At its heart, baseball is a battle to control the strike zone. There are plenty of other things going on, but the origin of the action is over the plate. Good hitters make good decisions about when to swing and when to take and good pitchers attempt to negatively impact that decision-making process. As the importance of walks and working counts became clear over the last generation, hitters who knew the zone and pitchers who could generate swinging strikes became very popular.
Throughout history, batters have been judged by their results. Things like batting average and RBI have given way to wOBA and WAR, but in general the average fan cares about the outcomes rather than the process. Plate discipline numbers are inherently process based. You don’t get credit in the box score for taking a pitch just off the plate, but taking a pitch just off the plate is probably going to help you do things that lead to runs, like walking and getting good pitches to hit.
Read the rest of this entry »
Perhaps one of the biggest objections people have with the current state of defensive metrics is that the stats don’t account for the starting position of the defender. Shift plays are excluded from the calculations, but when a center fielder plays in 20 feet, the system doesn’t know that he’s starting from a different spot than the average center fielder, which could obviously lead to some imprecise accounting.
This is true for every position except pitchers and catchers, as the starting location of the fielder influences the probability they will make a play, independent of anything they do from the moment the ball is pitched. If you start out of position, even if you run at top speed and take a perfect route, you might not be able to offset the initial disadvantage of not being in the right spot to begin with. This creates problems, but there’s a lot of nuance to these problems that are worth discussing, even as we get closer to having StatCast and rendering the discussing irrelevant (we hope!).
In addition to the daily analysis and normal statistical offerings, FanGraphs has added some pretty useful and powerful features over the last couple of years. Anchoring a lot of those features are the Depth Charts, which in addition to providing information on their own, power the playoff odds and projected standings we host on the site.
The Depth Charts are pretty simple in theory. They blend together two of the leading projection systems (Steamer and ZiPS) and then scale those projections to our expectations about playing time. The Depth Charts are updated constantly to provide the most up-to-date snapshot possible for the current state of a team, league, or position. You can think of the Depth Charts as the baseline projections for the entire site, as they are the input for the projected standings, playoff odds, and game odds.
As far as the basic Depth Charts are concerned, there are essentially three different views. You can look at a team’s Depth Chart, you can look at Depth Charts by position, and you can look at the summary data of both of those at one. To generate each the charts, we take a 50/50 mix of Steamer and ZiPS for the rate stats and then our staff manually allocates playing time based on what we expect teams to do with their lineups and injury histories.
Steamer and ZiPS update nightly throughout the season and our playing time estimates change every 15 minutes (if necessary). If a player gets hurt, we update their playing time. If a player gets moved to the pen or changes positions, we update the Depth Charts. Also, the Depth Charts are showing what we expect to happen for the rest of the season, not the stat line we expect them to end the season with.
As always, when you’re dealing with constantly updating information, there are occasionally bugs. If you see something that looks obviously wrong, it’s likely just a database error that will resolve itself once the system updates in a few minutes.
As far as viewing options, you can look at the Depth Charts in team view, in position view, or in summary view. In team view, you get a breakdown of a single team by position, meaning on the Blue Jays page there’s a box for catchers, first basemen, etc with the expectation that each position for each team will receive 700 PA per season. Obviously that will vary a bit, but it’s a good rule in general. Each team also has a box for all positional players and all pitchers, as well as a box on the right that shows you where they stand overall.
In position view, you can look every team’s Depth Chart at any one position. For example, here is the page for catchers. This allows you to compare positions around the league and see which group of backstops is most valuable. Obviously these rankings are based on the projection systems and our playing time estimates, so if you believe playing time will shake out differently that we do, you might expect to see a different overall ranking.
Finally, this handy grid collapses those two views into one. You can’t see all of the players in that view, but it puts together each team’s expected WAR at each position so that you can quickly compare how teams and positions stack up against each other.
The Depth Charts are very useful for a couple of reasons. First, they blend two projection systems together without you having to do any of the work, and that’s helpful because aggregate projections are better than any one system. Second, playing time is controlled by humans. While projection systems are much better at forecasting performance than people, projection systems aren’t very good at figuring out how much playing time a player is actually going to get. Finally, the Depth Charts gather a lot of information in one place. We’ve had projections on the site for years, but having them built into the system like this allows you to make a lot of comparisons and see where teams are strong or weak.
So as you get back into the swing of things this season, the Depth Chart pages will be a valuable resource if you want to look into the future. Obviously, the charts are only as good as their inputs, but if you care at all about the inputs, the way the data is presented is really helpful.
A baseball season is the amalgamation of a lot of little events. Each pitch fits into a plate appearance which fits into an inning which fits into a game which fits into a series which fits into a season. That’s a lot of little data points flowing into an overall end result. We care a lot about which players will have good seasons and careers. It matters to us that we can distinguish between good players and bad players, but doing so requires that we understand which chunks of data are meaningful and which aren’t.
