26 August 2012

Playoff Update: Oriole Magicks (August 26, 2012)

And, so, yes, the Orioles continue to defy their peripherals.  According to their run differential, the Orioles have have 11 more games than they should have won.  At the moment, they join the 2009 Mariners (10), 2008 Angels (12), 2007 Diamondbacks (11), 2005 Diamondbacks (11), and the 2004 Yankees (12) as teams that have exceeded their expected win total by 10 or more wins.  In other words, it is an occasion that has happened 1.5% of the time in the 2000s.  One remarkable trait of all teams is that they did well in close games.

2012 Orioles 23 6
2004 Yankees 24 16
2005 Dbacks 28 18
2007 Dbacks 32 20
2008 Angels 31 21
2009 Mariners 35 20
Close games do not fully explain their knacks for exceeding the run differential, but if you combine that with general deviation from the projected win totals...it makes sense that these events can occur.  It also makes sense that these events are likely not representative of any skills.  However, the data set is too small and the analysis is not incredibly in depth to make any solid conclusions.  Basically, the take home likely is that hopefully the ride continues this year and, come offseason, the team cannot rest on its laurels.

For this update, I have included "luck."  This metric can be defined as what our fWAR projection of wins compares with actual wins.  A positive number means the team is outperforming their projection.  As you can see, fWAR is wholly unimpressed with the Orioles.  Only the Indians have a lower fWAR than the Orioles.  This is quite remarkable.

As a reminder, the p30 projection considers the fWAR of the last 30 days.

Team Wins "Luck" Proj Final * p30 Proj
TEX  75 0 97 West 97
NYY  73 -1 94 East 93
CHW  70 8 88 Central 88
 TBR  70 4 88 WC 92
 OAK  69 7 87 tWC 88
 BAL  69 17 84 3 GB 86
 DET  68 0 87 tWC 90
 LAA  66 -3 85 2 GB 85
 SEA  61 4 77 10 GB 79
 BOS  60 -7 78 9 GB 77
 KCR  56 -2 73 14 GB 74
 TOR  56 2 71 16 GB 67
 CLE  55 4 70 17 GB 64
 MIN  51 -3 66 21 GB 67
One good bit of news in the past few days has been the Red Sox - Dodgers deal.  The Orioles have several games remaining against the Red Sox and they have become a weaker team by removing Adrian Gonzalez and Josh Beckett.  In their place will be the continually underwhelming James Loney and whoever the Red Sox can get to be their fifth pitcher.  The more Aaron Cook and Felix Dubront are in the BoSox rotation the better.  Of course, this makes the Rays and Yankees schedule easier, but it helps the Orioles against the teams in the other divisions vying for the Wild Card.


Matt P said...

From 2000-2011, when regressing fWAR (using all of the different components i.e batting WAR, fielding WAR etc) to actual win percentage, I found a significant relationship. However, it "only" had an r^2 between .8 to .82. Meanwhile, run differential has a significant relationship and has an r^2 of .8753.

I don't think using run differential is the answer but I think it's better than fWAR.

Your captcha thing has gotten more complicated. It's getting to the point where only robots will be able to read it.

Anonymous said...

Is the Orioles run differential really as bad as people say? here are some rought stats (might have made some mistakes

against the entire MLB (excluding AL West)
League - AL West 3

And as it makes sense, against the AL West

AL West -49

However, that -49 is really against just LA/TX

LA/TX -63

Now looking at the 7 teams they are going to play the rest of the way, their run differential against them is. They only have a negative against one.

White Sox 2
Yankees 3
Toronto 3
Tampa Bay -10
Oakland 4
Seattle 10
Boston 12

Jon Shepherd said...

I would expect run differential to be more related to record because that is what has happened. fWAR is more of a measure of skill, so it should be more accurate going forward as a predictive tool even though it is descriptive of talent framed as neutralized performance.

Jon Shepherd said...

What is difficult is that run differential will need more than a few games to be a decent statistical indicator. There are some good signs there though. I think you should write up and send us a post on it. We'll publish it as part of our expanded roster entries for September.

Anonymous said...

I'm not much of a sabermetrics guy, I just kept on hearing how everyone was harping on run differential, so wanted to see why the Orioles were so bad, so broke it down by team.

though the fact that they know how to squeak by the white sox was evi

Anonymous said...

these captches are a pita! (think it took me 10 times to get that last one posted)

Jon Shepherd said...

I am always signed into my Google account, so I never see the Captches.

Matt P said...

Fair enough about the WAR comparison. I guess the real article would be doing a month by month comparison and see which stat(Actual Win Percentage, Pythag Win Percentage, fWAR(all relevant parts) or brWAR) and see which one makes sense.

In order for the LA/TX thing to be meaningful, there would have to be two conditions.

We know that total run differential and actual win record have a large correlation. We would have to determine whether series run differential(i.e how many runs the Os score and give up to the Yankees over a given season) correlates with series win percent (how many games the Os win over a given season).

