Your Opinion of Royals Magic, Reviewed
I promise we’ll move on any moment now. The Royals are champions, but we’ve known that for a couple days, and fans of 29 other teams are ready to look forward. There’s talk about qualifying offers. Players declaring free agency. The offseason is beginning, and the offseason is fun to think about, because if they handle the offseason right, then your team can be the next team people don’t want to hear about anymore a few days after the World Series. If hope springs eternal in March, it begins welling up in November. Baseball’s weird calendar is already flipping.
So pretty soon we’ll talk about other stuff. Important events are right around the corner. But the World Series just ended. Like, two days ago, there was still baseball, and the Royals were as entertaining as any team I’ve seen in forever. Before I say goodbye to them, then, I want to re-visit last week’s poll. I don’t always re-visit the polls I post, but this one, I got particularly excited about. And the results didn’t let me down.
If there’s one thing that reliably turns our comment section into an Andy Capp-style fight cloud, it’s when people accuse one of our models of being flawed. Most frequently, the target is WAR, considered untrustworthy on the position-player side because of defense, and considered untrustworthy on the pitcher side because of also defense. WAR isn’t always at the center of it, though — you’ll see arguments about team projections, generally coming from fans whose teams aren’t playing at the projected level. Fans will accuse the projections of missing something or somethings. They’ll say you can’t reduce baseball to a sheet of paper, and of course they are right. Baseball is impossibly complicated. Projections are educated guesses.
Given that they’re just educated guesses, I do want to come to their defense — projections are so much better than nothing. But every year, a team will outplay its projection, and it’ll even outplay its own statistics. (Run differential, BaseRuns, whatever you want.) So then the same conversations always take place: some people figure the given team is special, and some people figure it’s not. Some people argue why the team is better than it seems. Some people argue why it’s actually exactly what it seems. You’ve seen these exchanges. You’ve participated in these exchanges.
Forever ago, the potentially exceptional team of choice was the Angels. More recently, it became the Orioles. Now it’s the Royals. The Royals just obliterated their projections. They did the same to BaseRuns, and a year ago, the same sort of stuff happened. So it’s the Royals who are now discussed as being greater than the sum of their parts. There are people who strongly believe this. There are people who remain ever skeptical. This is why I posted the poll.
The poll asked you to consider a hypothetical Royals projection. The projection put the Royals at 84 wins. The team would have all the same players. So I wanted to know where people would mentally project the Royals, given what the initial projection said. As an attempted measure of the community sense of “Royals magic,” I figured I could weight the results and then compare to 84. Any difference would be the estimated Royals factor, crowdsourced. The default response would be 84 wins. A vote for something else would be a vote for the Royals not being fully understood by the team-projection math.
Here’s how the lot of you voted:
Two quick things first. One, this would’ve been best done with money on the line. As is, there was no consequence, no downside to selecting an option you didn’t really believe. All this was was an anonymous vote on an Internet poll. Could be people lying. Look at the far left. There were people lying.
And two: I don’t know how to control for possible misinterpretation. I wanted people to vote based on the hypothetical 84-win projection, but some people might’ve just voted for their own Royals projection, having nothing to do with 84. That would mostly be a problem on my end. Gotta be careful how you design these things. I could’ve designed better. The results I have are the only results available, so let’s make of them what we can.
The most popular response: 84 wins. These voters saw the projection, and decided, no, the Royals don’t really do anything special to beat that projection. Whatever they’ve done has been fluky. I expected this to be the most popular response, but I didn’t have a guess for the magnitude. As it turns out, this got barely over a quarter of all voting support. Meaning roughly three-quarters of voters think the Royals do something the numbers don’t effectively capture.
I can’t think of a good reason for voting below 84, but anyway, 4% of voters put the Royals below 84 wins, and 70% put the Royals above. Notice the second-most popular response: 91 wins and up, with nearly a fifth of all votes. This captures how the FanGraphs community is torn. I think this also captures some of that misinterpretation of the question, but you clearly see the two camps. One large group sees the Royals as just another baseball team. A different large group sees the Royals as wildly exceptional. As a team with maybe some clutch character, and definitely contact-hitting and high-leverage relief. These things would be responsible for a considerable gap, it’s suggested.
Between 84 and 91+, it’s flat enough. There’s a weird emphasis on the even numbers that I can’t explain. If I plug in 92 for 91 and up, and if I plug in 76 for 77 and down, then I get an overall average of 87.2 projected wins. If I eliminate all the votes for 84 and just consider the input of those who wouldn’t agree with the projection, then I get an overall average of 88.3 projected wins. So, where we are: last week, the FanGraphs community decided the Royals are something like three extra wins better than they might otherwise appear. For the people who already think a projection doesn’t adequately capture the Royals, they see them as four extra wins better than they might otherwise appear. A good number of folks thinks the difference is even bigger.
I’m not sure how interesting this is. I don’t know how well I’m conveying how interesting I think this is. But, I mean, this is FanGraphs, with a FanGraphs audience, in which Royals fans are greatly outnumbered. Still, while there’s no clear consensus, the overall impression is that the audience sees the Royals as unusual. Usually we’re content to defer to the numbers, but the Royals have convinced a lot of people here that approach just sells them short. The relieving, the contact, the clutchness, and/or various other intangibles — people are on board. Some of them softly, some of them not, but 70% of respondents voted north of 84 wins. Royals magic now has a crowdsourced magnitude.
Of course, the Royals are going to change. The poll asked about a roster that’ll never exist again. But that roster won over a lot of people. It even won over the readers of FanGraphs.
Jeff made Lookout Landing a thing, but he does not still write there about the Mariners. He does write here, sometimes about the Mariners, but usually not.