Enter sample size. You’ve heard this phrase plenty over the last few years when talking about baseball statistics and it’s usually a conversation ended rather than a conversation started. Someone cites a stat and then another person says it doesn’t matter because the sample size is too small. What does that mean and how should we properly think about sample size in baseball?
As you might have noticed, our playoff odds and projected standings are now up and running for the 2015 season. If you’re a regular FanGraphs reader, or intend to be this year, you’re going to see a decent amount about the various numerical expectations we post on the site. While these odds and standings are a lot of fun and a great tool for taking stock of the league, it’s also pretty easy to misunderstand or use them improperly.
Before I run through the proper way to read the odds and standings, I want to provide a brief overview of how we arrive at the numbers you see on the site.
Our player projections are based on the FanGraphs Depth Charts which are generated by giving equal weight to Steamer and ZiPS (two projection systems) and then manually estimating playing time. Then based on the depth charts, we simulate the season 10,000 times and report the results as playoff odds and projected standings. We also host a Season to Date model and Coin Flip model which project the season based on the current year’s stats (instead of projections) or a 50/50 chance at winning each game, respectively.
Getting newcomers on board with Wins Above Replacement has a number of challenges, but the way we measure and evaluate defense is typically one of the biggest sticking points. Getting an open-minded person to believe in wOBA instead of average and RBI isn’t that difficult. Getting someone to accept that there’s more to base running than the number of stolen bases is pretty easy. Convincing them that it’s useful to compare players to replacement level is a bit harder, but nothing really compares to the questions people have about defense.
There’s good reason for this. Again, a thoughtful person can see the flaws in using errors or fielding percentage, but it’s harder to sell the merits of runs saved metrics for a number of reasons. If you want a little more information on how we measure defense and why we do it that way, check out our beginner’s guide to measuring defense. Today, we’re going to consider a corollary to the actual measurement of defense which is the positional adjustment.
Batting average is the most recognizable statistic in the game. It might be the most famous statistic in sports and it’s probably up there with Gross Domestic Product (GDP) among the most popular statistics about anything anywhere on the planet. Even people who don’t like or watch baseball understand what batting average means. Just like how you know a singer is famous because your mother knows who they are, you know batting average is huge because you never have to explain it to anyone.
Which is why it’s so difficult to remove it from our vernacular. Batting average is built into the language of the sport, but it’s simply not a useful statistic and if you want to analyze a player properly, it’s something you don’t want to pay close attention to at all.
Like any good acronym, the letters in WAR each stand for something. The “W” stands for wins, which is something with which we’re all pretty familiar. The “A” stands for above, which is just an adjoining word, but the “R” stands for replacement which is a place where newcomers sometimes get lost. What is replacement level, why does it matter, and how do you calculate it? If WAR compares players to replacement level, to understand WAR we need to understand R.
Let’s start from the beginning. Replacement level is simply the level of production you could get from a player that would cost you nothing but the league minimum salary to acquire. Minor league free agents, quad-A players, you get the idea. The concept is pretty tidy. These are the players that are freely available and if five of your MLB level players came down with the flu, you could go out and acquire replacement level players without really giving up anything you value other than their union mandated payday.
In other words, if you had no one on your roster on April 1st and just needed to populate a team, you’re generally signing replacement level players.
If you’re not someone who comes up with trade proposals, you’re someone who reacts to trade proposals. It’s one of the great baseball fan parlor games. They’re everywhere. They populate our chats, they dominate Twitter, and they even sneak into real live interpersonal communication. Would the Nationals trade Strasburg? What could they get? Who would they want? These are all very interesting questions, and while most trade ideas disappear into the ether, plenty do come to fruition.
We talk a lot about trade value on FanGraphs because a lot of our writers care about the roster construction aspect of the game. Certainly we cover what happens between the lines, but there’s a lot of interest among our readers regarding how those players happened to wind up on the teams in question.
Every summer, Dave Cameron runs a trade value series where he ranks players based on his reading of the baseball landscape. Jonah Keri has a similar series at Grantland every winter. This is a topic that generates lots of interest, so this post is going to lay out the variables you should consider when pondering what a player is worth to the rest of the league.
One of the things that makes baseball interesting is that none of the playing fields are the same. In the NHL, NBA, and NFL there are certain things that might make certain stadiums feel different than one another, but the measurements of each are the same. In baseball, the bases are all 90 feet apart and the mound is at regulation length, but the fences vary by distance and height. You can travel to all 30 parks and never see the same same dimensions twice, but that also poses a problem when trying to evaluate the game because there’s an additional variable influencing the outcome of every plate appearance.
