Neil Weinberg FanGraphs Q&A – 8/20/14
| 1:38 |
: Hey everyone, I’ll be here at 3pm to talk about our advanced stats, our features, our site, and anything else about baseball you’re interested in discussing. Fantasy and prospect questions are fine, but you’re better off asking people who know about those things.
I’m the Site Educator here, so if you’re looking to learn about stuff, that’s what this is for. I imagine we’ll have some fun Alex Gordon-Jeff Passan-Dave Cameron defensive metrics-y stuff to discuss. If you don’t know what I’m talking about, you have thirty minutes to solve the puzzle! Also, I’m @NeilWeinberg44 on Twitter, so follow me there and contact me that way if you have questions during the week. |
| 3:01 |
: Hey! Let’s chat.
|
| 3:01 |
: I’ll be here until at least 4, but as usual, more questions = more chat! Math already, oh brother
|
| 3:03 |
When I search for team stats and sort by the “Off” category for runs, there are a significantly more negatives than positives. Why does it work out this way on a number based around average? |
| 3:04 |
: One thing is pitchers, who are terrible. But mostly, -20, -20, and 40 average out to 0 even if there are more negatives
|
| 3:04 |
Hey Neil! It’s me, August Fagerstrom, FanGraphs author! What counts as a “defensive game” in UZR/150? I assumed nine innings but, in doing my own calculators, it appears to be slightly off. Any clue how UZR/150 is calculated? |
| 3:06 |
: Hey! So it’s actually based on the number of chances a player has. Essentially, what is this player’s UZR in 150 games of a typical number of chances for that position. Let me find the proper links! One second!
|
| 3:06 |
The section on rate stats vs counting stats is a bit vague. Could you please elaborate on the difference and advantages/disadvantages to both? Thank you! |
| 3:07 |
: A rate stat tells you how good a player is on an at-bat for at-bat (inning for inning etc) basis. A counting or cumulative stat tells you about their total production based on the rate stat and the quantity of playing time
|
| 3:08 |
: So a .350 wOBA is better than a .330 wOBA. That’s a rate comparison. But if the first player has 100 PA and the second player has 400 PA, their actual contributions are different. Make sense?
|
| 3:09 |
ROE don’t count towards batting average. Do they count for BABIP? |
| 3:09 |
: ROE is an in play out according to BABIP. You can certainly make a case for including ROE to be included.
|
| 3:10 |
I am confused. Dave Cameron in his chat refuted this morning that there is a positional ajustement for the defensive part of WAR. Is it true? To be more specific, Gordon doesn’t benefit in having loosy LF defenders in other teams and Trout is not penalized for having elite defenders in CF in their defensive WAR metrics? Thanks |
| 3:13 |
: Gordon loses 7.5 runs over a full season because he plays a position that lesser fielders play (-10 compared to a CF). The positional adjustment controls for the differences between the positions and is included in DEF and WAR. If you want to quibble about what the positional adjustment should be, fine, but it’s a decent estimate. One single poor LF doesn’t change Alex Gordon’s numbers by much at all
|
| 3:13 |
How is the run environment determined for calculating the 2014 wOBA weights? With the small year-to-year variations in weights, it doesn’t look like it can be using just one year of data each time. |
| 3:14 |
: The weights are relative to each other. So a home run in a low scoring environment is a little more valuable, but the big changes are really involved in the wOBA scale and average, which we use to translate things into run values and what not.
|
| 3:14 |
Maybe this is a stupid question, but why aren’t the league average heatmaps not all white? |
| 3:14 |
: Why would they be?
|
| 3:14 |
is bryce harper a bust? |
| 3:15 |
: Heavens no. Harper was one of the best 19 year old players ever! He was a better hitter at 20. He’s having a down year in the power department while dealing with an injury. He remains 21.