We would also need a hypothesis showing why the Orioles have struggled against LA/TX and not anyone else. LA and TX are good, but so are the Yankees and White Sox and we're +5 against them. LA and TX both score a lot of runs but so do Boston, New York and Chicago and again we're +17 against them. Why are we so bad against those teams but not against other good teams?

It's easy enough to test the first hypothesis. Using a dataset of all games played from 2000-2011(n~6800), we can determine that the variables are correlated with an r of .8266 and a significance of <.0001. While this is a lower correlation than the one for total run differential vs total pythag record, it's still a solid correlation.

Now all you need to do is find an answer to the second question. Why do the Os only struggle against TX/LA?

Matt P said...

It's also worth noting that we've been outscored 32 to 34 by Minnesota yet we've won 5 out of 7 games against them. We're -10 against Tampa but we took an even amount of games.

Even if you were to use the method I described above, it would mean the Orioles should be at 59-68 instead of 70-57.

In order to do what you're doing, you'd really need to also prove that the Os performance against those teams are statistically significant. Given the small sample size, that's going to be a tough challenge. There may be better approaches to explain this problem.

Anonymous said...

I think it might just be due to the way the O's play (or perhaps are managed?) that they are prone to being blown out by large #s.

perhaps in blow out games the pitchers are kept in instead of wasting the bullpen and continue to rack up runs, and sometimes they settle down, as in 5 run first innings, but solid the rest of the way.

It be interesting to see in blow out games what the distribution of runs to pitchers are and compare that to other teams.

Jon Shepherd said...

One key thing to remember is whether you are trying to figure out what metrics best describes what the Orioles have done vs what metrics best project what the team will do. They likely are not the same and I addressed this in a post a few weeks back to some extent.

Matt P said...

You're right and I do forget that. It's because I'm more interested in figuring out why the Orioles have done how they have so that's the way I think about the problem. The Orioles have been "lucky" for most of the season so far. But "luck" isn't a good explanation at this point in a season. As Fangraphs showed in an article, at this point, Current Win Percentage is a more accurate metric going forward than Run Differential. I can't prove it's also better than fWAR, but that's what I suspect.

Obviously, I can point to our record in one run or two run games as to why we've done well. Or I could point to our record when we score either two or five runs... but that's only a minimal improvement. It doesn't tell me why we're good only in those games.

If I find a metric that explains why we've performed the way we have(I've found one that's proven why Run Differential doesn't work for the Orioles but that doesn't really help much) then just maybe it will help me determine how we'll do going forward. But I'm mostly focused on that and that's why I get confused so frequently.

Although, at this point, if standard metrics have failed us so badly, I think a new solution needs to be found. Why would they start working now if they haven't worked for the first 130 games?

The Os are prone to being blown out. Therefore, they have a bad record in blowouts. However, they're still doing better than we'd expect them to do in blowouts. Given their run differential, we'd expect a .395 winning percentage instead of their current .447. But that difference is only two games which isn't all that surprising. It's the non blowout games which are interesting.

Also, relievers are more influential than starters in blowout games as opposed to close games. That makes sense as starters go longer when they're pitching well. Meanwhile, it's rare that a starter goes 8 innings if he's given up 6 runs.

I don't have good enough data to determine whether a starter or reliever gave up runs in a game because I don't know when a starter left the game. All I know is that a team gave up 5runs in game x while scoring 7. I could determine when those runs were scored/allowed but actually turning that into a metric would be difficult.

This makes it impossible to do the distribution you request. However, I will note that this team does excel its expected winning percentage when it scores under 5 runs. When we score more than 5, we're actually under the expected average. Same trend for runs allowed.

Jon Shepherd said...

What I go back to when thinking about this is a general way is this...why are teams unable to continually outperform their run differential at a high level? If there was consistency to this type of performance then it would seem like it was a skill that was being utilized. Historically that is not the case. One season a team exceeds what would be expected...the next, they don't.

That is what makes me not focus too much on this current team. We will see though. Maybe something pops up as an explanation.

Philip Taggart said...

I went to the Rangers game in which Beltre hit 3 home runs, and Moreland hit a grand slam.
The Os got Killed 12-3, but won 5-3 and lost 5-2. Remove the Pounding, and the net run diff is only -1
How many blowout wins have the Orioles had?
How many big innings?

The Run differential just shows that the Orioles lose BIG and win small.
in 2009, Seattle won 85 games with a -50+/- run diff, and it looks as if the Orioles will do better.
It might be very interesting to compare those teams, but it seems that the dominant factor this year has been the oriole ability to capitalize on opponent mistakes, and the bullpen's ability to pitch mistake-free ball.

Regarding the West, bear in mind that Angels, Rangers and Seattle(away from Home) have excellent offenses, but the Orioles haven't played the West that much(only 7 games all season against the rangers, for instance)

It will be much more dangerous to play against the Rays and Yankees pitchers.
What fun!
(what is a captcha?)

Jon Shepherd said...

