There’s some funny stuff going on in baseball. That’s more or less always the case, especially in September, but it’s probably worth pointing out nonetheless. The Minnesota Twins might make the playoffs. They weren’t supposed to be good at all but somehow have been. They’re 26th in fWAR for position players and 16th in fWAR for pitchers. They were projected to win 74 games by FanGraphs prior to the season but they’d already eclipsed that mark with 20 games left to play. They’ll probably win at least 85 games barring a major implosion.

Similar things apply to the Texas Rangers, and while both of these teams belong to the American League, this remains a relevant topic for Diamondbacks fans, or at least those who wish to think deeply about baseball. Because, in case you don’t recall, the Diamondbacks were projected to win a number of games totaling in the low 70’s, too, yet their reality is much different than that of the Twins or Rangers. So what the hell happened?

The most common response is that WAR is garbage and projections are meaningless. That viewpoint is entirely negligent of value metrics provide and ignores the fact that pretty much every major sports team is governed by them to some degree. Understanding the nuance of metrics isn’t easy, but you don’t have to understand them entirely to know they’re valuable. But no metric is perfect. No projection system is perfect. This is sports. These are human beings performing day in and day out. Life happens. Weird things happen. Players emerge and players crumble. Teams emerge and teams crumble. This is why we watch baseball.

Dave Cameron of FanGraphs wrote an abstract, yet poignant piece last week on this very topic and I’ll do well by borrowing from him. Addressing this very thing, he shared some overarching thoughts on evaluations, both of players and teams at large. He also addressed the danger of evaluating players and team retrospectively.

…the disconnect is mostly just a philosophical choice, and is the same choice that drives a lot of the disagreement around many of our less popular evaluations. Primarily, we evaluate players and teams by their inputs, not their outputs.

That last sentence is particularly important to grasp. When it comes to projecting a team, evaluations are done by looking at what a player does, not what happens. There’s no way to project exactly what the outcomes will be, but we have a good idea of how many times a batter will strikeout, how many walks a pitcher might allow and how many homers a player might put over the fence. We don’t know how many RBI’s that will produce because we can’t know who’ll be on base ahead of that hitter. We don’t know what a pitcher’s ERA will be since hits aren’t under his control and accounting for defense is always difficult and often downright mysterious.

If you’ve ever played fantasy baseball you know what I mean. What we expect to happen often doesn’t and that’s because we don’t have a firm grasp on what the outputs will be even though we do have firm-ish grasp on what the inputs will be. As Cameron rationalizes:

One of the primary goals of analytical research has been to attempt to isolate individual performance, as many of the more traditionally accepted measures of player value were based on taking a metric that involved contributions from many players and ascribing them to just one person. This has always been the core problem with pitcher Win-Loss records or RBIs, and has been one of the main reasons we’ve advocated for moving away from ERA as the primary way to evaluate a pitcher as well. These numbers are the result of a lot of variables working together, and we’ve generally preferred to move away from those kinds of metrics and towards things that focus more on measuring just the contributions of one player.

Hindsight is the enemy here, yet we can’t escape it. It’s easy to look at what did happen and try to rationalize what went wrong. Again, from Cameron:

We start with inputs and try to build up; I think most mainstream fans start with the output and try to tear down from there.

The problem with this is that it doesn’t teach us much since there’s so much in noise in what’s taken place over the course of an entire season. We can look back and try to make sense of the season, but we can’t throw out what was expected entirely. And just by the eye test, the Twins and Rangers didn’t look like contenders, let alone playoff teams. Alas, here we are and that’s mainly due to one thing: sequencing.

Sequencing refers to the order that events happen on the diamond. As far as anyone can tell, these are totally random. We literally can’t predict the order of events. Here’s an illustration.

  • Scenario One: a pitcher allows a single, another single, gets an out, gets another out, allows a homer, gets the third out. Result: three runs score on the homer, no runners stranded.
  • Scenario two: a pitcher allows a homer, gets an out, allows a single, gets an out, allows another single, gets the third out. Result: one run scores on the homer, two runners stranded.

The same events happened, but because of the order in which they happened, the outcomes are wildly different. There’s literally no way to predict ahead of time how the events will occur, if they’ll be spaced out or come in bunches. Some pitchers struggle to pitch equally well out of the stretch, but this is difficult to quantify. Hits happen more often with runners on base since fielders have to hold runners on, and this is considered, but it doesn’t account for the wild variance in which runs score and games are decided.

FanGraphs knows this and has created a handy metric to explore it called Base Runs. They strip away the sequencing of events and predict wins and losses based on the “normal” distribution of events. Early in the season, this was useful when examining the Toronto Blue Jays. They had scored roughly 80 runs more than they’d allowed, yet they had a losing records. Because these things tend to even out, the Jays have gone on to lead the AL East since they score so many more runs than they give up. To call them a poor team when they had a losing record was foolish – they’d just been unlucky. That luck turned to align with their talent level and the rest is history.

But that hasn’t been the case with the Twins and Rangers. Minnesota is 75-71, but by normal distribution of events (Base Runs), they should be 65-81. They’ve been lucky when it comes to getting the timely hit or stranding runners in scoring position. Texas is 77-67 but should be 72-74. The Twins, for example, are batting .279 with runners on base but just .249 overall. That’s a massive discrepancy and something you just cannot predict. For the record, the D-backs are 69-77 but Base Runs places them at 71-75. Arizona has been a touch unlucky. This variance happens but we just can’t project it.