|
| 3:15 |
How does UZR and DRS handle shifts? |
| 3:16 |
: If a 1) there was a shift on and 2) the shift played a role in the outcome, it is excluded. this means that if there’s an infield shift on but it’s a fly ball to center field, that play remains. If there’s a grounder to second, it’s tossed
|
| 3:17 |
Is there any way to get all the daily playoff odds in a single database or data source? |
| 3:17 |
: Not right now
|
| 3:17 |
Lazy question, I couldn’t find a good explanation with a quick search of The Book Blog, but do you know of any articles that succinctly describe reasoning behind the distribution of WAR between pitchers and hitters? |
| 3:18 |
: Don’t have one bookmarked. I’ll incorporate this into the WAR section when I update it, so stay tuned.
|
| 3:18 |
Does the position sort button have some quirks in it? For instance when I try to view every catcher with qualified seasons with the Royals it gives me like 8 guys which doesn’t seem like enough… http://www.fangraphs.com/le… |
| 3:19 |
: This is 100% right, if you can believe it! You need like 502 PA to qualify. Not many catchers do that in a single season.
|
| 3:20 |
Aren’t the colors relative to league average? So the league average heatmap should equal 1 and have a 100% correlation to league average and be white? |
| 3:21 |
: No, they’re relative to the numbers on the board.
|
| 3:21 |
do you have some back up for Base Runs that I can get up to speed on? |
| 3:22 |
: Here and the links in this post. http://www.fangraphs.com/bl…
|
| 3:22 |
Any thoughts on Rutney Castillo’s projections when he comes up to the bigs |
| 3:24 |
: This opinion is simply based on what I’ve heard from people who have seen him or have talked with people who have seen him, but it sounds like he could be a nice player with a small chance on helping right now. Basically, he could be a nice supplement to a winning team, but won’t have a chance to reach his potential until 2015/16, as he doesn’t seem to be as advanced a hitter as some other recent Cuban signings
|
| 3:24 |
Following up to the Castillo question his translation of his stats- what was his translated line? |
| 3:25 |
: Anyone have this handy? Saw it earlier but don’t have time to go find it
|
| 3:25 |
Could we please get the playoff odds for April 15, 2014 loaded? |
| 3:25 |
: Well that’s weird. Where did they go?
|
| 3:26 |
Hi Neil – I’m trying to find Base Runs among the team stats. I see BsR but I’m not sure that’s it. (I’m looking for it because Dave says that according to Base Runs the Royals are a below average team. However, BsR says that at 8.5 they are second only to the Nationals’ 11.4.) Can you help? If BsR isn’t Base Runs, what is it? It’s not in the glossary. Thanks! |
| 3:27 |
: BaseRuns are in our standings page. http://www.fangraphs.com/de…
BsR is base running runs, and I can see how that looks confusing now that we have BaseRuns, which is new to the site. Seems like someone who runs the Library should handle that…. |
| 3:28 |
Can you help me understand FG’s Clutch statistic? I understand WPA and LI, but I’m not clear on the distinction with (WPA / pLI) – WPA/LI. What are we subtracting from what? |
| 3:33 |
: The entry for this one isn’t very clear, sorry. It wasn’t super high on my list of updates until like every FG writer wrote about clutch last week.
|
| 3:34 |
: So it’s their WPA divided by their average leverage index minus their context neutral WPA/LI.
|
| 3:35 |
I noticed when Dave wrote a very intelligent article–which was well articulated and argued–there is a strong backlash among people who clearly don’t understand the points. How hard is site eduation, and how do you take their viewpoints into account? |
| 3:38 |
: The challenging part of my job is getting someone’s attention, if that makes sense. Dave writes an article that upset a population of O’s fans, for example, last week. If you have a handle on our stats and buy into them, Dave’s article is really about recognizing the O’s as an example of normal variation.
