Wednesday, 22 July 2015

Fun With Arbitrary End Points

This idea's been knocking around in my head for a couple of weeks or more and I've put it off, reluctant to do the drudge work necessary to parse the stats. But the trade-fever is rising to acquire pitching, almost certainly a starter, and with Mike Wilner's column Monday providing renewed inspiration, I took some time this evening to dig into the idea. What you are about to read is going to be rife with caveats, but that's okay because I'm not trying to be exact here but rather just give an overview of a general idea that you might have sensed but been sure about.

With time perhaps tthe impression has faded for many, but I noted before in this space (about halfway into the season) that there's a distinct difference between the performance of the team over the first 40 games, and thereafter. Coming into Tuesday night's game the team has played 94games which creates some convenient points of reference. Forty games - not mathematically 1/4 of the season but as close as you can get in whole games - followed by 54 games, 1/3 of a season. There's a big difference between the two. Why did I pick that date? If you're a regular reader (hey you two!) you can possibly guess. The 40th game was the one that the Jays surprised team watchers by inserting Todd Redmond into the rotation for one start to give the rest of the rotation a day of rest. It's hard to claim, of course, that this act had any direct impact at all for more than 10 days or so, but nevertheless, the numbers are striking.

Let's also distinguish that I'm not talking about the hitters here. Through 40 games, the Jays averaged 5.25 runs per game, since then it's climbed to 5.33 runs per game. Up, but not enough to point to any sort of trend. And with that, Caveat #1: This is going to be all about what are commonly called Arbitrary End Points (AEP) - primarily that one AEP on May 18. However, it's not quite as useless as that might make it seem because it's a straight "before and after" look at samples of significant size, it's not like saying "if you took out this bit and that bit then..." (although there are two minor points where I notice something by doing that further down). I believe ther is value in noting a "turning point" either for the good or the bad - you just can't get carried away with it. There's also value, in my opinion, in noticing if a player was very good most of the time but had one wee or 10 days or whatever when he went sideways. For example, that whole business about how Marco Estrada sucked last year? Yeah - for about 3 weeks. He got absolutely rocked for a short period, including a huge chunk of the homers he allowed, and outside of that stretch was just the same as he'd been all along. I think noting that was predictive of what we could expect this year.

This entry is about pitching. I'm not going to argue that I know why the breakdown looks like this, or that it's anything other than random in terms of which AEP you might have chosen. But it exists. In the first 40 games, they were 18-22 and since then, coming into the Oakland series, they've gone 29-25, and when you see these numbers you'll be scratching your head that the latter isn't way better.  What you are about to see has great risk of being presented in a confusing way, but hopefully I can organize it reasonably well. There will be two major sets of stats, obviously, separating the Starters from the Bullpen - with only Estrada making this difficult. I'll sort him out though. Another caveat: ERA is a bit of a blunt instrument, there are other stats of the sort that inform xFIP for example, which get more precisely to the quality of pitching, but for an overview like this, ERA will do.

The Jay's starting set has been remarkably stable this year. Three have made every turn of the rotation, Norris gave way to Estrada who's been a rock, and Sanchez is the one injury which triggered a revolving door which has not produced good results. What I've done, then, is charted  the five who held their job on quality as a separate set from those who did not hold on (and Doubront who seems destined to lose out to whomever is acquired). This also allows me to distinguish those who pitched on both sides of the AEP from those who haven't. Before you read further, do be aware I've caught myself in several minor errors already (like failing to remember Doubront had a bullpen appearance) so if you spot an error, don't be stunned (as if...).

First set. The first column of numbers is their ERA through May 18, and the second column is their ERA since.  Then a column with IP and ER before, and one with those after the AEP, and finally a column with IP per start before and one after. Note that Estrada did some quality relief work before moving into the rotation, the stat you see represents only his starts.The bullpen work will come later.

NameERA BeforeERA AfterIP ' ER BeforeIP ' ER AfterIP/G BeforeIP/G After
Buehrle5.362.0647 ' 2874.1 ' 175.887.41
Hutchison6.174.5242.1 ' 2961.2 ' 315.295.61
Dickey5.763.9550 ' 3270.2 ' 316.256.42
Estrada5.523.114.2 ' 1069.2 ' 244.896.33
Sanchez4.262.5738 ' 1828 ' 85.437
Total5.443.28192 ' 116304.33 ' 1115.496.48

Now here's a chart for the short-term guys

NameERA BeforeERA AfterIP ' ER BeforeIP ' ER AfterIP/G BeforeIP/G After
15.1 ' 11
6.2 ' 11
11.2 ' 6
23.1 ' 10
Total3.868.8123.1 ' 1033.66 ' 304.674.81

And finally, the total for all starters:

All Starters5.273.79215.1 ' 126337.33 ' 1425.386.25

Now, even in the totals it's striking. Over a run and a half down in Starters ERA, almost a full inning better in IP per start. But if you filter out the failure to thrive threesome in Sanchez's spot (7 starts) along with Norris (just to be fair), then the ERA swing is a fat 2.16 down, and the IP per start gets even better. To put these in some context, here's a list of AL teams ranked by starter's ERA on the season. Obviously the comparison is imperfect, since they are not split the same way the Jays' numbers are, but I'm not trying to prove where they rank since May 18, just giving you an idea where these numbers would rank if they'd been doing this all year.

