when Dave uses the 10.5 he’s playing conservative? I guess I could look at the graph when I get home. I’m on a road trip with my Blackberry. :)
Next question, does the fangraphs wOBA have Stolen bases, caught stealing, gidp, etc factored in? The only formulas I found from your site did not have these in there from what I could find, (again, I didn’t read everything, and my searches didn’t pull up anything I could find). So if it does factor in all of that, what’s the large formula?
FIP
FIP is fielding independent pitching, and simply takes a pitcher’s performance stats that doesn’t involve fielders (BB, HBP, HR, SO), and puts them together based on how much they impact his ERA. It’s pretty straightforward to remember:
FIP = (3*BB + 13*HR – 2*SO ) / IP + constant
You set the constant so that it matches the league ERA for the year in question. You can optionally add HBP and subtract IBB.
This measure can be considered a dumbed-down version of DIPS.
UZR
Suppose you know how often a ball is bit at the 45-foot mark between 2B and 3B, and suppose you know that the average SS turns all balls hit to that spot into an out 80% of the time. Suppose that Jimmy Rollins gets 100 plays like that, and he turns 90 of them into an out. We say that Rollins is +10 plays above average.
UZR looks at each foot mark between 3B and 2B, and looks to see how often all SS made an out at those marks, and goes through the exercise above.
(Actually, not each foot, but each angle. Same thing.)
Not only the vector, but also the distance from home plate is considered. Again, it’s the same exercise, but more variables are considered.
Not only vector and distance, but whether it’s a GB, FB, LD, Pop. And the hardness hit of the batted ball. And whether the pitcher is a GB pitcher or not. And the park. And the runners on base and outs. And if the batter is lefty or righty.
Now, it doesn’t necessarily break each play out into its own bucket, because otherwise, you won’t have enough other plays from other SS to compare against from the same bucket. So, it buckets a certain number of those variables, and applies adjustments for the other variables.
At its most basic, it answers this question: “Given that an average fielder faces this exact distribution of batted balls, how many outs did our fielder actually record against the expectation?”
So on the UZR subject, if Rollins turns 10 more plays into outs, how many runs or wins is that worth? And is that subjective to position, I assume a +10 SS is more valuable than a +10 1B, cause an average SS would see more opportunities to get outs than an average 1B correct?
Who created WAR?
Converting Plays into Runs
At its most basic, a hit is worth +.54 runs and an out is worth -.27 runs (all more or less, depending on the run environment), so converting one into the other is around .80 runs.
Please read this, carefully and in its entirety, and if you have more questions, post a followup:
http://www.insidethebook.com/ee/index.php/site/comments/why_saving_a_play_is_worth_08_runs/
JWay: please read my “Positional Adjustments” post and if you have more questions, please post a more specific followup.
Scott: WAR exists by many people by many names in many forms. I’ve invented my own version, of which I call WAR. Whether someone else also uses that name, preceding me, or following me, I don’t know.
Alright! Thank you so much for the information! I’ve easily cleared up 4 things today. Hooray learning!
Are there any reasonably good fielding metrics that work for catchers?
WAR v WARP v WSAB
So BP’s WARP has set the replacement level too low and WAR has it right. What about The Hardball Times’ WSAB? Is this stat on the correct scale?
How do you concretely measure replacement level? It seems like the number of AAA catchers/shortstops/starters/whatever that make emergency starts would make for a pretty small data set.
I second Jake’s question, and I ask the same with regards to pitching metrics. Thanks, btw, for this — it’s a great resource — and I’ve always respected the work you and MGL do over at your site.
I second Jake’s first question about catching defense, by the way.
Hmm… triple post, but I wanted to make it clear that I was asking if there are any good metrics that measure a pitcher’s fielding ability… now, I’ll stop.
Plus/Minus works well for pitchers.
http://www.fieldingbible.com
I just want to say that is offical, Fangraphs is not the best thing since sliced bread, IT IS BETTER! Thanks Tango and to Dave an ETC. Keep it up! My NFBC checks thank you ; )
Hey Tango,
I signed up for a subscription to billjamesonline so i could see Dewan’s fielding bible statistics. Now that fangraphs has UZR, I’m questioning which one is more useful?
Now for my question, and I apologize as it might be the biggest newbie question ever. I want to know the difference between +/- and UZR.
For example Tulo was something ungodly like +30 in 07. His UZR though was 5? Are those the same and I just dont know it? AS I have come to understand 10 is a win. Well one says his glove is three wins, one says its a half win. Help me please!
Thanks in advance
Since it just appeared on the site and it’s still not “officially” released. The wRAA is based off wOBA and is (wOBA – lgwOBA) / wOBAScale * PA
I’m in the process of phasing out RC and RC/27, and you’ll notice it has already been replaced by wOBA in the graphs section.
Thanks Tango for taking the time to answer FanGraphs reader’s questions!
This is fantastic. Thank you Tom for your time doing this.
My question is: You said that a hit is worth .54 runs and an out is worth -.27. I was wondering if there is an online resource to find out what the value is for every event that can take place, like ground outs, fly outs, sac flies, XBHs and such? I’ve tried looking online, probably poorly, and I can only seem to find bits and pieces here and there.
WAR v WARP v WSAB
WARP’s replacement-level is too low. This is my opinion as well as the opinion of its creator. Soon though, that will change. Until then, you simply cannot look at it.
WSAB is based on Win Shares. And rather than “replacement” level, which would be the 26th man, it uses a bench level, which, if I understand it correctly, is all the non-starters. It’s an excellent benchline, and I support its use.
The only thing to remember is that while the 26th man earns 0 dollars, the bench player will earn alot more than that. I worked it out on my blog, and I think if you use something like $750,000, then you can link WAR (converted to dollars) and WSAB (converted to dollars).
The current form of WARP has no logical underpinnings, and therefore, is not “real” in any manner. A future version may be, and when it comes out, I’ll comment on it.
UZR v Dewan
Both are based on BIS data. Both use a similar framework.
UZR uses, as I understand it, more parameters. And so, you have to presume that Dewan’s metric is a subset of UZR.
Run values of events
The Book has it.
My site has it (see last line):
http://tangotiger.net/RE9902event.html
Meauring Replacement Level
Since it’s theoretical, we’re on shaky ground. That said, you can create various frameworks to try to understand it. I can’t really explain it in one paragraph, but I will link you to a recent discussion that was based on an excellent article (and you should read both):
http://www.insidethebook.com/ee/index.php/site/comments/replacement_level_using_forecasted_players/
re UZR methodology, is the fielding prowess of the average fielder at each position calculated empirically over the course of every season or is it a hypothetical constant?