I think the interesting thing is whether losing big and winning small is an actual representation of a type of talent. The Orioles have defied the run diff by winning a lot of close games. Is that a true skill? It has not been one in the past. If it is, then we should expect that production going forward.

philip Taggart said...

I figured it out.
It's Showalter!
Every manager's job is to determine who plays under which circumstances, and to inspire/motivate/control.

That's really it.

The Orioles are full of worthy players. They are good by definition, because if they weren't, they wouldn't be there.

But how to use them, and when. That's Buck.
Inspiring them, making them give that little extra bit that gets the job done, however marginally(see Run Differential)
Thats Buck.
I haven't seen any managerial analysis in this blog, but I came late to the Orioles Club, and may have missed something.

But consider:
1) Showalter has a historic trend of vastly improving teams. He did it with-at least- Texas and NY, and now with the Orioles.

2) Every player Showalter has gottn has played BETTER than their previous spots. Quintanilla has 4 homers last 2 months, but only 5 in his career!
Lew Ford has two Homers for the first time in 5 years. Mclouth came to the O hitting about .180, now he's up to .217.
Everyone is being inspired to be better than his best.
The whole season has been a team producing better than it's on-paper best.

I love your "magick" description, because it brings to mind alchemy, the source of and my own phrase"the random factors align."

But need you look any farther than Buck?
Somehow, he has the ability to evoke the best from his players.
The White Sox would certainly agree!

Jon Shepherd said...

So we are ignoring 2011 when got poorer performances out of several players?

Philip said...

I wasn't following during 2011, but he took over in 2010 and went 34-23 or something to end the year.
Next year, like his first full year in Texas, he faltered(I don't know what expectations were for 2011) but now, 2012, the Os are responding.
I think it's a defensible pattern

Jon Shepherd said...

For it to be a defensible pattern, I think you need underlying hypotheses.

I just don't see a good reason to think there is a pattern.

Matt P said...


It's easy enough for you to test your hypothesis. There's coach of the year awards each year, yes? Find the coach of the year rankings and run a regression modeling win percent with that ranking(best is one, second vote is two, no votes is fifteen maybe) and pythag win percent and see whether both variables provide a better model than just run differential.

You should be able to determine whether a coach helps and by how much.

Best of Luck!

Matt P said...

"I think the interesting thing is whether losing big and winning small is an actual representation of a type of talent. The Orioles have defied the run diff by winning a lot of close games. Is that a true skill? It has not been one in the past. If it is, then we should expect that production going forward."

I thought I posted a comment about this but I guess I didn't.

The interesting thing about that is that winning games with a run differential of 1 is NOT a talent. I did a regression comparing actual win percentage for a team and a season to its winning percentage when it played a game decided by 1 run, 2 runs, 3 runs etc until 10.

The correlation between actual win percentage and a game decided by 1 to 6 runs(the large majority of them) is pretty much the same for all of them(it goes up for games won by three runs by a decently large amount of (.04) but all the other scenarios are within .035 of each other.

However, the correlation between pythagorean win percentage and run differential has a different relationship. In those cases, there is a large difference in the correlation between winning a game by 1 run and winning it by 4 runs(.20).

I can presume that a team has won a similar percentage of one run games as it has won total games. However, I can't do that for pythagorean win percentage.

Therefore, games decided by a small amount are one of the many areas where pythagorean win percentage fails to be a valid predictor.

It isn't a talent because teams aren't expected to win close games any more than they win blowouts. It's just a place where using pythagorean win percentage doesn't work.

Philip Taggart said...

I wish I had enough variables to substantially support my contention, but I can't.
What we're trying to do-and it's a pleasant conundrum- is explain WHY the Orioles have been successful.

We know they have been extremely successful,compared to pre-season expectations, but Luck, by definition, is short-term. The Orioles can't be "lucky" over 162 games.

if we assume that the information used to make that pre-season prediction wa accurate, then we are left with but one possibility, the omission of something significant enough to have caused this success.

I'm suggesting that Buck Showalter, who has won MoY more than once, is that significant something.... and in the absence of anything more concrete, it's as good a reason as any...

Jon Shepherd said...

Why can't you have a statistical anomaly over a 162 games?

Many things do not settle out over that time span.

Matt P said...

You very well can have a statistical anomoly over 162 games. That doesn't mean it should be the assumed choice due to the results of a few metrics.

There isn't a deadline for submitting blogs, is there? If not, I have something that partly explains the season.

Jon Shepherd said...

There is no deadline for the submitted pieces.

I think we there is an anomaly that last 120 games that it is certainly something to look into and investigate. That is exactly what we have been doing and have posted many items trying to figure out what is going on.

Within a population, it appears safe to assume that whatever the Orioles are doing...it is not going to be repeatable and is therefore not a skill. As such, one would expect their success to end.

It has not ended. That does not mean the team is benefiting from a long string of fortunate hitting and pitching and it does not mean it is not. We have tried to answer the question both at the individual and group levels.