Oakland - 3.00
TB - 3.42
LA - 3.56
Seattle - 3.80
Minnesota - 3.83
Houston - 3.97
Chicago - 4.00
Cleveland - 4.12
Baltimore - 4.22
NY - 4.22
Texas - 4.26
KC - 4.33
Toronto - 4.35
Detroit - 4.52
Boston - 4.84

So they are ranked 13th out of 15 right now, but if they'd done all year what they've done in the past 54 games, they would rank 4th - and if you take just the front five guys over that span they have the second best ERA on the list. Pretty remarkable, yes? One more tidbit on the starters. If you take the moment when their ERA peaked and they started trending downward, you of course come up with different dates for each of the main 5 guys, but it's noticeable how close they are (relative to the 5 day turn through the rotation) to that May 18 AEP.
On May 12 Buehrle peaked at 5.54, he's totaled 2.30 since (less if you factor in his shutting down Oakland tonight but I didn't).
Hutchison peaked on May 3 at 7.47, since then it's been 4.21
Dickey peaked at 5.77 on May 26, since it's been 3.49
Estrada doesn't fit well since his ERA was never again close to what it was after his first start on May 5, it's trended downward with a steady pacesince.
And Sanchez splits exactly with the AEP.

Sane treatment on the bullpen. This time it divides neatly into two groups again, those with 15+ IP (as relievers) who just so happen to all have pitched on both sides of the AEP, and those with less who only pitched on one side of the line. Noticeably, the first group represents the current seven pitchers in the bullpen, plus Todd Redmond who's at the fulcrum of this analysis. The one exception to my methods here is that I failed to distinguish Redmond's start from his relief work, but by the time I realized what I'd done all the charts were finished and I was well into this commentary, and damned if I'm going to go back and adjust all those figures. Just know that the pre-AEP starters figures, collectively, will go up just a bit more, and the pre-AEP bullpen figure for Redmond will too (I know it's weird but Redmond's first two bullpen appearances were so bad that a bad start still pulled it down some, meanwhile the collective bullpen figure in the "before" column will get just a bit better.

NameERA BeforeERA AfterIP ' ER BeforeIP ' ER After
Osuna0.863.4321 ' 221 ' 8
24.1 ' 5
Hendricks3.632.4917.1 ' 721.2 ' 6
Cecil3.465.2913 ' 517 ' 10
Loup6.284.1214.1 ' 1019.2 ' 9
Delebar1.84.265 ' 119 ' 9
Tepera4.262.196.1 ' 312.1 ' 3
Redmond11.882.358.1 ' 118 ' 2
15+ IP4.113.2785.1 ' 39143 ' 52

Notice that Osuna actually went up, as did Cecil and Delebar the collective total is still striking. Almost two full runs difference. Speaking of Cecil and Delebar, the latter's second section ERA was a sterling 1.81 until July got here, three recent rough outings ballooned it and puts him on the bubble as a candidate to farm out when Sanchez is activated. Cecil has an even more interesting, to me, story. For one week, June 15-21, he fell apart, giving up 8 ER in 2.1 IP over 3 appearances. Without those his ERA in the second column would be 1.23 and lest we leave him out, if Aaron Loup had called in sick on June 20, his second column figure would be 2.84 so...not going to guess what's up with Delebar because the blow-up is recent, but these guys are not as bad as their results, and that's before you add in Sanchez.

Unlike with the starters, the short-term guys really don't bust up that trend.

NameERA BeforeERA AfterIP ' ER BeforeIP ' ER After
12 ' 9
12.1 ' 6
10.2 ' 1
3 ' 2
2.2 ' 1
2 ' 1
3.382.2 ' 1
2.1 ' 1
0.001 ' 0
others >4.223.0042.2 ' 206 ' 2

The thing that jumps out at you here is the IP totals. Over 1 IP per team by a shuttle pitcher in the first set of games, less than 7 IP in 54 games since. That's called stabilizing, folks, and it shows up in the results. Almost a ran and a quarter lower in collective ERA in this group. Here's the bullpen totals:

Bullpen >4.153.26128 ' 59149 ' 54

Almost two full runs better, and not surprising given what the starters did, 3.2 bullpen innings per game in the first split, 2.76 in the second.  Those ERAs look good compared to the league too. The collective bullpen ERA in the last 54 games, if it were for the season, would land the Blue Jays 5th, just ahead of the Yankees.