I think that the bullpen full of excellent relievers prevents the manager from using conventional methods to determine that a lesser pitcher should be used in a high leverage situation just because it isn’t the ninth inning. I like the Royal, or any team with three or more ace relievers, to beat projections, and I think this number of wins can come close to approximation the effect of poor reliever sequencing by managers and pitching coaches.
11 wins based on coaching? I doubt it. And the Yankees’ top 3 relievers (Wilson, Betances, Miller) were just as good as the Royals top 3 (if not better) and they didn’t outperform their projections. Can’t just be based on the pen.
Defense is probably half the difference – or more.
The Yanks outperformed their PECOTA by 8 wins, and have consistently beaten their Pythag over the last several years.
Because, well, they have an awesome bullpen.
(This year was the first since 2012, when they were missing Rivera, that they haven’t)
“I can’t think of a good reason for voting below 84…”
If this is true, then the projection actually is terrible.
The whole point of this exercise (which, I agree with Jeff, may not have been properly understood by readers) was that the 84 number was unimportant. He wanted to get an idea of whether we think this team will outplay projections based on things that projections don’t capture. That is, do we EXPECT the Royals to be special? Do we think they do something that makes them win games that our numbers don’t capture?
The point was NOT to judge whether we thought that 84 was a good projection and substitute that number with our own best guess. We’re supposed to take that 84 number as the best available empirical guess, then think, “how many extra games will the Royals win just because they’re the Royals and they do Royals-ish stuff?”
Voting below 84 in this survey should have suggested that you expect the Royals to UNDERPLAY their projection, whatever that projection winds up being. That would mean you actually think the Royals have NEGATIVE “magic”, and they will play to a level that is less than the simple sum of their parts (or our best guess at the sum of the parts).
Knowing that, and knowing that the Royals significantly beat their projections and their Base-Runs/Run Differential expected win totals two years in a row, thinking that the Royals should be WORSE than their projections is a peculiar position to take. There is nothing to base that on unless you watch the Royals and say, “they could be winning more games with the exact same players performing exactly as well if deployed a different way or if they were less unlucky”. Again, that is NOT the same as saying “they should be worse than 84 wins”.
I can think of one reason to vote under the projection though, and that is that Yost is infuriatingly bad sometimes. The early-inning intentional walks, love of sacrifice plays, surely this isn’t winning them EXTRA games. One could make a well-reasoned argument it loses them some games.
“I can’t think of a good reason for voting below 84…If this is true, then the projection actually is terrible.”
I can think of as many good reasons to vote sub-84 as above 84. Assuming 84 is the mean projection, there should be as good a reason to vote for 83 as 85; for 82 as 86; etc.
I think you are completely missing the point of this poll and post.
Well, in this case it’s not a real projection, and there isn’t really any evidence that the Royals would underperform their projections in the abstract.
For the most part it’s probably people who are so fed up with the “Royals magic” narrative that they overreacted in the other direction.
But, anyone who (A) doesn’t believe in Royals magic but (B) does believe in the Gambler’s fallacy would vote this way; I’m sure there are some people out there who sincerely hold both positions and think the overrated Royals are “due” to fall flat on their faces in 2016.
@LK – Well, the Royals have generally had pretty good health, correct? And they’ve now also had two straight years of lengthening their season approximately 10% by playing through to the World Series.
So I suppose someone could have a view that a projection system will overestimate the future health of the Royals and that they are due for a rash of injuries in 2016, with pitchers and Perez being especially vulnerable.
I agree that the Gambler’s Fallacy is a more likely reason to vote that way, but there are other reasons someone could take that position.
It’s more than likely that that KC will not be as good next year, but they may fool us again. Thank you CAPT Obvious.
This is absolutely true and a great reminder that imperfect results do not necessarily invalidate a model. What makes this a more interesting question than just expected error is precisely the furor before the season even began that the projections were specifically underrating Kansas City and overrating Boston. Both were called shots by this community. That could still be coincidence and not necessarily some predictable flaw in the model, but having many people calling those failures beforehand makes coincidence and expected error less palatable explanations.
This year had a lot of misses : Indians, Tigers, Mariners, Rangers, Astros, A’s, Cubs, Brewers, Bo Sox, Royals, Nationals.
Hell the Padres were projected to win 84 games.
http://www.fangraphs.com/coolstandings.aspx?type=2&lg=div&date=2015-04-05
Every year has a lot of misses, what’s interesting is when those misses are predicted beforehand.
Every miss is predicted by some disappointed fan somewhere. And the more extreme the variance from median projected outcome, the more superficial purchase those predictions have. That’s sort of a truism.
What was missing was the thoughtful critiques of the models to accompany those predictions. Instead we get a bunch of mushy narratives based on misunderstandings of particular metrics and models and even some semantic disconnect about what it means to be “good” (specifically, whether it’s context-neutral or not).
Tough to give people too much grief on “understanding” models when the guts of the models are usually not public. People with Black Boxes shouldn’t expect a ton of trust when the predicted values diverge from reality.
*facepalm* The FanGraphs projections for the Kansas City Royals and the Boston Red Sox were hotly disputed by FanGraphs commenters, and some of those did give thoughtful critiques as to why they felt the FanGraphs projections were flawed. “You’re biased against my team!” complaints can be dismissed out of hand, even if that team ends up being better than expected, but criticisms of the methodology deserve more examination.
Blue, true that it would be better if the guts were public. However, if the authors can show us that the black boxes are with in X distance of the true outcome Y% of the time, we don’t really need to know all the details to assume that rate will hold up.
Jog my memmory JD, what were thoughtful critiques?
Sandoval was projected for his best offseason since 2011, after 3 years of decline.
The projections for Hanley didn’t account for the impact of the position change.
The low, low floor of the non-Miley members of their rotations was ignored.
People were saying this stuff in March.
And everyone thought in the preseason that Padres projection was too LOW. Looks like even while being wrong it was still better than popular opinion.
That’s one of the difficult aspects of making projections.