If you don’t know our stats or buy what we’re selling, Dave is spouting nonsense. So the first group doesn’t need me and the second group doesn’t want to hear it. I guess the right path is to talk to the second group on their terms before they get upset, because people who are upset don’t want to listen to you at all. |
| 3:39 |
how will moving to the OF change Kris Bryant’s value? |
| 3:40 |
: Depends on how well he can make the transition. An average defensive 3B who hits X is worth 10 runs more than an average defensive right fielder who hits X. But typically, you’ll do better in the outfield compared to average. So my guess is that when all is said and done, the move costs him half a win per year maybe.
|
| 3:40 |
Who’s having a better season, Jake McGee or Wade Davis? |
| 3:41 |
: Teaching moment! Custom player list. http://www.fangraphs.com/le…
|
| 3:42 |
: I’d say Davis, just barely. Although McGee makes up a little with a few more innings
|
| 3:43 |
Apologies if I’m missing something obvious, but what is “context neutral WPA/LI”? Aren’t those context-dependent stats? |
| 3:44 |
: Both of them are context dependent stats, but by dividing them, you are stripping out the context. A walk off home run earns you lots of WPA but it has a super high LI, so it washes out.
|
| 3:45 |
Is Chris Archer having quietly a “breakout” season in his first complete season in the majors? Relatively low walk totals, Top 10 in majors at inducing pop-up, slider/sinker combo helping keep hits inside park and on the ground. A few hiccups here and there but a 3-ish fWAR season is fantastic |
| 3:47 |
: I don’t know if it’s a breakout, per say but he’s pitching very well and showed signs he was capable of this in the past. I would classify this is as normal progression. A breakout for me is a huge leap forward, but that’s just semantics
|
| 3:47 |
I’m kind of new to FanGraphs, why do some players have Positive UZR but Negative DEF, and vice versa? |
| 3:48 |
: DEF = UZR + positional adjustment. So a good fielding 1B has a positive UZR, but they get a big hit in the positional adjustment because they play 1B
|
| 3:48 |
Where are the explanations for the Def and Off columns that were added to the site this year? I don’t see them in the glossary anywhere. |
| 3:49 |
: They do not exist! Basically, Dave wrote about them as an article when they came out, but no one was hired to update the glossary until I arrived last month. So there are lots of things to catch up on! Stay tuned, but for now: http://www.fangraphs.com/bl…
|
| 3:50 |
If I have a lengthy question about defensive WAR, is it better to post here, put it in a comment section (where), or send it to you another way? I saw on Twitter where you said you read every comment. That must be exhausting. |
| 3:50 |
: Depends, is it a long question or one that requires a long answer? Try it here and if it doesn’t work get me on Twitter and we’ll do email
|
| 3:51 |
So, you’re Andrew Friedman and you have the stats of Jake McGee and Brad Boxberger from the last 3 years in front of you and you have to figure out which player is more likely to maintain dominant/near-dominant seasons ahead, who do you sign to an extension and who do you try to sell-high on? |
| 3:53 |
: So McGee is better from that perspective, but he’s two years old. I would rather have McGee going forward, but I would not likely sign either for lots of money. That’s a cop out, but I sort of answered.
|
| 3:54 |
Is there a baseline number of PA, or another other metric’s number of occurances, that you feel changes a small sample size to valid proof of a how a hitter fairs against same handed pitching to determine if a platoon is the best option? |
| 3:55 |
: No baseline, just increasing confidence as you add more data. So after 50 PA, there’s almost no value. After 200 there’s a little more. After 600 there’s more. Once you’re getting close to 1000 you’re seeing something real, usually.
|
| 3:57 |
I understand that Fangraphs uses the formula (1+pLI)/2 to determine reliever leverage. I understand how this works for a closer or someone with a leverage higher than 1. But how about a long man? Wouldn’t using this formula overvalue his contributions? |
| 3:57 |
: Why? It will give them less credit?
