The Role of Context in Determining the Best by Dave Cameron September 22, 2015 Here’s a statement I think most people would agree with: Bryce Harper has been the best player in baseball this year. Harper leads the majors in batting average, on base percentage, and slugging percentage, so naturally, he also leads in wOBA, wRC+, and just about any other offensive metric you can find here on FanGraphs. In most cases, it isn’t even close; his 205 wRC+ is 30 points better than the next best hitter (Joey Votto), and no one is within +1.5 WAR of his current total (+9.7, with Josh Donaldson’s +8.1 coming in second). Harper is having one of the best offensive seasons of all time, and while some other guys are having excellent years as well, no one is really performing at Harper’s level this year. Now here’s a statement that I’m guessing would be a bit more controversial: the Washington Nationals are having the best season in the National League East. The Nationals are first in offense among NL East teams (4.5 runs per game) and are second in the division in pitching and defense (4.0 runs allowed per game), so they lead the division in both run differential (+78) and expected run differential based on BaseRuns (+82). Led by Harper’s monster season, the Nationals are tied for the fourth best BaseRuns record in baseball, dead even with both the Pirates and Cubs. The Mets are also having a good year, but trail in both actual run differential (+62) and expected run differential (+69). Of course, the Mets also happen to have a 6 1/2 game lead over the Nationals in the standings, which is the metric of choice for most baseball followers when determining which team has had “the best” season. It’s the only number that gets tallied up when postseason berths are being handed out, after all, so in the end, it’s the only number that really matters. Which is why I’m guessing that a good number of people — probably including everyone actually playing for both the Mets and Nationals — would disagree with the idea that the Nationals have been the best team in the division this year. I bring up these two statements because I’ve been thinking about context a lot lately. Over the last few weeks, both Jeff and I have written multiple posts about the role of sequencing in a team’s results, and how large an impact the order of events can have on the outcomes of games and seasons. And I’ve found it interesting to see how different the reactions can be when you suggest using context-neutral or context-included numbers to evaluate players or teams; for reasons I don’t totally understand, many in the baseball community seem to have settled on using context-neutral numbers to evaluate players but context-included numbers to evaluate teams. For instance, if you missed Dave Studeman’s piece over at The Hardball Times last week, his post suggesting that the timing of events should be included in the MVP discussion generated 156 comments (as of this post getting written, anyway), about 10 times the average of a normal THT piece. Many of those comments pushed back against his assertion, believing he was just trying to stir the pot and write a “hot take” piece, arguing for anyone other than Bryce Harper as MVP. The context-neutral stats are so strongly in Harper’s favor that it’s pretty common to see people suggest that there’s no credible argument for any other player, and the fact that Harper has been relatively terrible in clutch situations shouldn’t be considered a factor in the awards race. Which is a perfectly rational belief. After all, Harper can’t really control who gets on base in front of him, or what the score is when he comes to the plate, and if his teammates sucking meant that he ended up hitting a lot of solo home runs or launched a bunch of bombs in games that were already decided in favor of the opponent, well, that wasn’t his fault. So, instead of arguing over the nebulous idea of how to define “valuable”, many prefer to just give the award to the best player in the league, based on metrics that assign a single run value to every type of event, and just count up the number of events that player contributed to his team’s total without regard for when those plays happened. But the reality is that part of the reason the Nationals have underperformed their BaseRuns record by six wins is because the distribution of Harper’s events have skewed more towards situations where his hits don’t have as significant an impact on the outcome. In low leverage situations, he’s running a 213 wRC+, #1 in baseball. In medium leverage situations, he’s running a 218 wRC+, #1 in baseball. In high leverage situations, he’s running a 117 wRC+, #70 in baseball. Harper’s 2015 season is the player-version of having a great run differential but a less great win-loss record; he’s destroyed everyone else in the quantity of positive events he’s been involved in, but he’s distributed those events in an inefficient manner, leading to fewer overall wins for his team that you’d expect from his raw batting line. This isn’t meant to denigrate Harper’s season; even with leverage impact taken into account, he’s still having an amazing year, but when you add the order of events into the picture, it’s a bit less amazing than it looks from a context-neutral perspective. So, every year, we end up having this same tired debate about “the best” versus “the most valuable”. I don’t really want to get into that debate again, honestly. I’m tired of having it, and I don’t find it particularly productive. But I do think it’s interesting to point out that the method for evaluating “the best player” seems to be quite a bit different from the method for evaluating “the best team”. At the team level, wins and losses are the currency, with everything else being seen as a peripheral, or a way to evaluate what might have happened most often had the season been played 1,000 times. But as I’ve heard repeatedly throughout the years, when it comes to awards and retrospective analysis, we are only supposed to care about what did happen, not what should have happened. But when writers apply that same line of thinking to the MVP discussion, the tenor of the discussion seems to change pretty dramatically, even when it comes from a well respected member of the statistical community. And I’m not really sure why that is. Why do we have such a strong preference for all-events-are-equal-and-context-is-noise metrics when evaluating players, but then tend to move towards sequencing-is-a-thing-that-happened-and-should-be-accounted-for numbers when evaluating teams? Or, I could even ask it another way; why is there still a strong preference for context-included pitching metrics like ERA when determining the Cy Young Award — which doesn’t have any vague “value” terminology associated with the name — for the best pitcher, but we tend to still want to evaluate hitters by context-neutral numbers like wRC+ and WAR? Last year, I actually was an MVP voter, and had to wrestle with this question in order to fill out my ballot. I ended up deciding that, for a retrospective award simply trying to determine how much each player contributed to his teams wins and losses in that season, that context did matter; hitting a three run homer is better than hitting a solo homer. So I did include the effects of the order of events in my MVP balloting, even though it led to me being the low voter on Andrew McCutchen — I had him sixth — even though he’s one of the best players in the game, and one of my favorite players to watch. As I expected, I received a lot of pretty negative feedback for that pick, as will anyone who doesn’t put Bryce Harper first on their MVP ballot this year. Because we’ve become conditioned to evaluate hitters by the numbers that treat all events the same — which is the correct thing to do when trying to determine a player’s overall talent level, or to project his future performance — and Harper is running away with the context-neutral numbers, it seems like a good portion of the baseball community won’t even consider arguments that perhaps a guy like Anthony Rizzo (who has been much better in high leverage situations) has had a greater impact on his team’s win-loss record this year. I guess what I’m suggesting is that we strive for logical consistency. If we’re going to say that context-neutral metrics prove that Harper has been the best player in baseball, let’s also use context-neutral numbers to evaluate team performance. But if we’re going to judge a team’s season on the numbers of wins and losses they accumulate, maybe we should also be willing to consider the situational impact of a player’s performance on his team’s record, since we know that sequencing is a pretty big part of a team’s wins and losses. If I had an MVP vote this year, I’d vote for Bryce Harper. But I think it’s reasonable to consider the situational performances of the various candidates, and to not shout down those who might consider voting for a guy like Rizzo because of the fact that he ended up hitting better in more important situations. We can acknowledge that the ordering of events has an impact on the final results without buying into narratives about whether a guy had the right mentality to live up to the pressure or whatever other post-hoc character judgments have been made about players with lower RBI totals before. The sequence of events matter to the results. 1B-1B-HR is better than HR-1B-1B on the scoreboard. There’s a huge place for context-neutral metrics when it comes to evaluating and forecasting player’s abilities, but when it comes to past performance, I do think we should be willing to consider the situations in which their events occurred. If we’re going to judge teams by win-loss records, then maybe we should also judge past player performance (for the purposes of handing out awards) by the kinds of metrics that more directly correlate to wins and losses, and those are the ones that include context.