We’re guessing what a group of people will do in the future, and we’re not even assured that the group will contain the same people a month from now.
Projections necessarily assume that we know about how is going to be playing in the future, and they assume that these players are going to progress in their careers about how similar players have progressed in the past.
They’re not too good at guessing when an entire pitching staff is going to fall apart and give up about a million homers.
It’s very possible for the projections to have been too low and STILL to have been closer to the final result than popular opinion.
For an extreme example, suppose a projection system had the Packers winning 8 games and finishing in third place. That’s too low. Then suppose Aaron Rogers breaks his leg in the preseason and misses the whole year, then the Packers finish 4-12. The fact that something unlikely happened to ruin the final record doesn’t NECESSARILY mean the projections weren’t too low.
It’s almost like we are not good at predicting the future.
“Two quick things first. One, this would’ve been best done with money on the line.” Vegas itself doesn’t have any sort of serious favorite. The Cubs are tops at 11-1. The best team in baseball is hardly distinguishable from mid-pack.
1) Vegas doesn’t work on projections or trying to actually predict anything.
2) We’re actually pretty good at projecting the future. It shockingly looks worse when we specifically choose to look at an outlier.
There are two different questions here. One is whether the Royals winning 95 games was a fluke, the other is whether the Royals doing so well in the postseasons the last two years is a fluke. I am inclined to say that the Royals benefited from a certain amount of luck to win 95 games, maybe 90-92 would have been more what should be expected, but in terms of their postseason performance they have been the team best suited to doing well in the postseason, so this isn’t a fluke. Deep relieving and contact hitting is a bigger factor in the postseason than in the regular season.
In the 5-game series, they were trailing in the 8th and/or 9th innings in 4 out of the 5 games. They won three of those four games. According to Baseball Reference, MLB teams do that once out of every 2,820 times. Of course there’s a good level of fluke involved!
If you want to say they’re better equipped for extra innings or tight games because of their bullpen, then that’s fine. But let’s face it – *no* team is suited for 9th inning comebacks against an elite closer on a consistent basis. Even if they’re *better* equipped for it, they’ll still lose a lot more than they win in the grand scheme of things.
I’m curious if the likelihood is similar in the postseason.
The Royals did have a very strong bullpen, and their opponents had tired arms from playing all those extra innings, and this was at the end of a long season and a couple of difficult series against strong teams. In the regular season, you need a more conservative approach to using your relievers. Maybe the postseason takes a toll on relief arms and makes them more likely to give up hits (against tougher competition than what they face in the regular season, too — but that might be mitigated by postseason teams having better relief pitchers, maybe).
I’m not saying this is the case; it would need to be demonstrated with data. It’s plausible, at least, and it should show why we can’t just count outcomes in the regular season to determine postseason probabilities.
Of course it’s still luck on their part, but it’s probably not on par with surviving a trip into an asteroid field to evade the Empire.
Never tell me the odds.
Well then why isn’t it true of the Royals’ relievers too? They gave up 1 run in the 7th inning or later. Shouldn’t they also be tired?
Exactly stathead.
This is all just more #narrative. The only way we can find out who has this “very strong bullpen” is to see if they survive October. The Astros had a better bullpen in the regular season by WAR. The Cubs were tied with the Royals and the Dodgers weren’t far behind. Why didn’t they get this same bump? Only because they didn’t win it all, so we aren’t talking about them.
This is silliness. You’re telling us its more of a fluke to do well over 162 games than over at most 19? I get that playoff baseball is slightly different from the regular season, but the gross difference in sample size just can’t be over come.
Beating the regular season projections by a bit over a standard deviation once is lucky. Twice is very lucky.
Three time in a row, with the third time by about four standard deviations, is a little too lucky for comfort.
However, that’s mitigated somewhat (only somewhat) by the fact that we’re effectively cherry-picking our results by looking at the team that just won.
On the other hand, we’re looking at one of the teams that beat its projections without a clear, obvious reason for it to have beaten them. This isn’t the same as the Astros getting a bunch of good young players or the Mets and Blue Jays making deadline trades, and the Royals have been doing it for longer.
Sure, it’s possible for a team to finish four standard deviations above its projected level after destroying projections two years in a row, but it’s really, really unlikely, especially in an era where at least one other team used a similar philosophy for its roster and had a run of success that also included beating projections (though not by nearly as much).
There’s a lot that needs to be investigated here, even though the answer is probably that the Royals had some players who played poorly long enough for the projection systems to think that they were awful. Then they got better, and the team played Moneyball for a few years.
Actually, the Royals did make a couple of deadline trades. Do you think they get to 95 wins playing Omar Infante instead of Ben Zobrist? I doubt it.
Perhaps, the trade deadline allows a team to beat the projections consistently. If they’re good because they’re getting lucky (like the 2015 Twins), they have a chance to become better by trading players, and then they play at their real level, which is closer to the previous state. Maybe that’s what happened with the Royals this year.
The post season has brought a lot of attention to the advanced scouting that the royals do. I don’t know to what extent other teams do this but the comments regarding Prices pickoff move, bautista always throwing to second etc are really interesting. Rusty Kuntz said they scouted david wright and duda and they had even practiced a similar play to the duda/hosmer play earlier in the season and he couldn’t believe that it actually happened in the world series
I would like to think teams always do their due diligence with advanced scouting…but then I’m reminded of that quote from Alfonso Soriano a few years ago saying that he had never been given any instruction on how to play the outfield, despite being signed to a 9 figure contract to play outfield
Suggestion: no more single point estimates for projections, of any kind; always publish the standard deviation for your predictions. This will, real quick, start imposing some restraint on claims made about those predictions.
I’d also suggest that the underlying models be presented with underlying variance explained (R-squared NOT RMSE) as well as regression diagnostics. Again, this would show the relatively modest precision that underlies existing projections and hopefully cause people to up the game in the underlying models.