The former would be intuitive, but then the value of a guy like Everett or Vizquel can vary greatly over the course of their careers as positional depth and quality fluctuates, although what they bring to the table is the same.
Pitcher fielding
I don’t know if MGL has expanded UZR to include pitchers. He could. I do something very simple: PO + A minus GB * lg
where lg = league average (PO+A) / GB
All I’m saying here is that if Maddux gives up 1000 ground balls and he gets 100 PO+A, and the average pitcher gets 60 PO+A per 1000 groundballs, then I count Maddux as +40. It’s simple and basic, and that means it has its shortcomings. But, it’s the first step to understanding PO and A, in context.
If the Crown of England has the authority to prevent the Canadian parliament from conducting it’s business, is Canada truly a country or is it really an English colony?
BTW, what’s up with Quebec? What can’t it be more like Vancouver?
Canadian Politics
The Governor General of Canada:
http://www.gg.ca/gg/rr/index_e.asp
So, The Queen is purely a figurehead. While she maintains the title of “head of state”, it is the Governor General who in fact performs as the head of state.
This is her (yes, she is black):
http://www.gg.ca/gg/rr/cc/index_e.asp
One of her primary roles is to be responsible for the parliament.
The governor general is appointed by the prime minister.
Any other political question should be made off-thread.
Fangraphs FIP differs from THT FIP (i.e., they have different FIPs for the same pitcher). Is that because one accounts for HBP and IBB and one doesn’t?
Thank you
wRAA represents a game state agnostic linear weight value, and WPA/LI is a game state sensitive linear weights value.
As the value of offensive events can vary given the game state, is it correct to think that after enough of a sample size WPA/LI will be more indicative of a the player’s actual value than wRAA because WPA/LI will capture the ability of a player to adapt their offensive approach to the demands of varying game states, and thus better reflect the player’s actual offensive contribution?
What would be a large enough sample size to allow us to look more to WPA/LI than wRAA?
Hizouse: FanGraphs FIP does include HBP and IBB. I think our + ~3.2 constant might be slightly different, though I’m honestly not sure how. I adjust by season, they might be adjusting by league too?
Mike Ketchen,
I believe the Fielding Bible’s +/- score is expressed in plays above/below average, while UZR is expressed in runs above/below average.
I think you want to convert the Fielding Bible’s “+/- plays” to “+/- runs” to compare it to a player’s UZR.
wRAA v WPA/LI
Correct that since one treats a HR as always worth around .13 wins per PA, while the other oscillates around that .13 figure (from .06 to say .20 wins), then it is possible that WPA/LI will come out better in the long-run. That is, to the extent that a batter/pitcher actually change their approach for each PA. WPA/LI captures the “moving runners over”, while all other measures don’t.
The question, as you are correctly asking, is WHEN does this happen. Since WPA/LI is capturing both greater signal and greater noise, at what point does WPA/LI surpass wRAA as the better measure.
I don’t know. I suspect the answer is after 2000 PA or so, but I can’t confirm that as other than a figure I pulled out of the air.
David, why don’t you adjust FIP by league also? I’m pretty sure you should be, and since we have Tango on an answering-spree maybe he can clear this up too.
Park impact
To followup on my previous post, tangentially to park impacts: if you have Barry Bonds actual HR rate at PacBell (3Com? ) after thousands of PA, then why would you want to know the HR rate of the rest of the league at the same home park? Clearly, each park affects each player differently. Clearly, if you have a large enough sample, you can ignore how the park affects everyone else, and you focus only on the player in question.
However, WHEN does this happen? I don’t know. If all you have is 100 PA from Bonds at SF home park, that won’t tell you how that park affected him. If you have 10,000 PA, that’s good enough for me. But, when does it switch over? I don’t know.
I suspect that if you have 500 PA at home and 500 PA on the road, that that becomes the point at which what you do, and what everyone else does, works out as equally indicative. Maybe as much as 1000 PA? Just guessing.
For FIP adjustments, I would want to do it by league, if only because of the DH issue. But, I don’t think it’ll matter than more than .10 in FIP, would it?
I don’t because of the way the leaderboards are set up. Adjustments by season are easier to do, but doing things by league is considerably more difficult, mainly when doing on the fly calculations (like the leaderboards). This is not to say I won’t change it at some point.
Yeah, for current seasons it’s not more than .10 in FIP. For earlier seasons it can sometimes be as much as .2, but it’s not really going to change your evaluation of a particular player.
Tango the glossary is at odds with you on WPA/LI. Which is correct?
You:
WPA/LI
WPA/LI is the most sensitive to the context. A bases loaded, bottom of the 9th, tie game HR is IDENTICAL in reality to a walk, and WPA/LI reflects that. The other measures don’t. Is a HR and walk the same in this illustration? If so, then this is not for you.
Glossary:
WPA/LI (context neutral wins / game state linear weights): How many wins a player contributes to his team with the Leverage Index aspect removed, invented by Tom Tango.
Steve. If you click on the link and read the “why you should care” section, you’ll see it reads:
Why you should care: Unlike standard linear weights, WPA/LI does take into account the situation. So at times when a walk would be just as valuable as a home run, WPA/LI accurately weights the walk and the home run, where linear weights would still give .13 wins to the home run and the walk .03 wins.
I agree it could probably be worded better in the original glossary page, I’ll try and get that cleared up.
Doesn’t that mean that all standard linear weights are derived from something like WPA/LI?
David,
When you were described the formula for wRAA, can you explain to me what ‘wOBAScale’ is?
Thanks, this is a great Q&A thread.
I assume that the lack of responses about effective catcher fielding metrics means that most of the favorites don’t work for catchers. Are there ANY fielding metrics for catchers besides the basic numbers?
Catchers’ Fielding
I wrote an extensive article in The 2007 Hardball Times as a followup to this article:
http://tangotiger.net/catchers.html
Chris: the “wOBAscale” is a divisor (around 1.15-1.20) to convert wOBA into runs. You can get all the technical details on my blog:
http://www.insidethebook.com/ee/index.php/site/comments/woba_year_by_year_calculations/
Average ERA is different for relievers versus starters. Why not normalize FIP for relievers against average ERA for relievers etc?
Thanks to both Tango and David for this thread….
What is your favorite publicly available DIPS ERA-type stat? Does tRA (quoted often on fangraphs blog) give you more meaningful information than regular old FIP–enough that it’s worth heading over to statcorner when you’re trying to get a quick idea of a pitcher’s true talent?