|
| 3:58 |
Do you think Shane Greene, T. Bauer, K. Gibson and Stroman will have their innings limited this year — will they be shut down in mid- to late Sept? |
| 3:58 |
: Probably. Guessing more of a fewer innings thing rather than complete shutdown for Bauer or Stroman perhaps
|
| 3:59 |
Jesse Hahn and Alex Torres combined aren’t worth more than Brad Boxberger alone, does this fact alone give him a mulligan with the Balfour deal? |
| 4:00 |
: I don’t think Balfour was a bad gamble. He was showing signs of decline, but it’s not like they paid him a crazy amount of money. It’s a really good environment on paper for him too, with a good park and good defense in the outfield.
|
| 4:01 |
Here’s the long question. I really appreciate your work on the site, and this may or may not require a long answer. I have three points and then a question at the end about defensive WAR, prompted by the whole Alex Gordon thing. (A) Looking at all OF chances in 2014 by Inside Edge, Gordon gets 34% more non-routine, non-impossible chances per inning than the league average OF (1 every 19 innings, vs. every 26), so he’s getting relatively more opportunities this year to accumulate positive or negative value. It’s a larger percentage difference than what you might see from player to player with batting opportunity (which is essentially batting order, since both are affected by playing time). (B) Sample size is a huge issue with defensive stats, since we’re talking about ~50 plays for Gordon in that range, and they have a relatively large run value per play, since those plays are effectively what makes him a +20 fielder vs. the league average in 3/4 of this season. (C) MGL also says himself in the FG UZR primer that he would regress single-season UZR heavily for determining performance level in the same season. Presumably orders of magnitude more than he would regress hitting stats. Question: Based on these points, If WAR is trying to give us our best guess as to a player’s value, why treat their runs the same in fWAR? fWAR tries to isolate skill in the pitching WAR equation by using FIP, so why not go to the same lengths for defense? This isn’t meant as a huge criticism because I try to have positive assumptions and just assume there is a good reason for things to be the way they are. And my reasoning could be wrong, so that could also be the answer. |
| 4:03 |
: The trick with defensive metrics is that they have wider error ranges than offensive statistics, but it’s also wrong to say that offensive statistics aren’t open to error as well. Defensive stats just have a little extra measurement error built in.
Let’s think about it like this… |
| 4:06 |
: Let’s assume Alex Gordon is has actually saved 10 runs in LF this year. Let’s say we know this from God.
Now let’s say we’ve done our best guess with UZR and we’ve estimated he’s worth 18 (made up numbers). Now let’s take Mike Trout who has been worth 0, let’s say, and we’ve marked him at -5. We’ve missed on both players, so it looks bad. But it’s not. We’re going to overshoot some players and undershoot some. Which is why we don’t ever say WAR is perfect to the precise decimal place. But we have no way of knowing which player we undervalued and which we overvalued…. |
| 4:08 |
: So absolutely, be skeptical of Alex Gordon’s +20 or whatever. But it could also be too low! And Trout’s could be too high. Probably not, but they could be. Unless you can find systemic bias in UZR, it’s one of our best estimates.
I think the real trick in all of this is that as saber-minded people we use these stats with a deep understanding of their flaws but we don’t talk about their flaws every second of the day. So when someone who is skeptical comes into the conversation, they’re like HEY MAN THOSE ARE FLAWED. But we already built those flaws into our brains. |
| 4:09 |
: I now realize that was a boring way chat experience. Let’s move faster now!
|
| 4:09 |
Gordon has actually saved 14.3 runs this year. Trout, not so much |
| 4:09 |
: I’m now very concerned that God has chosen to read a baseball chat.
|
| 4:09 |
By the way Neil, this is my favorite chat of the week. Thanks for doing this. It’s not as sexy as discussing trade rumors and Mike Trout, but it helps the readers understand the site. |
| 4:10 |
: Thanks! I appreciate that. Tell your friends!
|
| 4:11 |
Will Jake Odorizzi get shut down early? |
| 4:12 |
: Maybe a start or two? But he threw 155 innings last year. So 180 seems reasonable and he’s got 50 left with about 8 starts, so he should have room.