There are deep, underlying problems with current, public projection systems as currently implemented but two in particular are holding public analytics back: 1) The link between season aggregate statistics and the ways W-L are actually generated (as a summation of discrete contests) is seriously underexamined. 2) For some odd reason existing modeling uses a sledgehammer in forcing regression to the mean for projections and then adds particularistic additions to this projection. This is in contrast with standard modeling approach which would focus on identifying orthogonal variables in the existing data sets (via PCA or cluster technique) and then try to figure out how much of the underlying variation in the datasets can be explained using a concise set of orthogonally-situated variables. In proper modeling procedures regression to the mean happens naturally and is not imposed by brute force.
The Royals (and the Giants and Angels before them) pose analytical challenges to be addressed; they are not flukes to be dismissed nor evidence of some supernatural and unknowable phenomena. A lot of the pushback is coming from people who either know, or sense, the underlying lack of rigor in existing models and how that contrasts with the grand claims made for the results generated from these models.
My biggest problem with the claims that the Giants/Angels/Royals/Orioles are not flukes is that there is no viable hypothesis that explains them other than the fact that they were the team to end up on that end of the distribution. Every year a team underperforms by just as much yet no one complains; it’s simply the distribution.
In terms of the hypothesis, everyone seems to think effective bullpen management is a huge inefficiency, but the Royals got over half their clutchiness from their hitting, which is fundamentally impossible to clutch. Being a clutch hitter means you don’t try 90% of the time…
The main problem with that is that doing that much work usually entails getting paid. The skillset to do that is worth a lot of money in business.
When you put it that way, it sounds fun, but that’s leaving out the long hours of obtaining and cleaning the data. That’s hard to stomach even if you’re a baseball fan. Every analyst and data scientist loves the 5% of the time for doing the actual analysis. The dirty work has to be done first, though.
That dirty work is why you rarely see THAT much rigor in this kind of analysis done freely on the internet.
On the other hand, it’s not too terribly hard to see how imprecise these models MUST necessarily be just by taking thirty seconds to do some calculations about coin flips, and the variance I get when doing that isn’t too far off from the variance I get when I simulate 100,000 baseball seasons and analyze the results.
Long story short, even with perfect information about the probability of a team winning each game, we could still see a two dozen game swing in results without anything funny going on. There’s nothing inherently shocking about a .500 team winning 90 games. Or 72 games.
Then again, a .500 team winning 90 games once is easy to explain. Several times in a row? That’s harder. Several times in a row and it happens to go against the grain of conventional wisdom? Billy Beane got rich by doing that.
So, I don’t think we can conclude anything until someone with too much time on her hands gets around to doing the analysis.
Yeah, it’s definitely true that data cleaning (and flat out basic data understanding) is the true work in any analysis (and it is amusing because so little time is spent on it in formal education). For everyone of the major projects I’ve ever done the cleaning was measured in months and the bulk of the analysis was generally done in a week or two.
And you’re absolutely right–people generally get paid for that level of work (or they do it as a dissertation or very sparky master’s thesis)…and I’d bet people ARE being paid for that work right now. It’s just not public.
Which is why I just thing people rolling out the simple, public models just need to act with a whole lot more humility than they do when assessing moves by ballclubs at this point in time. Sure, maybe they are being boneheads. Or maybe they’ve discovered some signal in the data and are acting on it.
+1
I’ll link to this every time some meatball replies “where’s your better model, smartass?” to a thoughtful comment.
I’m pretty shocked my answer of 85 was not more popular. I think, objectively, it is the most defensible answer, maybe 86 if you want to be less conservative. 87 might be defensible as well I suppose, though three wins is a huge effect.
I think it’s clear that there are probably some effects not captured by the projection systems involving contact hitting and elite fielding, things like that. There are been new studies along these lines. So I think it’s fair to expect the projections underrated the Royals a bit.
But the idea that this effect amounts to 7 wins or more – LOL.
I think this is reasonable. It seems fairly intuitive to me that elite bullpen arms should strand more inherited runners and thusly allow a team to give up fewer runs than projected wOBA or Base Runs would predict. But as you imply, the effects of these things are likely pretty marginal.
I’d like to address your post with data, however the only links I am findings to Fangraphs preseason projected standings for previous years go to the currently updated in-season projected result.
I did find that the Royals projected wins result for 2013 was 80 games. They ended up six games over that amount. They were at least six games greater than projected in 2014 (given that they won 89 games probably more). They were *16* wins off this year with 79 instead of 95.
Not sure how picking 6 or more wins above projection is unreasonable when that’s precisely what the Royals have done for three years in a row.
The extent to which those 6 games mean anything is exactly the question.
The biggest thing I could see that would screw with the projections is that the Royals don’t behave like your average team with respect to pitcher usage. Lowest in the American League in starter innings pitched.
I would be interest to see if the projections are smart enough to understand that the team expects to get five innings out of underwhelming starting pitching instead of seven. If the projections are saying ‘you are relying on getting quality starts from a bunch of cruddy pitchers’ when the game really is ‘have those cruddy pitchers keep us in it for five innings then go to the dominant bullpen’ you could see a reason for a disparity.
I guess the question is whether or not projections take into account specific team behavior, or if they work on the model that every team tries to get as many IP out of their starters as possible.
The team projections are running off the play time allocations that FGs staffs parse through “manually.” So if the projections assumed too many innings by mediocre starters, that’s a flaw with FGs authors/staffers, not the robots.
It would be very easy and somewhat interesting for FGs to compare the actual play time allocations with its pre-season projections and calculate some kind of match rate.
I’m not sure it’s that simple, Ernie. By limiting a pitcher to 5 innings instead of 7, you’re making him a better pitcher on a rate-basis. With the ‘times through order’ penalty, you’re using him at his best and avoiding him at his worst.