When you say a pitcher is “a .550 pitcher,” for example, what exactly does that mean? You can refer to starters and relievers separately if you wish.
Ignore that last question.
Hi, Tom,
I got the print of “The Book” earlier this year. It was sent to me across more than 4000 miles, and I’m really pleased with the knowledge in it. Do you mind autograph it for me? and also ask Bill James when will he release his new mystery stories?
By the way, why are the results from the UZR sometimes way off from its counterpart in Chone’s or Dewan’s +/-??
What technologies did UZR use? (the “how” behind it?)
And what is the baseline of the “150″ from the UZR/150?
Thanks Tango,
So wRAA is the same as bRAA?
Chris – Yes, they are the same, but there may be some differences between wRAA and the bRAA you find on StatCorner. wRAA includes SB/CS, but not ROE, while bRAA doesn’t have SB/CS and does have ROE.
What do you see as the next breakthrough(s) in analysis? Is it the batted ball equivalent of pitch/Fx? Accumulation of pitch/Fx data?
Assuming a WAR calculation isn’t too far away (given UZR, bRAA, and David’s general openness to adding awesome stuff to the site), how about a play-by-play baserunning stat and an outfield arms stat (pretty much the same methodology, right?)
My question is a bit more simple. What’s “league average” (AL/NL/ML) for some of the advanced statistics. Specifically, I’m curious about:
For pitchers
BABIP
LOB%
IFFB%
HR/FB
IFH%
For batters
BB%
K%
LD%
IFFB%
HR/FB
IFH%
Contact%
What’s the translation to convert LD% for hitters into the ‘expected’ BABIP?
And what’s the difference between the various contact%es?
Thanks!
By the way, let me say again that this is great.
I would love a total WAR thing on this site, but part of the fun of having it piecemeal is that I can figure it myself and feel like a real “analyst” (even if all I am really is a data entry guy).
But second on the baserunning suggestions Sky made.
Where can I find xBABIP for every player? The recent BABIP article on THT sparked my interest in xBABIP but THT doesn’t list the stat or the formula. It’d be nice to see it on Fangraphs.
Any chance for seeing the stats regressed towards a player’s xBABIP AND towards his personal mean?
Well, first off, I don’t even know how to calculate xBABIP. I understand it takes into account a number of parameters, some of which I don’t have.
You’re basically asking for a projection here, and we already have those.
What do you think are the shortcomings of UZR? I mean, it’s great and all, but what are the gray areas that we need to fill to get a more complete picture of a fielder?
Obviously it misses an outfielders arm strength and accuracy. And it’s not park specific.
Is it possible to make a true, comprehensive fielding statistic?
thanks for taking these…
Hope you don’t mind if I crash this party, but I just want to add that THT’s FIP is indeed based on season and league. Plus, we include HBP in the formula and we don’t include intentional walks.
How do you calculate the pitcher Win% in your WAR calculations? Please show your work with an example if possible. Thanks!
vr, Xei
How much control does a pitcher have over his:
1) BABIP
2) HR/FB
3) LOB%
How much control does a batter have over his HR/FB?
Just wanted to second Sky’s post. I think a WAR-style statistic would be absolutely wonderful (I wouldn’t have to calculate my own anymore), and would be easy to do (I would think) given what you already have on the site. Adding to that info on baserunning and outfield arms would be icing on the cake.
…Since everyone else is plugging their stuff, I thought I’d mention two things that are relevant to questions in this thread:
2008 WAR for all MLB players:
http://jinaz-reds.blogspot.com/2008/10/preliminary-2008-total-value-estimates.html
My methods series, for those interested in what goes on under the hood:
http://jinaz-reds.blogspot.com/search/label/player%20value
-j
Thanks again, fangraphs is probably my favorite site on the internet now, and I read everything i can on it.
One quick question about fielding. Would it be possible to try a fielding WPA/LI? Or would this just not really work at all? Would there really be fielders who make more errors with 2 outs and nobody on then with runners on? If such a fielder exists who makes no errors with no outs and runners on base, but makes a few errors with two outs and bases empty, he would surely be better than a fielder who does the complete opposite. Ive had a debate with myself over whether a stat such as this would actually have any value, and after seeing this thread, I think you can answer it better than I can.
Thank you Tango-san to give me a chance to ask.
I am confused about “Plate Discipline”.
I think O-Swing is for batter how many percent he swings at out of zone balles and is a indication of “Plate Discipline”.
As for pitcher it will be a indication how he could fool the batter and let him swing on out of zone pitch.
What are Swing % and Zone% ?
Also what is F-Strike % ?
Is only strike/(strike+ball) in first pitch or Z/(Z+O) for first pitch?
Z/(Z+O) on first pitch might be a good stat how pitcher fear a particular batter.
DrNaka: You have O-Swing correct.
Swing% is just how often a batter swings at pitches in or out of the zone. You could consider it a sign of how aggressive the batter is.
Zone% is the percentage of pitches a batter sees in the strike zone. It can be an indication of how aggressive pitchers are against a particular batter.
F-Strike % is the percentage of first strike pitches seen or thrown by a batter or pitcher.
Hey Tom — thanks again for all of this. I have a question — I’ve been rummaging through the sites recently trying to search for information about hot/cold zones (you know, red for .300 and above, .200-.300 gets neutral, and below .200 gets cold). I am one who believes they exist — some hitters turn on a middle-in pitch better than others. FOXSports has that information, but it’s limited in so much that it only gives you one year of data.
Would it all be possible, using the Gameday data that we have, to develop an archive of hot/cold zones… and how would one go about doing that?
Thanks again.
What about creating a “Little Things” stat, which would be WPA/LI-wRAA? This would show how much a player contributes than what traditional linear weights shows?
Thanks, Tom.
I love Fangraphs.
FIP for relievers, starters
I remember looking at this, and I don’t think there was that strong a bias that one needs to be worried about it. My personal opinion is to do as few adjustments as needed, because each adjustment brings in 10% new readers and loses you 20% of them. It’s always easier for someone else to take what you have and have them apply their own adjustments, rather than me doing all the adjustments, and then others either accepting the black box, or trying to reverse engineer some of the adjustments out.
It’s a tough call when to do which. With FIP, it is such an extraordinarily easy metric, that I definitely don’t want to tamper with it.
tRA
Yes, tRA is one of the best, if not the best, one out there. Indeed, I have something very similar to it, and is useful in its simplicity. Along the same lines is Studes’ Batted Ball stats in the Hardball Times annual on his old blog. They all do the same thing. Gassko’s DIPS v2.0 I think is the same thing as well.