If we want to properly evaluate players and teams we need to have some way of adjusting for the fact that every park is different. More specifically, each park plays differently for reasons beyond the outfield dimensions. If you pitch at Coors Field in an identical manner to identical hitters as you pitch to at AT&T Park, your results will be different due to the ballpark. We want to try to control for this when we create statistics, so we apply something called a park factor to even out the differences.
These park factors are imperfect for a variety of reasons, but what they’re after is on the money. The parks influence the game and we want to strip that out of our evaluation of individuals.
It’s not just the dimensions. The dimensions matter, obviously, but deep fences don’t automatically make a pitcher’s park and short porches don’t always favor hitters. In addition to the dimensions, the weather matters, the air density/quality matters, and topology of the surrounding area matters. The ball tends to travel better in warm air and thin air, and the surrounding buildings and ballpark structures can influence how well the ball carries.
Petco Park, for example, has a marine layer that doesn’t let the ball fly. You probably know that Denver is way above sea level, making the Coors Field air thin and ripe for plenty of carry. Beyond that, the arrangement of the stands can influence how well the ball flies and the average temperature certainly affects the game play.
So while “big” and “pitcher’s park” are often used synonymously, there is more to it than that.
If you had the power to do so, you’d want to know how every single plate appearance would play out in all 30 MLB parks. If it turned into a single in the park of interest and then went for a single in 25 other parks, an out in three, and a double in one, you’d have a good sense of the way the parks played. The park that allowed the double would be a hitter’s park and the ones that created outs would be more pitcher friendly. But unfortunately, we don’t have that kind of data.
We want to know how parks influence each moment of the game, but we simply don’t have granular enough data to really get there. A ball hit at 15 degrees directly over the shortstop while traveling at 93 miles per hour will travel how far and land where? That’s basically what we want to know for every possible angle and velocity, but we just don’t have the data and we don’t have it for every type of weather in every park.
Instead, we have to settle for approximations.
There are many different park factors out there. We have some. Baseball-Reference has different ones. Stat Corner has more. Individuals create some. It goes on and on. We use 5-year regressed park factors and you can dive into our method here.
At the end of whatever process you choose, you wind up with a number that communicates how much more offense is produced in that park than you would expect to be produced in an average one, and when we display them on the site, we cut them in half so that you can more easily apply them to player statistics.
A league average park factor is set to 100 and a 105 park factor means that park produces run scoring that is 10% higher than average (halved so 110 becomes 105 in 81 games). We also provide park factors for each type of hit and batted ball, and for handedness, although we use the general ones when making park corrections.
For example, if a player has a .340 wOBA, but their home park is hitter friendly, they we need to adjust their wOBA down as a result. We don’t calculate a wOBA+, but some do. Instead, we jump over to wRC+ for our park adjusted offensive metric. This stat, among other things, applies a park adjustment to the player’s batting line. Stats like ERA- do this as well, and pretty much any time you see a +/- stat, it’s park adjusted.
And our park factors are applied with the additive method, meaning that we’re essentially adding or subtracting a little production based on how much offense is affected by the park in our estimate, but remember that we only apply half of the full park factor because a player only plays at home for half their games. We assume the rest are played in a pretty average setting.
As I said before, park factors aren’t perfect for a variety of reasons. They do a nice job on average, but in specific cases they fail to properly capture the nuances of the game. For example, Target Field is actually a slightly above average park for hitters. It’s on par with Yankee Stadium in fact, despite the much different dimensions. However, if you’re talking specifically about left-handed home run power, Target Field is a desert and Yankee Stadium is an oasis.
The problem with park factors as they stand right now is that while we’re trying to adjust for the run environment, the run environment is difficult to capture is a single number. Lefties and righties experience the world differently, but so do ground ball/fly ball guys and guys with speed and guys without.
It’s safe to say that AT&T Park is a bad place to hit and Coors Field is a good place to hit, but parks don’t affect every player evenly and our park factors sort of assume that they do.
In the future, you could imagine a world in which we could know what the average outcome of a batted ball might be (i.e. the average outcome across all 30 parks of that swing is .25 singles, .15 doubles and so on) so that we can compare the observed outcome to the expected outcome, but we aren’t there yet.
This isn’t to say you should ignore park factors. The park factors we have and use are much better than pretending all 30 parks play evenly, but you have to be aware that in some cases the numbers we use aren’t going to make the right corrections. For example, a right-handed hitter who spends 81 games at PNC Park is going to hit fewer HR than if he played at Great American Ballpark on average, but if it’s a righty who happens to have more power the other way that to his pull side, the PNC park factor is actually going to overcompensate.
It’s a tricky business and one that requires caution. You really just need to be careful and to look closely if you think something looks funny. The parks play differently and we need to pay attention to that, but we also have a long way to go before our estimates are perfect and we can say for sure exactly how much of a boost or deduction is necessary.