That won’t come into play with the depth charts. Whether you give someone 150 innings or 170 innings, it’s all based on the robot’s projection (which is held constant). In reality, the robot would need to bump up the projection if it knew the pitcher wouldn’t go through the order as many times.
And that’s a good point by Ullu. For most teams, it’s not worthwhile to limit SP innings as much. Someone like Harvey, Lester, etc, are still better the 3rd time through the order than most teams’ 6th inning reliever. The Royals are in a unique position where their 6th inning options are as good or better than most teams’ 8th inning options. So they can extract the most value from mediocre starters, then replace them with a solid middle reliever.
This is an excellent point that I completely overlooked.
That said, I doubt this explains much of the Royals’ “over”- performance relative to the models this season. The Base Runs analysis suggests that sequencing mattered far more than, say, the effect of usage on reducing wOBA-against.
Actually, it might not be quite as good a point as I initially thought. The Royals rotation averaged 24.2 batters-faced per start, compared to 24.7 for the AL overall. Maybe that difference reflects real benefit by reducing aggregate time-through-the-order-penalty, but we’re probably not talking about a very big difference.
Don’t think it’s particularly hard to make an argument that a team with a deep bullpen is going to out perform projections by a couple games. Possibly defense too although I think it’s more obvious that reliever chaining is a clearer problem. Can’t imagine I would project the royals for more than 2-3 games better than the projections though.
Random aside- most of us know that managers don’t really matter much and the vast majority of managers are just bad equally and make the same mistakes over and over, but I wonder whether we should be giving teams about a 2 win bump from having one of the few competent managers in the game. (note- i don’t include YOST in this obv)
It has never been shown that managers are bad “equally”. It has been shown that managers aren’t the source of a ton of extra wins, however people apply the converse to that (managers aren’t the source of a lot of extra losses) which had most definitely NOT been shown to be true. There are also a lot of reasons to believe that poor managerial choices can hurt a lot more on the downside in many different ways–for example, Trey Hillman’s absolute abuse of Gil Meche blew up his arm and cost the Royals many games over the last three years of his contract. Poor tactics are much more likely to lose a game for a team (Grady Little, Terry Collins keeping in their SP too long) than managerial brilliance can squeeze out wins. While team chemistry is tough to quantify it’s reasonable to assume that managers can also poison the clubhouse in such a way as to generate additional losses.
Fair enough. I mean that imo probably 20 MLB managers are equally bad/standard with bunts/bullpen stuff, etc
You offered 7 options greater than the projection. When you do that, you’re likeliest to see a mode of 84 but you’re also likely to see a higher proportion of people pick options greater than 84 just because of poll design.
Yeah this was an exercise in poor pool design.
I thought the whole idea behind Sabermetrics was to find a better way to compete and analyze a situation or industry. To me, Sabermetrics has always been a fancy word for the combo of strategic planning thought and economics. Those disciplines have been around forever in corporate America, but not always in MLB to the extent that modern baseball has it now. Essentially, it boils down to if you don’t like the status quo or hearing the words, “well, that’s the way we have always done it” then you blaze your own path using statistics and math. Sometimes you may have to parse your own data to the granularity that you need in order to express the way you think about the game. This leads to the development of new fields inside databases and new formulas or algorithms, that hopefully prove your perspective right and you are able to justify the financial decisions made to better a ballclub, company, or organization. To that end, is it so hard to realize that what the Royals did with defense, speed, and especially their offense, won a World Series?
On offense, its not about batting average, OBP, SLG, or OPS, its about the value proposition of plate appearance outcomes. PA = backward K’s + K’s + PO + HBP + IBB + BB + fb + GO + FO + LO + PE + ME + 1B + 2B + 3B + HR, from left to right is the most negative of outcomes that do not help the individual player or his team, to the most valuable of outcomes. To reduce the PA outcomes to make it simple to understand it would be, PA = SO + W + CO + ROE + H, meaning plate appearances = strikeouts plus walks plus contact outs plus reaching on an error plus hits. But there is a trade off or equilibrium of sorts between WALKS and STRIKEOUTS vs. the HR and DOUBLE vs. the SINGLE and CONTACT OUT.
Walks and strikeouts, together as a group have very little value precisely because 1) strikeout have no value at all, 2) walks don’t get the individual who walked into scoring position, and only move other base runners one base at a time. Walks can score runs, but they don’t drive them in, 3)Contact outs can move runners forward on the base paths and garner RBIs, but cannot score a run, 4) hits are the highest value proposition, can drive in more than one RBI at a time, does not get you or your base running teammates out, you can still score a run yourself, and move other base runners more than one base. The Royals propensity to choose the contact out, single, and error to score their runs over HRs, plus their speed helps out to take the extra base, go first to third, second to home. It would stand to reason then, that the less walks and strikeouts you have as a team, the rest of the PA outcomes are all contact related. So they had higher contact and a materially higher value proposition. Its why I created HEWCO, BSM, and CCR to evaluate all the good batters do while hitting. The Royals of 2015, are a 1980’s team, just like in 1985 when they won it all. In fact, the 1980’s in terms of offense are kicking the collective keisters out of modern era offenses in total runs. Its nothing new, the more your plate appearances end with a walk or strikeout the more you need to hit for extra bases to make up for it (doubles and homers), but you do so at the expense of consistency and overall contact.
Average Team Runs and RBI Gained/Lost
Year Runs RBI
1980v2010 -16 -25
1981v2011 -46 -53
1982v2012 -4 -12
1983v2013 24 14
1984v2014 31 20
1985v2015 12 4
w/ 81-11 1 -52
w/o 81-11 47 1
81 was a strike year with only 2/3rds of games played so probably not valid.
Average Team Differential: Hitting Components Gained/Lost
Year Hits 2B 3B HR XBH 1B CO + E
1980v2010 49 -39 13 -35 -61 110 377
1981v2011 -5 -47 8 -48 -87 82 374
1982v2012 46 -32 6 -35 -61 107 411
1983v2013 37 -25 14 -28 -40 77 358
1984v2014 51 -32 10 -14 -37 88 359
1985v2015 11 -28 6 -25 -47 58 381
w/ 81-11 189 -204 57 -186 -333 522 2260
w/o 81-11 194 -157 48 -137 -246 440 1886
(the 1980’s winning the war would be all + numbers, losing the war for the 1980’s would be a negative number in the above comparison tables) Keep in mind there was 26 teams in the 80’s instead of 30.
I concede that fielding defenses in the modern era are way better than the 80’s and that speed is better in the modern era in terms of percentage but not in raw number of bags swiped, nor in my metric of BSM, bases moved on offense. Pitching was better in the 80’s and so was hitting production (average and total runs). Making the 1980’s the last great decade of MLB.
The top 18 in HEWCO adj for 2015
Year Player Team Pos HEWCO adj
2015 Altuve, J HOU 2B 800.987
2015 Pujols, A LAA 1B 790.063
2015 Arenado, N COL 3B 772.195
2015 Machado, M BAL 3B 755.779
2015 Kinsler, I DET 2B 742.025
2015 Pollock, A ARI CF 741.459
2015 Seager, K SEA 3B 740.138
2015 Betts, M BOS CF 739.082
2015 Donaldson, J TOR 3B 735.213
2015 Fielder, P TEX 1B 735.119
2015 Cabrera, M CWS LF 734.289
2015 Posey, B SF C 721.309
2015 Rizzo, A CHC 1B 718.383
2015 Beltre, A TEX 3B 714.177
2015 Cespedes, Y NYM LF 706.308
2015 Escobar, A KC SS 704.346
2015 Cano, R SEA 2B 703.742
2015 Andrus, E TEX SS 703.063
Year Team HEWCO adj Walks Strikeouts Total W+SO
2015 Kansas City Royals 6241.828 460 973 1433
2015 Toronto Blue Jays 6106.373 624 1151 1775
2015 Boston Red Sox 6025.071 524 1148 1672
2015 Oakland Athletics 5971.393 515 1119 1634
2015 New York Yankees 5916.844 617 1227 1844
2015 San Francisco Giants 5844.299 506 1159 1665
2015 Arizona Diamondbacks 5752.108 523 1312 1835
2015 Texas Rangers 5751.921 579 1233 1812
2015 Cleveland Indians 5735.847 572 1157 1729
2015 Colorado Rockies 5732.093 421 1283 1704
2015 Detroit Tigers 5727.997 496 1259 1755
2015 Cincinnati Reds 5713.468 538 1255 1793
2015 Los Angeles Angels 5705.17 493 1150 1643
2015 Miami Marlins 5644.604 414 1150 1564
2015 Pittsburgh Pirates 5635.655 550 1322 1872
2015 New York Mets 5608.035 556 1290 1846
2015 Atlanta Braves 5599.622 515 1107 1622
2015 Seattle Mariners 5580.073 514 1336 1850
2015 Chicago White Sox 5578.81 469 1231 1700
2015 Los Angeles Dodgers 5550.47 623 1258 1881
2015 St. Louis Cardinals 5547.337 572 1267 1839
2015 Minnesota Twins 5545.811 479 1264 1743
2015 Philadelphia Phillies 5512.999 441 1274 1715
2015 Baltimore Orioles 5512.981 469 1331 1800
2015 Tampa Bay Rays 5492.093 520 1310 1830
2015 Houston Astros 5485.659 542 1392 1934
2015 Milwaukee Brewers 5454.792 453 1299 1752
2015 Washington Nationals 5399.111 583 1344 1927
2015 San Diego Padres 5378.377 466 1327 1793
2015 Chicago Cubs 5122.6 641 1518 2159
169873.441 15675 37446 53121
5662.448033 522.5 1248.2 1770.7
I just find it funny that the Royals took a page out of the 1980’s when every other team is playing the three true outcomes (I consider it 4 true outcomes and count doubles in there) game and mindset. Who better to dip into the 80’s than the Royals, since they won it all in 1985 using basically the same concept? The pendulum has swung too far on TTO, look at the average team runs and wins table comparing the 80’s to the 2010’s. I know of 5 teams from 2013-2015 that would love 2.5, 3, and 1 more win respectively in those years (that is, if sabermetrics has it right and 10 runs = 1 win)
Not sure if this has been discussed – it was an op-ed about luck in the Cardinals’ 2006 postseason run:
“The World Series does not establish the best team; it just compacts 162 games into seven or so frenzied ones. This lottery nature is what makes it so exciting. Ya gotta believe, because you never know when your number’s coming up.”
Birds of Pray
http://www.nytimes.com/2006/10/18/opinion/18leitch.html?_r=0
I realize that as a Mets fan, I’m biased – so take this comment with that large grain of salt. But I think this was a much harder fought World Series than many people are giving the Mets credit for. Statistically, it does seem like the Mets regressed to the mean in too many areas to expect they could win the Series. But they had a lead in every game and even took two leads into the ninth inning. To me that is a bright spot to build on for the Mets team next year and can’t be explained away as the Royals somehow “spotting” the Mets a lead to set up their relentless comeback.
What I’m talking about is that elusive judgment about errors and luck when the ball leaves the hitter’s bat or reaches the opponent’s glove. I’m curious about the ongoing statistical analyses that will come out of this World Series, in particular whether the Royals’ approach to hitting means they put more unpredictable balls in play, especially in late innings, that were harder for opponents to field, and how much that has to do with the skill of Royals hitters vs the luck that can happen when a ball is put in play.