Finally then, the totals for the whole team, before and after:

Total > 4.853.63343.1 ' 185486.1 ' 196

How does that play against the league?  That 3.63 nestles between Baltimore (3.73) and Tampa Bay (3.59) for 6th in the AL and for love of 1/10th of a run, tonight's effort by Buehrle might have put them in second place (#1 is Oakland at 3.35 followed by the Angels at 3.52).  The major point of all this is when you hear broadcasters or journalists say "the Jays have some bad pitching" - they really don't. The HAD some bad pitching, over the first six weeks of the season. Since then they've had perfectly reasonable pitching on the whole.

How then to reconcile this with Wilner's description of how the bullpen has somehow managed to lose games at a higher rate than they should have? If you haven't read it, go do so for context, but the basic idea is that they pitched really well in low leverage outing but sucked in high leverage situations. Wilner reports accurately that the Jays lead the majors with 15 blown saves, so let's start there. But that will have to wait until next time, because sorting that out pre- and post-AEP is too big a chore to begin at 3 AM.


After another long session with Baseball Reference while listening to the Jays and A's I'm prepared to say a little something about Wilner's reporting on how much trouble the Blue Jays bullpen has had in high leverage situations. I didn't bother to try and re-create his reporting or be more clever than BR in terms of what qualifies as high-leverage. Rather I just wanted to see it for myself and I needed some sort of measure to quantify what I'd seen. So what I decided was to look at every situation in which a Toronto reliever pitched when the score was within 1 run either way (or tied of course) and how these turned out. I'm not entirely comfortable with the method because if, for example, three pitchers throw 3 innings of relief while tied and the third one give up a homer, the first two actually did their job. But this is more about collective results and the finer the detail, the longer it would take me to parse out.
ETA: Tonight's extra inning loss is a good example. It was both a success and a failure. Osuna gave up the game winner in a tie game, however,4 relievers combined to throw 3 shutout innings while the team was down a run, or tied. The reports will say the bullpen blew the game, but that won't do justice to the good work they did to stay close. That's where those stats Wilner cited work better.

So what I found was that in 34 games so far, such a situation occurred. and the Blue Jays only won 11 of those. In 10 of those games, the bullpen did it's job and four times the Jays lost anyway (situations where they were 1 run down), while 24 times they failed, but in five of those the hitters bailed them out. That's just awful and honestly, it defies reason given the overall quality of the players involved. In reference to my AEP, 12 of these cases occurred during that first 40 games and the bullpen failed in nine of those. Six of those nine were at least partially the fault of Castro or Francis but that's all that can be taken as a "positive" from this period.

Since the AEP, the success/failure record is 7-15 which is some better but still not good, but to introduce another AEP, since June 1 (the win streak started the next day) it's been 6-8 which is more reasonable. Honestly, I'm not really sure what would be considered an acceptable rate in a profile like this so I'm just defaulting mentally to .500 as a point of reference - which means over the last seven weeks they were only one game below where you might reasonably ask them to be.Still, on the season, if you expected a .500 record that would mean that, setting aside the 4 games they were already behind, and the 5 the won anyway, they might have won 6 or 7 more games than they have. If you do the same but disregard the games before the AEP, they might have won 3-4 more since then than they have (again, disregarding the Oakland ongoing series). Four more wins and they would be a game and a half behind the Yankees right now.

Another remarkable aspect of all this: Who among the relievers is considered by everyone to be the single most reliable reliever we have? Osuna right? Yet Osuna has been involved in more of those 24 failures - eight - than any other member of the 'pen (and also involved in six of the successes). Cecil is next with six (4 of the successes), Loup five (4), and Castro four (1)...and on the other side, Delebar was involved in only one of the failures, but in four of the successes. Which brings me to the one big take-away from all this drudgery: You can't just assume that adding any given pitcher will fix the problem because it's a problem that really shouldn't even exist. If your best reliever is involved in more failures than anyone else, and you trust him - as you should! - then how can you say of any pitcher you might add that such events won't happen to him?

This all strikes me as akin to the annual examination of records in 1 run games. You can certainly wise it were better, if it's bad - but if it's bad that doesn't necessarily prove you're doing anything wrong. All you can really do is hope the scales eventually balance.

No comments: