The Orioles Are Better Than We Thought, Again
There’s a certain urgency to a post like this. Yesterday, I talked about the Cubs, and I had little choice but to mention the Cubs probably won’t win the World Series. The Orioles, one has to figure, are worse than the Cubs, so the Orioles probably won’t win the World Series, either. The odds are strongly against every individual team, meaning fans of every individual team are likely looking ahead to crushing heartbreak. If and when that heartbreak occurs, it’ll be a little while before people want to reflect upon happy memories.
So instead of waiting, I want to slide this in today. For all I know, some hours from now, the Orioles’ 2016 season will come to an end. They have something like a 50/50 shot to move past the Blue Jays, and then they’d just be rewarded with another tough match-up against another tough roster. The playoffs are hard and the playoffs are draining. But no matter what happens soon, it was another strong season for Baltimore. And it was another strong season that the projections didn’t expect.
The Orioles won 89 times. That wasn’t enough times to win the division, but it was enough times to get them to the Fourth of October. They’ll play another team that won 89 times. But there is one big difference here. Just before the start of the season, we had the Blue Jays projected to win about 84 times. We had the Orioles, meanwhile, projected to win about 78 times. Believe it or not, the Orioles had our second-lowest American League projected winning percentage. I know that doesn’t make us look great, but better to embrace your flaws than ignore them.
The Rangers were the team that most defeated the projection game — they cleared their projected win total by 16. Now, the way the Rangers won would’ve arguably been impossible to project, but, whatever, the season’s complete and wins are wins. The Rangers beat their projection by the most. The Orioles beat their projection by the second-most. And if you’ve been paying attention, this isn’t even necessarily anything new.
We can make things really simple! Leaving aside any issues of magnitude, this was the fifth year in a row that the Orioles won more games than they were projected to win. Five years in a row, the Orioles have been projected to finish at least a hair under .500. Five years in a row, the Orioles have finished at at least .500. Let’s say a given team has a 50% chance of beating its projection, and a 50% chance of falling short. It’s a bit of an over-simplification, but it’s good enough. To beat your projection five years in a row: 3.1% chance of happening. Slim chance. And the magnitudes matter.
I’ll show you what I mean. There’s not really anything special about examining specifically the last five seasons, but that’s what I’ve done. Here are five-year preseason projected win totals:
The order of teams shouldn’t be shocking. We’ve all observed and discussed these projections. But go ahead and search for the Orioles. They’re sixth-worst, between the Marlins and the White Sox. Teams go through cycles and there are ups and there are downs, but over five years, on average, the Orioles have been projected to be fairly lousy. Here are five-year actual win totals:
The Cardinals are first. They were fourth in the projections. The Nationals are second. They were first in the projections. The Dodgers are third. They were second in the projections. The Orioles are fourth. They were 25th in the projections. I don’t think you could say that in any given year recently, the Orioles have been dominant. But over this five-year stretch, the Orioles have been almost as successful as any other organization, after having been projected to be pretty bad. I genuinely don’t know what that means, but I know that it has to feel good.
This plot is the difference between the above two plots. This shows the difference between five-year actual wins and five-year projected wins.
However you want to examine this depends on how you frame it. I’m not saying the Orioles have definitely been the trickiest team to project, but they’ve been the most underrated by the projections over the last five years. And by, you know, a ton. Shown here is a difference of 58 total wins. The Pirates are in second with a difference of 35 total wins. The Rockies are in last with a difference of -34 total wins. It’s the Orioles who stand out by a mile, and, say, I mentioned yesterday I have projections going back to 2005. Here are the most exceptional five-year stretches, in terms of actual wins – projected wins.
| Team | Years | Difference |
| Orioles | 2012 – 2016 | 58 |
| White Sox | 2005 – 2009 | 44 |
| Pirates | 2011 – 2015 | 42 |
| Angels | 2007 – 2011 | 42 |
| Angels | 2005 – 2009 | 37 |
| Orioles | 2011 – 2015 | 37 |
| Marlins | 2005 – 2009 | 36 |
| Rangers | 2008 – 2012 | 36 |
| Pirates | 2012 – 2016 | 35 |
| Marlins | 2006 – 2010 | 34 |
It’s the Orioles, again, by a bunch. They were most exceptional in 2012. The big projection-defeating years have been 2012, 2014, and 2016. The 2012 – 2014 stretch was nuttier than the 2014 – 2016 stretch, in these terms. But as fans are concerned, it’s all been the same run. The Orioles built themselves an extended competitive window, and the projections just haven’t seen it.
One more plot. I can’t not bring up BaseRuns. You could argue that differences between actual record and BaseRuns record are just about un-projectable. BaseRuns strips away a lot of the randomness. Over the last five years, the Royals have won 25 more games than BaseRuns would’ve expected. The Orioles are tied for second at +24. Why don’t we take that away, just for fun? Here are five-year differences between BaseRuns wins and projected wins.
It’s still the Orioles in first, by 10 wins. So the projections haven’t just missed on account of the un-projectable events. The Orioles have very legitimately out-played their projections, even when you take away what’s mostly explained as random. In four of the last five years, the Orioles’ BaseRuns record has been better than the projected record by at least six wins. In the fifth year, they were the same. The point being, the Orioles have knocked their projections out of the park, and they’ve done it far more than anyone else.
Once again, I don’t have an explanation. I’d like to pretend that I do, but the best I can muster is some of this credit should go to Buck Showalter. And, you know, that back of the bullpen that we’ve all discussed. I’m not an authority on why certain teams over-achieve, and I might never be. But the Orioles have pretty consistently over-achieved, to the point where, over five years, only three other baseball teams have won more games. That’s outstanding, and to some extent it should serve as consolation for various moves that haven’t worked out. Orioles fans have agonized for a few years over the loss of Jake Arrieta, and that’s fine and justifiable. Losing a guy like that sucks. Yet by and large, despite everything, the Orioles of the last half-decade have been terrific. I’m not sure how long they’ll keep it up, but three trips to the playoffs are a heck of a reward, considering the math might’ve argued for tanking.
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.
When you look at the Royals and Orioles there are two huge similarities between the two clubs. They both have elite back end of bullpens and have both built their teams to their ballpark. I think the second one is the key. Both these teams play in ballparks which are extremes on the opposite ends of the spectrum. While neither team look great on paper, they are both perfectly built for the ballparks they play in. This is how they can beat their projections
The ballpark angle fits the 2016 results, but in the prior two seasons Kansas City was the A.L.’s best road team while Baltimore was the second best A.L. road team in 2014, so I’m not sure it’s actually that much of a factor.
Also, don’t the projections implicitly include the ballpark effects anyway?
It does fit the results. They were 50-31 at home and 39-42 on the road in 2016. But their pitching was better at home. 3.81 era at home versus 4.64 era on the road. The xfip was closer but still better at home. Same thing in 2014, although the road record was better. So yeah not sure it is much of a factor.
Fangraphs should just make this a featured article already. And/or give the Orioles their own metric…call it the Buck Factor.
I’ve been over this here before. The missing factor is wOM+ (weighted Orioles Magic plus).
The Buck Factor in the wild card game, however, was *disastrous*.
One bad decision did it, but it’s easy for us to 2nd guess too
And, just like the Angels, they torment their fans by surrounding a generational talent with heaps of flaming garbage.
Downvotes not getting Britton up, guys.
Where can we find projected wins by team going back to past seasons? Is that available somewhere on the site?
I don’t know why this is happening, I’m just enjoying the ride. Go O’s!
I’m gonna refer to my (late) comment on your other projections review article and note that those actual-expected graphs look more or less like a bell curve to me. Maybe a bit flatter/fatter than I might have expected, but it would be more surprising to not have outliers than to have outliers.
Comment to reference: http://www.fangraphs.com/blogs/instagraphs/the-projections-in-review-briefly/#wc-comm-5137028_0
This doesn’t actually mean anything – the residuals would still look like a bell curve even if the projection model was just “expect every team to finish at .500.” The actual value of the variance matters.
(Dave Cameron wrote a pretty bad article a couple years ago trying to make precisely the point you’re making here – amusingly, that one was also about the Orioles).
I’d like to see how FanGraphs’ projections stack up to simply projecting teams to finish with a record equal to the average of their last three seasons (perhaps with some weighting function to give the more recent seasons more weight).
It would look like a bell curve yes, but a much wider bell curve than one with more accurate projections.
Of course, which is why I said “the actual value of the variance matters.”
Unfortunately, I haven’t seen anyone do an analysis (Monte Carlo, maybe?) which gives us any idea of what the variance *ought* to be if outliers were purely due to “random” effects (sequencing and whathaveyou), so we can’t really draw any conclusions.
The biggest effect has got to be the bullpen. I don’t think people yet have a clear idea how to actually value an elite bullpen in terms of expected contribution to win total, but my intuition says current modeling probably vastly undervalues it. The thing about a bullpen is that, unlike with starting pitching, the manager can pick and choose the situations in which he uses his best pitchers. The effect one might expect from that is that a well-managed bullpen with a really strong back end and a few garbage men is going to be a lot better than a bullpen with a bunch of pretty-good relievers (but no outstanding ones).
On the flip side, BaseRuns (and all pythagorean-based win projection systems) also suffers from not being able to account for strong lineups beating up on aforementioned lousy long relievers in already-decided games.
Both of these are situations in which ordinarily descriptive stats like WPA might actually hint at a real effect.
If you look at the top 3 teams in terms of beating the projections (BAL, PIT, KC), and you think over these last few seasons, the bullpens from these teams stand out to me. No data to back this up so it is just my gut reaction to what I am seeing in the tables, but I have to think that a common thread here is the very strong and deep bullpens. There are probably other commonalities, but this one stands out for me.
Kinda meh SP too….BAL and KC especially have had weak rotations and PIT’s back end has been a revolving door of short outing tryouts and reclamation projects behind Cole/Liriano (when they still had him).
Does this have anything to do with exit velocity? Oriole players that Fangraphs and WAR overlook have great exit velocities (see Trumbo, Alvarez, Schoop, Kim, and even Nelson Cruz). Does Fangraphs still overlook hard contact? The Orioles don’t.
Sure is tough to overlook the exit velocity off of Encarnacion’s bat last night. I wonder if Buck overlooked it when he watched Jimenez give up three hard contact hits in 5 pitches and didn’t turn to his vaunted bullpen.
Sorry to rub salt in the wound, I just couldn’t stay quiet.
Hell yes they are better. I am glad someone on this site finally acknowledged that.
Well projecting bullpen value fails both by WAR and by BaseRuns:
First, the replacement level is set too high, minimizing their contribution (if two teams got equal performances from their staffs, but Team A had 30% of the innings thrown by relievers while Team B had 20% of innings thrown by relievers, then Team B would have more WAR, even though the performances were identical).
“To beat your projection five years in a row: 3.1% chance of happening.” That’s crazy. That’s 1 out of 32. That means you would only expect one team out of the entire league to do it over the course of five years – oh…wait a second.
I love the article, and I agree it’s about the magnitude. I definitely don’t think it’s entirely by chance. I just don’t find it hard to believe that one team beat its projections five years in a row. In fact it is exactly what I would believe.
I’ve been trying to ask DC all year why he thinks the O’s keep blowing away the projections to no avail. Glad to see someone here had the humility to actually address this even if no reason is evident. At least we’re not continuing to be fed ‘random variation’ as the answer.
This would be a terrific research study topic bc it appears to be something out on the frontier – something that current projection systems aren’t capturing. My instinct has been underweighting of bullpens and defense, and maybe that’s too simplistic but maybe it isn’t.
Until tonight! 😉
Ubaldo Jimenez in 1st innings in September: 10.80 ERA, 2.40 WHIP, 9.15 FIP. Not that you’d expect a high-performing manager to know that.