The idea is to look at the type of batted ball hit, and PRESUME the league average hit/out rates for each type, and IGNORE the actual result of whether it was a hit or out.
150 in UZR/150
That’s per 150 games. Perhaps David should call it: UZR/150G
Autograph
Arranging autographs seems a bit complicated at the moment.
Next breakthrough
There’s no question that it’s PITCHf/x. HITf/x, if ever it gets implemented, would be a close second.
The breakthrough is us getting data that we’ve been trying to infer. And so, the more data you can record, the less inferences we make, and the more we actually know.
Baserunning play-by-play
It’s fairly straightforward in concept, but might be too much effort to code. For example, David simply has to code for something like this:
http://tangotiger.net/destmob1.html
He has to presume the runner will take these number of bases those number of times. And whatever he does over and above that, he gets credit for it.
It’s alot of work to get something that won’t necessarily give alot of payoff. I imagine it’s a question of priority.
League average
I agree that it would be nice to see a “league average” line at the bottom of each chart, just so that we have a frame of reference. Either the 2008 league average, or the 2002-08 average.
UZR shortcomings
Actually, it is park specific. And he does have arms, though I don’t think MGL included it in this version. The shortcomings is none really. As long as you use as much of the data as possible, and try to infer things as intelligently as possible, then you have to conclude that it’s doing as much as it possibly can.
There is an uncertainty level, of course. The shortcomings is in the data collection, not the data processing.
Control over stats
I don’t like the question of “how much control”. The question is really “how much does the metric show how much control…”. The batter has alot of control over his power, but how much does HR/FB show that? If you only have 50 flyballs, then the answer is “not much”. If you have 5000 flyballs, then the answer is “ALOT”.
I recommend reading MGL’s regression.pdf file found here:
http://tangotiger.net/mgl/
Fielding play-by-play
You could try it, but boy would it be complicated. I’m not sure the payoff is there.
Hot cold zones
Yes, I love this chart:
http://tangotiger.net/halejon/allcounts.html
And I agree that it would be great to have it by hitter. All it takes is effort by someone who processes GameDay data.
The Little Things
Yes, WPA/LI minus wRAA would give you the “little things” the hitter contributes, or, more specifically, how much he tailors his approach to the game state. I highly support the suggestion.
I posted this in another topic, but it’s probably more relevant here:
I’m having some trouble finding the correct formula for calculating a pitcher’s WAR. Does anyone know the correct replacement level modifiers for SP and RP? Also, I’ve seen some large discrepancies on what the league average ERA should be – resulting in more than a win difference depending on the formula. If anyone would be willing to run through this process using an example, I would greatly appreciate it.
Replacement Level – Pitchers
League ERA is around 4.30 these days. I set the starter replacement level at about 125%-130% of that, and the relief replacement level at 105-110% of that.
There’s a bit of difference between NL/AL.
Tango, I’m looking at Justin Morneau’s page on Fangraphs (http://www.fangraphs.com/statss.aspx?playerid=1737&position=1B). Marcel has his wOBA going down by a few points; it also seems to project that he’ll lose approx. 90pa from this year to next, which is enough to cut his wRAA by a third. Why would Marcel make such a prediction (seeing that he’s gotten 660+ for last three years)? I believe I’ve heard that Marcel isn’t the most accurate predictor, unless I remember wrong, but that projection seems a bit strange.
My brain hurts.
I would love to see fangraphs either develop or inport some kind of xBABIP stat (based on things like LD%, GB%, FB%, HR/FB, maybe a speed factor) that can then be used to general an xAVG/xOBP/xSLG line. xwOBA would be a similar option probably preferred by the statheads but not as easily understood to the masses (myself included).
Basically try and correct for luck and translate that into a player’s overall hitting value. As I currently understand it, wOBA does a good job giving us a picture of a hitter’s overall context-neutral value. But as I understand it, there is no consideration of luck, so a player with an extremely lucky BABIP (say Hunter Pence in 2007) has a huge wOBA (.384 for Pence in 2007). But that doesn’t really have much predictive value for us since a good chunk of his wOBA was presumably based on seeing-eye grounders, etc. xwOBA or xAVG/xOBP/xSLG (or something else) would give a context-neutral, luck-neutral valuation for a player. It would answer the question of “How valuable was this hitter’s performance if he had league-average (or neutral) luck?”
I suspect something like xwOBA would be extremely consistent for most players. It would be even nicer if this stat was then translated into some sort of run calculation based on 150 games so that one could easily combine it with UZR/150 to get a fast measure of how valuable a given player was in a given season.
The holidays are almost hear, so you all better get on it…
TT – Which book about baseball – that hasn’t been made into a movie – would make a great movie?
Nothing like using terms like “xwOBA” to impress the ladies. I think I just burped up some vomit onto my shoe in reaction to what I just wrote. Must be Friday afternoon.
Thanks, tom. I love this stuff but I have a ton of trouble coming up with any good stats. Have any suggestions?
Thank you David Appelman-san to answer my question.
Still confused about
“F-Strike % is the percentage of first strike pitches seen or thrown by a batter or pitcher.”
So it is based on PITCHf/x
in zone/(in zone + out of zone)
So if pitcher throw an out off zone pitch and batter fouls it it will decrease F-Strike %.
I have other questrion.
Are there any studies about PITCHf/x date vs umpires?
Who was the consistent umpitre by year?
Who had small strike zone and who had big strike zone?
DrNaka
when trying to evaluating a possible trade, what are the most important statistic(s) to use to determine a players worth on the field? (disregarding salary and service yrs remaining)
1) in a position player for positon player trade?
2) a pitcher for pitcher trade?
3) a position player for pitcher trade? (is there a ratio for playing time envolved ?)
short version, no formula’s please.
thank you very much.
O Swing %
How does this measure work for pitchers?
Additionally, it seems as if the Little Things turn out negative for most players. Anyone know if this is because of a general discrepancy in the two stats? If I’m wrong, please feel free to tell me.
My prior question should have been expanded to ask…
How can swing percentages be used to foreshadow the outcomes of a pitcher or hitter?
For example, JJ Putz saw his O Swing % drop 7% from 2006 to 2007 which lead to a 2% drop in Swing % in the same time frame. While this didn’t really catch up to him immediately, it appears as though it did in 2008 as as his Swing % dropped to 42% (or so). Is this a mere coincidence, or is this a factor to consider?