Also, as you cover in the Murphy’s error article**, there is a reason earned runs are not awarded for errors. In many cases, an error amounts to giving the other team an extra out or two. So even though one can easily argue that an error-prone player makes that team “worse,” MLB rules still do not give the “better” team credit for earning a run they were allowed to get as a result of that mistake.
In World Series Games 1, 4, and 5, the Royals won on unearned runs. Of course, that is not the same as winning a basketball game on free throws, but it’s inconvenient to the narrative I’ve heard that the Royals “thrashed” the Mets. This World Series really bothered me not only because I’m a Mets fan but also because I want to be honest with myself about any bias I have, and yet I can’t shake the feeling that the Mets gave away the Series as much as the Royals took it.
**
http://www.fangraphs.com/blogs/daniel-murphy-and-the-costliest-errors-in-world-series-history/
Couple of comments; I don’t think I’ve heard anyone say that the Royals “thrashed” the Mets–anyone who says that didn’t watch the game.
But unearned runs don’t just happen–it isn’t like the error meteor hit the stadia each night. The Mets made a conscious choice to favor a set of skills based on offense over a set of skills that favored defense (as the Royals did). Those choices have consequences and, in these games, the consequences for the Mets from going stick-first overwhelmed the benefits.
And that happened specifically because the Royals are such a contact-heavy team.
I think the Mets would definitely have beaten the Blue Jays, whose offense doesn’t revolve around forcing the opposing middle infielders to make plays.
Thanks – guess I just read too many comments on the game boards written by Royals fans.
I see what you mean about the Mets tradeoffs. True.
And you’re right – you articulated what I couldn’t before – the Royals contact hitting was strategic and paid off, even when the outcome was expressed versus the weaker defense that the Mets had. I knew Murphy has always been a less than stellar fielder. But even in the case of that throw from Duda, a decent throw gets Hosmer but the pressure by the Royals forced Duda to get such a throw off. Makes sense based on the Royals’ aggressive strategy that Duda blinked, as it were, and was forced into a throw he was less likely to make in that situation.
I don’t think that the Mets gave it away so much as the Royals had a great matchup in terms of roster construction.
The Mets give away defense for offense on the right side and up the middle, because their pitching staff is built around preventing hard contact on the right side and up the middle.
It’s a smart strategy in general…but the Royals entire offense is built around hard contact on the right side and up the middle.
I love the graph. It very clearly depicts that the vast majority believe the Royals will outperform their projection.
You mentioned the Angels and this KC team kind of reminds of 2002, great defense, speed & contact, shutdown bullpen (Shields, Frankie, Percival).
But I remember thinking, “I hope they don’t re-sign all these guys, there were some career years (or a career month).” They did and mediocrity ensued. That’s how I feel about the Royals, they’ll fall in love with some guys when they should move on, the GM self-convinces that he has the secret sauce and gets narrow minded, and some winners naturally relax their effort. They’ll settle into just another team. (That’s the easiest prediction, take the field).
I would like to suggest that another reason for all the 84-and-lower projections is plain old sour grapes: “Well, you Royals just won the World Series, but I still say you’re just an 84-win team. So there!”
Ahh, now that I’ve gotten that off my chest . . . There are three elephants in the room that all this hand-wringing and invective are missing. The first I’ve already ranted about enough: human activity is not 100% quantifiable, and it’s folly to pretend that it is. The second: you’ve missed the boat on the Royals for three straight years; this year in particular the projections were wildly off. I am not a numbers guy by any stretch of the imagination, but neither am I a fool, so don’t talk to me about “small sample sizes” or use the magic words “random” or “luck”: we are talking about 486 regular-season games plus five rounds of league playoffs (counting last year’s wild card) plus two World Series. How big does a “sample size” have to be? And, as I’ve already pointed out in another post, in this context “random” and “luck” tell me no more about what’s really going on in the real world than “magic” and “destiny” do. Bottom line: if your models have failed to account for the reality they are supposed to elucidate, then it’s time to tweak your models. No need for hand-wringing or getting angry or giving up sabermetrics entirely.
Finally, in these discussions we all seem to have lost sight of the whole purpose of playing baseball in the first place: to win games and championships. The Royals have accomplished this most basic of goals. The whys and hows are secondary to that–interesting, intellectually challenging, but secondary. On this most basic of levels, our debates are really irrelevant, however important they may be to us. In other words, there seems to be a disconnect between our analyses and what we are analyzing. This, in the end, is what I think has set people off: analysis that suggests why a team won or lost is one thing; analysis that purports to explain why a team shouldn’t have won as many games as they have three years running is just ridiculous.
I suspect the lower than 84 answers are some variation of the Gambler’s Fallacy–that the “extra wins” will be balanced out by “extra losses” in the long run.
Let me clarify my last comment: It is, in fact, interesting to consider analysis that suggests that a team has wildly out-performed its talent, based on projections that the sabermetric community considers valid. The problem has been privileging the projections over the actual outcome, and suggesting that: the actual outcome is somehow fraudulent/the Royals are undeserving/their fans are poop-heads, etc.; there has been too much of this, and THAT is what is ridiculous. However, I do stand by my contention that there is a disconnect between reality and analysis here that should be addressed.
“to win games and championships” – the Royals definitely won more games this year than last year, but look at this:
The 2014 Royals won 89, lost 73. They were a wild-card team and almost ran the table until SF (another wild-card team) narrowly stopped them.
The 2015 Mets won 90, lost 72 – i.e., they won 1 more game than the 2014 Royals did. Then they beat the Dodgers and Cubs including pitchers Kershaw, Greinke, Arrieta, and Lester.
So this year vs last, the Royals won 6 more games, the Mets 11 more games. In terms of wins, the Mets improved almost twice as much as the Royals did.
The Mets did lose the 2015 Series in 5 games, but in two of those games the Mets took a lead into the ninth inning. Sure, the Mets fell short, succeeding more in the “win games” department than in the “championships” department. But after reviewing these past two seasons I really think the Royals are a bit over-hyped. The Mets held a lead in every game and fought 4 out of 5 of them really hard.
I admit I am a Mets fan so was extremely upset to see the Mets lose and it’s no doubt clouding my judgment, but I do think at least some of my thinking is justified. It annoys me at the way many Royals fans on comment boards are characterizing the World Series as a thrashing or how the Mets were never in it, or how the Royals are just “better.” If we define “better” by World Series wins, then the 2014 SF Giants were better than the Royals, and the 2015 Royals are better than the Mets. If you go by other standards, such as the ones in this article about how the Royals had the best postseason of any team ever without winning the World Series, well that is another set of criteria. Yet I see Royals fans saying “Gordon should have tried to score” etc. etc. – my point being a lot of the commentary amounts to paining the Royals as “better” in both years by switching criteria back and forth.
http://www.si.com/mlb/2014/10/30/world-series-royals-game-7
I think what we’re seeing with the Royals is real, but as a few others have written here, may be more a strategic shift than a question of something magical about this Royals team. Swinging a lot and assembling an aggressive team built around great defenders and excellent relief pitching seems to be disrupting the previous model of super-strong starting pitching and a mix of selective hitters with home run swingers.
The Royals sure seem to be ahead of the curve on this, and at the moment they are executing it better than other teams, but it’s not like they didn’t have their fair share of struggles even with this much-praised strategy – Houston took them to 7 games in the ALDS and Toronto to 6 in the ALCS. I also just went back and watch Game 7 of the 2014 World Series. Although it is not easy to say hey, just have a Bumgarner on your team and you can beat the Royals, it sure seemed like he had their number.
2-0, 0.43 ERA, 21 IP, 1 SHO, SV
http://www.si.com/mlb/2014/10/30/madison-bumgarner-giants-world-series-mathewson-burdette-gibson
I’ll be very curious to see if they can do this again next year with the same approach. I have my doubts, if only because we may see other teams copy this style as well and disrupt the strategy landscape even more. If the Royals use this style to reach the World Series again, I’ll be convinced that they’re dynasty-level/magic. Until then I see them as being early adopters of a style that other teams will either adopt themselves or will solve and defeat.
** painting the Royals
Grr sorry ALDS was 5 games. D’Oh!
For as many comments as you have read about the Royals thrashing the Mets, I have read as many about how the Mets gave the series away. It is as if when the Mets take advantage of Royals’ miscues that’s part of baseball but when the opposite occurs that’s the mets giving it away.
In game 1 the Mets had a late lead because Eric Hosmer booted a ground ball. If Hosmer makes that play the game goes to extra innings even if Gordon doesn’t hit a ninth inning homer.
In game four the two batters after Murphy’s error produced base hits. Even if Murphy makes that play the Royals eventually take the lead.
In game 5 the Royals don’t need Duda’s bad throw to tie the game if Hosmer doesn’t make an error that leads to the Mets’ second run and Rios doesn’t forget how many outs there are.
Both team made mistakes. But the Royals made less of them and fought back from the ones they did make. That’s the difference between winning and asking “what if”.
“One, this would’ve been best done with money on the line. As is, there was no consequence, no downside to selecting an option you didn’t really believe.”
SO MUCH THIS. For anyone who voted 91+, when the Vegas over/unders (or PECOTA projections, or whatever) come out next year, I would honestly like to lay money against you on KC beating their forecast win total by 7+. Even money of course since you voted for that as the most likely outcome .
Does Fangraphs have money on the line for its projections? The site would probably have lost a lot of money this year if it took over-under bets on its pre-season Royals projected wins total.
??? You realize that when you are talking about projections made yearly for all the teams, talking about the projection for one team is kind of asinine?
Flyball pitching, defense, deep bullpen, coaching to strengths/development of team philosophy, player retention and impact on team building, momentum, farm depth for trades/call-ups. All things that the Royals seemed to have at the beginning of the season that are undervalued by the projections or are just unquantifiable.
+1 for the most concise summary I’ve seen so far.
It’s probably a little higher because Chris Young always pitches better than his peripherals, the bullpen is probably a lot of added value. The defense might also be more consistent than the projections suggest. All of this might add up to 4 wins of value, which sounds about right. I wonder if this is a case of the average of a large group of people guessing right about the number of jelly beans in a jar.
Two reasons off the top of my head why people might have chosen less than 84 wins:
1) Straight-up Gambler’s Fallacy. They beat their projection these last two years, so they’re “due” to fall short now, right? Right?
2) Pretending they DO have money on the line! If this poll were a betting pool, and I knew or guessed what it would look like, I would absolutely bet under 84 wins if I trust the projection at all: It’s every bit as likely as the Royals beating their projection again, and I’d stand to win a lot if they lose the virtual coin flip and have a down year instead, because “my” winnings would be divided only with the precious few others who bet the same way.
It is sad that even on a site like this over half the commenters don’t seem to understand Stats 101.
Given the comments about baseruns not completely accounting for a strong relief staff, out of curiosity I ran a bivariate regression on the 2015 team stats for all mlb teams to see what the relationship, if any. was between reliever’s WPA (which does account for leverage) and the gap between estimated wins from baseruns and real wins. I used the gap between real and baseruns wins as the y variable and reliever’s WPA as the x variable. For example, the Royal’s relievers posted a WPA of 9.8 and the Royals won 11 more games than baseruns said they should given their overall stats.
The correlation coefficient was .6. The slope of the best fit line was .932 with an intercept of -1.86. For a team like the Royals with a fantastic relief corps about seven of the 11 wins above their baseruns estimate could be explained by this model.
I haven’t really thought through whether using WPA is a reasonable way or the most accurate way to reflect the impact of reliever leverage but I though I’d throw this out there for comments to see where I may have messed up.