|
| 4:13 |
Colin McHugh has been pretty great thanks to k rate…. smoke or fire? |
| 4:13 |
: Which is the good one? It’s a little heavy, but he has the breaking ball for Ks
|
| 4:14 |
Not sure if this is your brief here, but is there anyway to add minor league players to the custom teams section? Would be great to be able to track their performance in a dynasty league rather than having to go page by page… |
| 4:14 |
: Not right now, sorry!
|
| 4:15 |
I’ve kind of grown more in love with SIERA as I look at it and read about it. You’d think that it would get more mentions because it’s a better predictor of future success than FIP (if only slightly) but it’s rarely used. Looking at what happened, does FIP > SIERA? Seems like SIERA would explain things FIP can’t |
| 4:18 |
: The problem with SIERA is that it’s dependent on a lot of related assumptions that get pretty complicated. It’s a good idea, but I don’t think it’s dramatically more useful than xFIP, even though it’s way more complex.
|
| 4:18 |
Does a high weighted pitch value for a batter represent his true talent at hitting that pitch type? If so, how long does it take to stabilize? |
| 4:20 |
: Correct on the direction, but they’re not really predictive. You’d need lots of data, but by then the old data isn’t so good anymore. Don’t have an official answer to link to though at the moment
|
| 4:20 |
Pick one: A year of UZR or a recent scouting report? Which would you trust more? |
| 4:21 |
: What is the question? I’m not being snarky. Specifically, what am I using it for and where does the scouting report come from.
|
| 4:21 |
Hi Neil – it bugs me more than it should that 99% of the time, Jordan Zimmermann autolinks in FG articles go to Jordan Zimmerman because the parser recognizes Zimmerman and doesn’t get to the 2nd n. Is this something you could fix? |
| 4:21 |
: Can you send me a link to an example? I can’t fix it but I can tell someone about it
|
| 4:21 |
How useful is a pitcher’s actual fielding ability in determining how valuable he is to his team? |
| 4:22 |
: Good pitcher could add half a win realistically. Maybe a touch more.
|
| 4:23 |
http://www.fangraphs.com/fa… here’s an example in the chart. Thanks! |
| 4:23 |
I believe Different Jordan is referring to (another example of) the famous Johan Santa problem. |
| 4:23 |
: Thanks, I will pass this along.
|
| 4:25 |
I know their careers aren’t over yet, but who would you say is more likely to make the HOF of these 2, Carlos Beltran or Adrian Beltre? |
| 4:25 |
: Beltre.
|
| 4:26 |
puig has a 312/394/515 line, while j. upton is 287/363/522. upton has 24 HRs while puig has 13. how is it that puig’s slash line is better nearly across the board but upton could have twice as many HRs? they have identical walk rates and puig strikes out 19% to upton’s 26%. where does this power difference show up in the slash line? |
| 4:27 |
: Puig strikes out way less. http://www.fangraphs.com/le…
Gives him 10 more singles, couple more doubles, eight more triples. |
| 4:28 |
: Going to step away for one sec, get your questions in if you have them and I’ll take this thing to 5pm
|
| 4:31 |
Thanks for your patience, Neil. RE: Clutch, your explanation of context neutral LI makes sense, but I’m not understanding how an individual player’s average leverage index differs from his context neutral LI. |
| 4:35 |
: Ah, yes. WPA/pLI is total WPA divided by total pLI. WPA/LI is the WPA/LI of each individual event added together.
So if you have WPA’s of 1, 2, and 3 to go with LI of 1, 1.5, and 1.8, let’s say, the WPA/pLI is 6/4.3 = 1.4…the WPA/LI is 4. Those are very different, obviously! |
| 4:35 |
It will give them more credit. Suppose you have a long man with a pLI of .8. If you use that formula then he’ll be given credit for having a pLI of .9. |
| 4:36 |
: Oh, I see what you mean. I’ll have to look at that, don’t know the formula off hand.
|
| 4:37 |
Follow up question, would a conversion from wOBA weights to run values look something like this: rHR=(wHR-wOBA)/wOBAScale? Because that gets me rHR=1.40 in 2014, but also in 2004 and 1968. Am I missing something? |
| 4:38 |
: Run values above average. The average value of a PA is way higher in 2000 for example, though
|
| 4:38 |
Are there any stats that tell if a pitcher is drastically worse out of the stretch as opposed to the windup? |
| 4:39 |
: Look at their performance with men on base relative to their performance with no one on. Not always perfect but it’s a decent proxy.
|
| 4:39 |
Is there a primary driving factor in the increase in league wide BABIP (talking post WWII to present), there wasn’t a season above .290 until 1987, then a jump to around .300 post 1993? Does Coors field really affect the league numbers that much? |
| 4:40 |
: Don’t know off hand, but Coors is probably a factor. Might also be a matter of offensive increase, meaning that more fly balls carrier the wall instead of being caught. Would have to check. Also, could have something to do with expansion in general?
|
| 4:41 |
When you see a young SP like Stroman come to the majors and dominate early, what do you look for in forecasting future success? Do minor league stats play a significant role in your evaluation? |
| 4:42 |
: I typically look at their minor league numbers and the scouting report. If the reports and #s say he’s a medium K guy and all of a sudden he’s a 10 K/9 guy, I’m skeptical. I’m always up to change my mind, but if it happens right away, it’s probably just a matter of the league having a bad scouting report or something.
I like to see a guy in the show for a full year before really changing my mind about him, but that always depends |
| 4:42 |
As a followup, Where can I find stats of a pitcher with or without runners on base? |
| 4:43 |
: Go to the leaderboards and use the Split dropdown. Or in a player page, there’s a tab called “splits.”
|
| 4:43 |
I’ve posed this question to Jeff and Dave several times through the season. They tend to disagree but the comparison keeps getting more and more apt: Oscar Taveras is a LHH Delmon Young. Nearly Identical hype, minor league numbers, and awfulness against MLB pitching combined with lack of game-awareness and surprisingly poor defense. Buy the comparison? |
| 4:45 |
: Too soon for me. But I don’t think you’re crazy. It’s easy to forget that Delmon was once a highly touted player with lots of potential.
Here’s why I disagree for now. A huge factor in your analysis is his performance against MLB pitching, but we just haven’t seen enough of that. Taveras has some potential flaws, but I haven’t moved from potential flaws to these are for sure his flaws yet |
| 4:45 |
Dave C mentioned that LHPs systematically outperform their FIP, possibly due to their increased ability to manage baserunners. Are there any studies on this you can link us to, or do you know about how much the difference is? |
| 4:47 |
: I don’t have one off hand, sorry. Basically, lefties 1) limit stolen bases and 2) limit extra bases taken against them because of an ability to limit the size of a player’s lead/jump. That should allow them to get hurt a little less when a batter gets a hit. I don’t know the exact amount, though. Worth looking into if someone hasn’t done it
|
| 4:47 |
To Answer A.Lane, do you think the BABIPs have gone up because players are not worried about striking out and thus when they actually make contact the ball is hit harder? |
| 4:48 |
: Plausible. but that huge spike in 1993 probably means expansion or Coors.
|
| 4:48 |
If I wanted to know every baserunning event in a game’s contribution to BsR score, how could I find that? |
| 4:50 |
: We don’t have play by play UBR, I think because we don’t calculate it ourselves. So I’m not sure that you could do this. You can do it for SB and CS, but not extra bases taken because we don’t have the run values or averages available. Sorry.
|
| 4:51 |
: Alrighty. Let’s call it a day. Remember, you can find me on Twitter @NeilWeinberg44 if you need me during the week. And look for new stuff in the Library, probably xFIP on Friday.
Can’t believe there weren’t more defensive metrics questions after that Twitter fight last night! Enjoy your weeks! |
Neil Weinberg is the Site Educator at FanGraphs and can be found writing enthusiastically about the Detroit Tigers at New English D. Follow and interact with him on Twitter @NeilWeinberg44.
Thank you, Neil.