Is there a place where I can find the correlation of stats on a year to year basis, ex. K%, BB%, GB %, BABIP? I had a chart with all that at one point but I can’t seem to find it anywhere. Thanks.
Question about the “Little Things” suggestion: To get them more on the same “scale,” should WPA/LI be put on a runs scale? Should it be divided by 10.5 to do so?
ON the other hand, one could convert wRAA, but again, what is the multiplier? 10, or 10.5?
re Devil Fingers:
Yes, it should. Sorry I forgot to mention that.
Any chance of getting an offensive/150 stat? That way we could directly compare that to UZR/150. I was just looking at Brett Gardner, and I figured I need something like that.
Thanks, Tom.
And it doesn’t matter wins or runs. I personally prefer wins, it is easier when evaluating a player’s contributions.
samg — thanks. I think I mixed up “multiplied” and “divided.”
What sorts of things are covered in “little things.” Runner-advancing sacs and groundouts, moving up a base on those outs, going from first-to-third on a single, etc?
Basically, yes. According to my preliminary findings, many players are in the negative. The main purpose is to see whether a manager is truly justified in justifying playing a player because he does the “little things” well.
Anyone know a good formula for finding the conversion factor for runs to wins as well as wins to runs accounting for specific changes in run-scoring environments?
Thanks in advance.
I hear you talk about bUZR vs. sUZR, and I get that it’s BIS and STATS, inc. but where can I go to find the bUZR info? Doesn’t fangraphs only post the sUZR?
Also, as we know, CHONE’s projections where he regresses UZR (or whatever he uses) to the fans scouting report is really the best way to assess defense optimally (if that’s the right word?) given the less sure we are of advanced defensive metrics. However, now that we have UZR info using both BIS and STATS, inc. data, would you combine both numbers into a better UZR average? Maybe that’s what we need to start doing when evaluating players defensively, especially those like Ichiro where the numbers differentiate more than average.
You mentioned in this blog, “The only thing to remember is that while the 26th man earns 0 dollars.”
While this makes sense at first thought, don’t we always add on minimum salary for a replacement player? I’ve always seen the guys here and at other blogs do a $4 mil + 500K (or whatever is min. salary) to determine total value. Am I missing something here or talking about something else?
Tango, on my comment, I mean where can we find sUZR data to compare with fangraphs bUZR…does MGL only have access to that? That correction should help you better answer my question.
Samg: There’s a few reasons why the “Little Things” could mainly be negative, double plays not being included in wOBA comes to mind, but I think for a truly apples to apples comparison, you’d need to use the same play-by-play data and attribute the average run value of each event to each actual event and then add everything up at the end. This would be a play-by-play version of linear weights and maybe it’s something I can do for 1974 – 2008.
And I think the conversion for runs to wins for 2008 is almost exactly 10.
-David
Could you tell me how to calculate UZR over replacement? Is it available anywhere? If not, could you guys add it to the stats pages?
Marcel and Forecasting Playing Time
Marcel simply goes on what has already happened to tell you its forecast as to what will happen. I recommend these two links:
http://www.insidethebook.com/ee/index.php/site/comments/forecasting_pujols_ab/
http://www.hardballtimes.com/main/article/forecasting-2006/
Future Baseball Movies based on books
From what I heard, Playing With the Enemy. Also, The Book, where I end up with Jessica Alba.
***
ball in play: I have no idea how to answer your question.
***
cody K: go here
http://www.fangraphs.com/blogs/index.php/confused-says-what-getting-to-know-fangraphs-stats#comment-56094
And look at “Control over stats”
***
devil: yes, for the “Little Things”, the wRAA should be converted to wins first. So, wWAA.
***
offensive/150 stat: you can take wRAA divide by the player’s actual PA and multiply by 650.
***
Fangraphs posts bUZR, not sUZR. And sUZR is not available to the public.
***
Right, the 26th man is 400K… kinda, but not really. We really should just talk in terms of marginal dollars. Take everyone’s salary, and remove 400K. Then what I said makes sense.
***
UZR over replacement: discussion is happening here
http://www.insidethebook.com/ee/index.php/site/comments/sabermetric_moves_of_the_2009_pre_season/#288
Learning all of the formulas and studying each glossary on here, THT, BP, etc. is becoming more and more overwhelming every day.
My question is: What are the best ones to use? I mean, is there one stat each for pitchers and hitters that would rank them from best to worst? Which is the best stat for offensive production? How about defensive production? Thanks!
Yes Tom, I know it is around 10 now. However, I want to be able to convert it for any theoretical season. Would it be W=(LGAVRUNSSCORED+(LGAVRUNSALLOWED)?
Runs to Win Converter
http://www.insidethebook.com/ee/index.php/site/comments/more_updates_at_fangraphs/#1
***
Single best stat? Sorry, but you won’t get an answer from me. You use whatever you need to use to answer whatever question you have. Sometimes WPA/LI fits, sometimes WPA, sometimes RE24, sometimes wOBA.
if BA is the poorest stat to determine offensive production, which is the best?
if wins is the poorest stat to determine pitching effectiveness, which is the best?
That doesn’t quite answer my question. Isn’t it just Wins = Runs/Game?
sorry, i now see you wouldn’t commit to my question, asked by mike. my bad.
Samg: runs to win = (1.5 * Park Run Environment) + 3
So, if you score 4 runs per 27 outs, then the runs to win converter is 9.
***
ball in play: the reason I can’t answer is because the adjective you use is so ambiguous. “Best” can mean just about anything, just like “valuable”. Establish a specific objective, ask a question that uses that objective, and then I’ll commit to an answer. Otherwise, we’ll be arguing the equivalent of Jessica Alba, Beyonce, and Halle Berry. If I give you my answer, only one-third of you will agree with me.
Any thoughts on what I wrote/asked (using BABIP, LD%, GB%, FB%, HR/FB%, and whatever else to try and remove luck from wOBA)? I’m wondering how well the new stat (I called xwOBA, but you can call it whatever you want) would be used to predict future performance and overall context-neutral offensive value. Seems like a stat that would do a great job both of summarizing a player’s offensive contribution while trying to remove luck from the equation.
mymrbig: this is exactly the approach used by Graham at statCorner and I support it for pitchers. For hitters, it won’t work quite as well. Studes at HardballTimes does something along those lines (linear weights by batted ball type), and I think it’s a fantastic way to make the presentation.
Indeed, this kind of presentation is what is done in their annual, and is probably the best stats feature of the book:
http://www.baseballgraphs.com/battedballs/index.html
If I had my druthers, I’d merge David A and Dave S, and make a Super Dave monster.