So how does this apply to our situation? When it comes to looking ahead at 2016 and 2017, the seasons we’ve been hyping up here at Inside the ‘Zona for over a year now, we can do our best to project, our best to predict. But how it all plays out is still up for discussion. Seasons like the Twins and Rangers are having are extremely uncommon and have almost never been seen before, but they do exist within the realm of possibility. And while we look at the inputs, and the potential inputs to the Diamondbacks over the next couple of years, it’s important to realize that things can go wrong and sometimes do. It’s also possible that the team exceeds expectations like we could have only dreamed. But it’s most likely that they perform at or near the level projected and that’s what they should be building for. We’re all excited about what that level looks like, but we’ll have to watch it all play out on the field, and this is the best part.

So before people bemoan the numbers, remember that they are most often right and they don’t take away from the magic on the field. I’m pretty sure it’s fun to be a Twins fan right now, or Rangers fan, or a Cardinals fan. It’s never fun to be a Dodger fan. And whether your team performs as expected (Cardinals) or blows the numbers out of the water beyond what can be imagined (Twins and Rangers), just keep in mind that we do our best, but the ball won’t necessarily bounce our way. That’s baseball, and it doesn’t invalidate the work we do to understand it.

Tagged with:

5 Responses to Projections, Evaluations, Weird Things and Life As We Know It

  1. Sam says:

    Don’t throw in the towel so quickly. The goal of any prediction system is to accurately model reality, and saying that you literally can’t predict something suggests that maybe your model just isn’t powerful enough yet.

    Adding up the pitcher’s and batter’s contributions as if the game state doesn’t matter is just a first-order approximation. Analyzing performance differently with runners on is indeed the first second-order correction to consider. You could also look at situations where the infield is brought in, distinguish innings where one run is all you need from those where you need multiple, distinguish the first batter a reliever faces from later batters, and measure to what degree the hitter on deck matters to the outcome of an at-bat (beyond walks).

    All of these are still on the input side, but can hopefully help us actually get to the goal of predicting what happens. They’re already pretty common slices to look at, and maybe some players have skills making contact when contact is needed (infield brought in), or getting the first batter they face out.

    As a Diamondbacks fan, I’m not surprised that they’ve underperformed. How many solo home runs have we seen recently, while they waste opportunities with runners on base? How many times has Goldschmidt hit a two-out single with no one on, but ground into a double play if Inciarte or Pollock gets on before him?

    • Jeff Wiser says:

      The point I was trying to make isn’t that the projections don’t work – they do a fantastic job by and large. The sequencing is the part no one can nail down because there’s just so much variance in what happens when the ball leaves the pitcher’s hand. I hope that much rang through.

      In terms of preseason projections, the D-backs should finish the year about four wins over their preseason predictions, maybe five. So it’s hard to say they’ve “underperformed,” it’s more that they’re starting to regress to the mean. Given the roster on hand, I think that’s something I can accept.

      And on the topic of “sequencing,” the power lies in timely hitting. There is virtually no correlation between hitting with runners on base from year to year. There are often wild swings and if you catch it on the upswing, you’re going to be a powerful offense. If you catch it on the downside of average, it’s going to be a tough year. The problem is that players just don’t seem to carry much consistency from year to year in this department.

      Thank you for the thoughtful comment, Sam! Would love to see a few more of them here!

      • Dave-Phoenix says:

        I have to agree.

        In April, on paper, this team was expected to have good offensive production and good defense, but had no pitching. As the season played out, that is exactly what they had.

        This team did not underachieve. By winning as many games as they have, the 2015 D-Backs actually over-achieved. I give the credit to the offense doing more than projected.

        So, what will the 2016 D-backs look like?

        I think almost any team would be happy with the D-Backs 2015 offense and defense numbers, so the FO’s goal for 2016 should be primarily to keep what they have. Maybe a few tweaks here and there. The biggest tweak, will probably be 2nd base.

        So that leaves the D-Backs with pitching. If you lead the league in offense and defense and finish below .500, it is pretty obvious where the problem lies. If the D-Backs don’t acquire better pitching, they won’t be any better in 2016, than they were this year. End of story…..

  2. Anonymous says:

    I like that there is a little unknown because otherwise there would have been little reason to watch the D-backs this year (gotta have hope). Although I can’t imagine how frustrating it would be to project to be really good and then turn out to be really unlucky. But the game is better for it.
    With that being said, I think the only people who are demeaning the use of analytics are the people who never gave it a chance in the first place. It’s a convenient time to rag on it because of what’s happened with the Twins. But quantifying value to analyze what’s most likely to happen in the future is the best way to go. I look forward to other advancements in analytics. Even though a moneyball-type breakthrough will probably never happen again, unfortunately

  3. OJ Carrasco says:

    It is an interesting study in perspective. If you like projections (as I do) you see it as acceptable for a variation of 3% (5 games) or even up to 6%. If you don’t like (not sure how it affects somebody, but it appears to happen), you use any error to denigrate the system. The projections generally come with the caveat that differing factors (injury to player on team, player developing a new skill, injuries to players on other teams, et cetera) can cause swings.
    I like to view the projections at the beginning and end of the year and see how close they got, it’s a pretty neat system.
    An interesting parallel to baseball projections is weather projections. Last year a lot of meteorologists projected a major snowstorm on the east coast, with lots of damage, it didn’t land and people mocked them. The problem was that a wind current switched, slightly and the storm hit off the coast, by about 2 miles. These people missed by 2 miles, a minuscule distance in terms of the ocean and were mocked as idiots.

Leave a Reply

Your email address will not be published.