Tom, I had a question about the process one would use to “adjust” a players avg/obp/slg line to account for xbabip. For example, if a players babip was lower than what we would expect, could we then add on the number of “missing” hits back to the players line in proportion to their non home run rates?
For example, if it was determined that the player was unlucky to the tune of 10 hits, and 80% of his hits (that aren’t homers) are singles, then we could add 8 singles and say 2 doubles back to his line and this would adjust the line to expected levels?
Am I on the right track here or completely off base?
Tom, I just want to make a point here. There is a point at which we must stop adjusting for things. Otherwise, we would simply end up with everyone having the same results. I simply feel that this point must be made.
Don’t know if anyone is still reading/answering here, but for those of us too lazy, and who want to go to fangraphs for all the work, for an individual year in question, could we just use that year’s RE24/REW to get the runs to wins conversion?
I think you’d be better off doing: ((lgR * 9 / lgIP) * 1.5) + 3
lgRE24/lgREW will get you a very very close number to that, but I’d think it’d be easier to do it the other way.
devil: yes, you could do the RE24/REW, but if you have low numbers, you may get thrown off a bit because of rounding.
samg: exactly what am I or Fangraphs or whoever is doing that is “adjusting” too much? That is, I have no idea if you are talking about something specific, so that we can have a productive discussion, or you are simply making a general point, to which really, it becomes a philosophical discussion with no framework for debate. I’m not opposed to the latter, but this thread is impossible for that kind of discussion.
Thanks, Tango and Dave. I already did some experimenting with a few players RE24/REW, and the numbers did seem a bit low (around 9.5 or something). I’ll try the formula Dave suggsted — even just messing with b-r’s Runs/Game seemed to get closer to something “right” — 10.17 or something like that.
Should I take this to mean that runs-to-wins should be taken as different for AL and NL, even when estimating WAR for players (which is already on a different scale for position players)?
The problem with dividing a specific player’s RE24/REW is you’re going to get some sort of park adjusted runs to wins conversion and it’s done by game, so if a player played 10 games in DC and 5 games in Philly it would reflect that.
Oh, and I should also add that it’s more like if he had 50 plate appearances in DC and 25 plate appearances in Philly since it’s done on a play-by-play basis. Anyway, unless you’re dividing by lgRE24/lgREW I think it could lead to some issues.
Sorry, one more thing. I know this isn’t technically a “suggestions” thread, but here is one. I don’t know how tough this would be to do, but one of the many cool things with fangraphs is that ability to output player/team/league stats to excel/csv (even though my csv outputs puts the full html script for each player in that field for some reason).
Anyhow, I was wondering if there was some way that would allow users to “customize” what fields get output? This would allow us to individually output, e.g., both the WPA/LI and wRAA to our personal spreadsheets/databases to do out own “Little Things” stat for the whole teams or whatever. Ok, obviously, this is something I want to do for my own blogging stuff and you might implement that particular thing yourself. But it would be cool in general, for generating, say, our own personal relief pitching stats combining different pitching results and leverage ratings, and so on.
It’s a great resource either way, just thought I’d pass on my two cents.
Is there anyway to measure a catcher’s ability to call a game besides the ERA of the pitchers when the catcher is catching? (I know that’s a stat, and I know it’s useless b/c Rod Barajas is gonna end up better than Gerald Laird last year no matter who is actually a better game caller)
[I think it was called “Catcher’s ERA”….which would make sense, but I can’t remember0
Could you take us through the process of calculating a player’s free market salary? I’ve tried to reconstruct the process based on articles on the site, but I have a bit of confusion over playing time and just want to make sure I don’t forget anything.
Thanks
In a recent debate with a fellow fan and friend, we got into it about Scott Kazmir vs. Andy Sonnanstine. The argument rested on who was the better pitcher last year and who was more valuable. Taking a look at their traditional statistics, one would easily come to the conclusion that Kazmir was superior last year. FIP and other defense-independent statistics seemed to favor Sonnanstine. The FIP advantage for Sonnanstine was about .4, but Kazmir held the advantage by 1.2 in RA. Generally, park factors and defense would seem to be the culprits for the difference in these figures (as well as luck and LOB%), but since they play on the same team these factors should be mitigated.
My question is thus:
When reviewing past performance to determine which pitcher was more valuable to the team, what statistic seems to shed the light more? I argued that the one based on real runs was a better indicator of past performance rather than figurative runs. The reasoning to me seems logical; the pitcher that (for whatever reason) allowed less runs in a somewhat context-neutral environment did more to help his team win. In theory, FIP or tRA or DIPs strip luck out of the equation, but in reviewing past performances are they better equipped for measuring value. Why? While luck plays a large part in RA or ERA, it also plays it in these other statistics to a lesser extent.
Thanks
I am a big fan of tRA, but I was always under the impression that a pitcher had little control over if a ball in the air was a FB or LD. How much control do they really have?
Year to year correlation on line drive percent for pitchers is pretty much nil, so it appears they don’t have really any control at all over line drives.
Would you consider that a weakness in tRA then? Personally I would rather make all balls in the air worth the same for tRA because, as you said, a pitcher has little control of LD vs. FB.
You could group LDs and FBs together, I suppose.
But you could also use tRA as a value metric — all those LDs given up actually were more costly than the FBs. If you wanted to judge the true skill demonstrated by a pitcher, use tRA*, which regresses appropriately.
Just because a metric doesn’t correlate doesn’t mean there isn’t a skill. If for example, it is not at all crucial that a pitcher be able to K batters, what will we see? Well, since pitcher won’t be selected based on their ability to K, then you’ll get lots of high K and low K pitchers.
What if it was absolutely critical that you must have K to succeed? Then, in this case, you’ll have lots of high K and higher K pitchers. They will be bunched up.
One of the ways to get a higher correlation is for the underlying population to have a high spread in talent. So, when you see a low correlation, it means it’s either: a) no skill, b) low population spread, or c) both
In the case of Line Drives, it will be impossible for a pitcher to survive in MLB if he had a high line drive rate. Therefore, the population spread must be rather tight. You can be a successful high GB pitcher or a high FB pitcher… you CANNOT be successful as a high LD pitcher. So, those guys are weeded out of our MLB population. So, we know the answer is at least b), and, since I believe players are humans, a) can’t apply, and so, it’s c).
The question on our plate is always the degree to which the metric reflects ability, and not whether a player has the ability.
Tango I am having some trouble understanding UZR. I believe that the statistic is too volatile and the variables need to be adjusted. I would like to use Curtis Granderson as my example here. First lets look at his wOBA (using his 2006,2007, & 2008 seasons ). The are very strongly correlated .333 , .395 , & .374 respectively. I think it was easy to predict that CG would regress toward the mean for his 2008 because of how well he performed in 2007. Now lets look at his UZR data. 12.1, 10.4, & -11. To regress that much completely makes no sense at all if you look at the other data available. His PO/Inn were .293, .330, & .308. So he is still making the same number of PO per inning resonably. So lets look at the other periphels available Assist’s and Errors. A’s were 3, 10, and 5. E’s where 1, 5, and 4. So 2006 was his highest UZR and the least number of errors. 2007 UZR shows the statistics ability to take in acount Assists. 2008 shows the flaw in the statistic, IMO.UZR has the same flaw that fielding percentage does by overvaluing error’s. In reality it is possible that a average centerfielder doesnt get to the balls that CG gets to.
1, 5, and 4 errors is not big difference. It’s just a matter of 3 or 4 plays here. It’s nothing compared to the 20-run difference (25 plays or so) that UZR purports between 06/07 and 2008. So, you can scratch that off as the reason.
Now, you see something more interesting here: 293, .330, & .308
Those are his putouts per inning. Multiply those by 9 x 162, and you get: 427, 481, 449.
So, between 2007 and 2008, you have Granderson making 32 less outs per 162 G. That’s pretty much what UZR says!
The bigger concern is how he could have been considered so good in 2006. The explanation there would be that there weren’t as many flyballs (or “catchable” FB in CF).
But, the 2007/08 data is consistent here. It’s the 06 that’s the outlier.
And, also note that Granderson himself said he sucked terribly in 2008.
Is there any way to reconcile between counting and rate statistics? ie. Total Bases and SLG. This would be especially useful for career evaluations, peak v. longevity arguments.
Thanks.
“Reconcile”? I’m not sure what that is supposed to mean here. You should choose a more appropriate word.
In any case, the “wins above replacement” is the metric of choice in terms of balancing rate stats and counting stats.
Tom, i think you are misunderstanding what I am saying. I want something for individual stat “groupings” like XBH and SLG. I think this is necessary especially for evaluations over a career. Thanks.
It sounds like what you want is very specific. Isn’t the data available enough for you to do what you want to do? You should provide concrete examples if you want to see something that specific implemented.
Well, Tom, let me use the Koufax example. He had a good K rate, but a low number of career K’s. And the issue isn’t data availability. The issue is I don’t know how to “combine” the rate stats and the related counting stats.
Tom, what is the reason for the difference in the value of a replacement player between the two leagues. Is it because there is a difference in the quality of play, or because in the NL there is a lot more subs because of the pitcher batting?
There are better players in the AL than NL.
Here’s a question: Are there park effects on UZR and Plus/Minus besides the “Manny Correction” that accounts for giant walls? Presumably playing 81 games in Wrigley is going to have a different effect on your stats than playing your home games in Florida or San Francisco or Houston, right? Specifically I’m thinking about outfielders since infields are pretty standard besides foul territory.
I guess infields aren’t entirely standard. Turf vs. grass. The quality of the turf. All these things are variable, I suppose.
Yes, alot more. Here’s an old version:
http://www.insidethebook.com/ee/index.php/site/comments/mgl_archives/
Thanks, TangoTiger. I’m buying a copy of The Book for everyone I know.
Also, what is the formula for calculating league wOBA?
Same as for players though it may exclude pitchers.
Could I reverse engineer it, or would the numbers be skewed? Not that I would, just wondering.
I’ve been trying to figure out how UZR/150 is calculated from UZR.
It seems as though it would be simple, but I’m not sure if it’s based off of expected innings played in 150 games as a starter, or what.
So I’m curious as to how that is calculated.
I’m sure I could figure it out myself, but once I saw this I couldn’t resist.
“DG” is defensive games and that’s the denominator.
Wow, thanks.
Can’t believe I never saw the DG.
I can.
Things move VERY FAST on Fangraphs. Kinda like arriving at a party late, or being in the wrong room when the action happens.
Here is the blog post that David had on UZR:
http://www.fangraphs.com/blogs/index.php/uzr-on-fangraphs/
Tom, any chance of a new book?
99.44% nope.
Thanks for the interest though.
And by the way, on the Little Things, change it to RE24-wRAA.
Wow, that was stupid. FORGET I EVER SAID THAT
NEVER
You better!
Any chance of commenting on the scoreboard, and adding live updates of all your stats to the game?
Yes to the commenting.
Probably no to the all of the stats. There’s certain things that don’t come in live, so we can’t add those and other things take too much processing power to calculate on the fly, so those can’t be added either. What’s in the boxscore pages is pretty much what I have to work with, though there might be slightly more this year.
Maybe just wOBA and FIP and the standard stats? And possibly making the graph interactive? I realize how hard most of this stuff is, but I’m just wonderin’.
Thanks guys, and keep up the great work.
What about game score?
Tom,
I have a couple questions about FIP…
Isnt the name Fielding Independent Pitching misleading? “FIP helps you understand how well a pitcher pitched, regardless of how well his fielders fielded.” Shouldnt the FIP formula incorporate some form of stats to account for the defense being played?
I did a real quick look at the 08 marlins
here are the ERA/FIP of the 10+ GS pitchers from FLA last year-the worst defensive team IMO
Olsen-4.20/5.02
Nolasco-3.52/3.77
Miller-5.87/4.00
Hendrickson-5.45/4.76
Johsnon-3.61/3.37
Volstad-2.88/3.82
Sanchez-5.57/4.87
If anything, FIP suggests the marlins might have a good defense, but I dont think many people would make that claim.
Anecdotally, it seems a lot pitchers with ERAs>5.00 have much lower FIPs, is this just coincidence?
I don’t understand the question.
The idea is to look at how the player would perform on a neutral defensive team. We don’t want to penalize the Marlins’ pitchers for their poor fielders, since that’s not something they can control.
As far as FIP for 5+ ERA pitchers being lower, well, sure. There’s two reasons for it:
1) Those pitchers tend to be unlucky in the play of their fielders – the range of pitching performances is smaller once fielding is removed.
2) FIP is linear, while pitching isn’t. So the spread of FIP is (slightly) smaller than it would be in reality.
Tom,
Let me try to clarify, I know what Im trying to ask, but maybe its hard for me to type it concisely.
If FIP was truly a “defensive independent” stat, why does FIP stat say that Olsen, Nolasco, Volstad benefitted from the defense being played behind them when, at least to me, was among the worst, if not the worst in baseball? Shouldnt their FIP be lower than their ERA?
Also, it seems due to the formula used, its very hard for a pitcher to have a FIP over 5.00, while in real life, pitchers have ERAs > 5.00 bc they arent good, not because their defense performed poorly.
My original post was also asking you to comment on the phenomenon of many pitchers with ERAs > 5.00 have a FIP < ERA.
I absolutely hate these kinds of discussions, because you’re trying to start from a conclusion and back into a process. That rarely works well.
For what it’s worth, the Marlins were essentially average in UZR.
Team FIP for the Marlins was 4.58, compared to the team ERA of 4.76. (I used 3.46 as the constant, to line everything up with the 2008 NL.) That’s probably within the margin of error.
My last comments for me on the subject, Is there a link discussing where the constants from the FIP formula came from, I tried searching tangotiger.net but couldnt find it?
To paint a more complete picture in 08, FLA was below league avg in UZR, 3rd worst in OOZ, 2nd in TE, worse than avg in FE, 4th worst in INF UZR & OOZ, above avg in OF UZR, and below avg in OF OOZ.
Why did you use a constant of 3.46 for the NL in 2008? Using the FIP formula that doesnt account for IBB or HBP, the constant would be 3.22, using the FIP formula that does account for IBB or HBP, the constant would be 3.21.
Still using the 3.2 constant, FLAs 2008 ERA was 4.44, their FIP, using both formulas was 4.31 and 4.30, both much less than their team ERA. Using the 3.22 and 3.21 constant the FIPs would be 4.33 and 4.31.
To me, it seems you changed the constant in order to make the data fit ia essentially is statistical “cheating”.
Finally, I dont know where you got the 4.76 ERA for the marlins but thats not correct.
Im stupid,
In my last post I disproved my own point, so basically you can ignore it but I still would like to know where the constants from the FIP formula came from.
Because I looked at RA instead of ERA. I apologize for the confusion.
WAR makes sense when talking about how much money player A is worth versus player B (i.e. players production can be compared on an apples to apples basis).
But when looking at real wins and the pythag (RS/RA), position adjustments aren’t really kosher are they? By that I mean a shortstop didn’t contribute 15 more runs to the RA/RS difference than a corner outfielder just because of the position he played. His production may be more valuable money wise because of where on the field it came from but there is no difference between a +10 bat/+10 glove at short vs a +10 bat/+10 glove when it comes to a team’s won-loss record.
Would you agree?
No.
That should have read: “His production may be more valuable money wise because of where on the field it came from but there is no difference between a +10 bat/+10 glove at short vs a +10 bat/+10 glove in leftfield when it comes to a team’s won-loss record.
I dont think it affects the overall production of the team, but finding a +10 bat/glove in the corner outfield is a hell of a lot easier than finding a +10 bat/glove at SS, and thus the SS is more valuable
Is there a statistic that quantifies the affect a baserunner has on a pitcher? For instance does the threat of Jose Reyes stealing a base cause the pitcher to become distracted and more prone to give up hits? Comparing a players stats with a basestealer on versus without a basestealer on might tell the affect potential stealers have on a pitcher
See study in The Book, basestealing chapter.
Any chance we can keep/ put up all future and past projections?
If you mean can I put up the 2007 & 2008 projections, then probably not. Chances are after the 2009 season I’ll remove the 2009 projections for the 2010 ones. If you want to see older projections, you can probably find them on the authors original sites.
What is the random variation that should be expected in different things like BA. OBP, SLG, wOBA etc? I know it should depend upon sample size but how can it be calculated?
1 SD = sqrt(p*q/n)
n=AB or PA as the case may be
p = success divided by n
q = 1-p
Now, this works for BA and OBP. For wOBA, q=1.1-p. For SLG, that’s more complicated.
and the uncertainty in x-bar is x-bar over radical n?
Shouldn’t estimators for pitchers be non-linear because pitchers create their own environments?
How do first basemen get a rating for double play rate? I saw Justin Morneau had a 0.8 run boost for DPR. This sure isn’t a large boost, but I can’t imagine how he could have contributed a run by adding double plays. Does his scooping ability on double plays really add a run of value? Or is this just random variation?
I think we need a positional adjustment for pitchers.
Any chance of Fangraphs adding a play by play baserunning stat?
It’s somewhat annoying to go to BP for their baserunning stats in order to get the complete picture of a player’s value.
Why do BB% and K% (for batters) not have the same denominator?
The reasoning behind this is that you can’t strike out if you’re not swinging the bat or not getting pitches to actually hit, so K% is per AB and BB% is per PA. The point is to try and isolate ability to make contact with the ball using traditional stats as opposed to actual swing/miss rates.
Thanks!
I had 3 questions, if you have the time.
> 1. I know wRAA is not park adjusted, but is wOBA park adjusted on a
> player’s page?
>
> 2. I heard UZR is not adjusted by position like Plus/Minus is. Is there
> any truth to that? If so, what is it normalized against?
>
> 3. Positional Adjustment Runs
>
> I’m a bit confused by how these are given out.
>
> In a post by Dave Cameron, he says that the positional adjustments are for
> 150 games played.
>
> Yet, when I look at player page like Adrian Gonzalez, I see that he is
> only given -12.5 runs despite playing in 161 games, starting 159, and
> playing 1417.1 innings. More confusing is that he started 161 games last
> year, played more innings, but was given -12.4 runs positional adjustment.
>
> Is there a cap on the positional adjustments unlike replacement level
> adjustments? Even so, why would Gonzalez get 0.1 more credit for playing
> more innings in 2007 over 2008?
>
> I’m confused by that. Maybe i’m looking at it completely wrong. Thanks in
> advance.
1. wOBA is not park adjusted on the player pages. “Batting” under the player value section is park adjusted wRAA.
2. UZR is absolutely adjusted by position. It’s set up so that it’s based against the average at each individual position, not all the positions combined.
3. Positional run adjustments are based on these: http://www.fangraphs.com/blogs/index.php/confused-says-what-getting-to-know-fangraphs-stats/#comment-55754
Those positional adjustments I linked to are per 162 games, which is what we’re using for the calculations so that’s why he’s at -12.5 for 161 games played (rounded to the nearest .1 on the player page). I see where in the Cameron article it said 150, but that’s now fixed.
It’s based on games and not innings for simplicity and to keep the DH positional adjustment easier to handle.
Thank you very much, this clears up a ton of confusion for me.