In the wake of the Marlins blowing up their roster last week, Eno Sarris took the dispassionate analysis route, noting that the Marlins go-big-or-go-home roster construction strategy has helped them win a couple of World Series titles, and they’ve avoided the trap of perpetual mediocrity that has ensnared other smaller revenue franchises. Putting the Marlins aside, his key paragraph is this one:
One of the main tenets of the statistical approach to baseball is honesty. Numbers can help see past any mystical optimism into the stark reality of a team’s competitiveness. How much the Marlins actually depended on statistical forecasts in their decision-making is debatable, but one thing is clear: they knew when they had a shot at winning, and they knew when they didn’t.
There are plenty of ways to spin their approach more negatively. Pump and dump. Boom and bust. Fun, then fire sale.
But if your team is not going to be competitive, why keep high-priced assets around?
This is a pretty common line of thinking. For many, the idea of winning 80 games and finishing out of the playoffs is no better than winning 65 games and finishing in last place. The value of a win is clearly not linear, and for many, every win before 85 or so is basically irrelevant. Some even view additional wins that still don’t get you into the playoffs as a net negative, as they result in lower draft selections, or they can “trick” an organization into prematurely giving up future assets to make a run with a roster that isn’t quite good enough to stand as a real contender.
Here’s my problem, though. Eno notes that teams should be honest with themselves about their expectations for the following season, but because of the massive amount of uncertainty in forecasting, an honest assessment of nearly every team’s 2013 forecast is that they could be good, bad, or somewhere in between.
To be clear – this isn’t a knock on the quality of publicly available projection systems, or any kind of criticism of their methodology. It’s just an acknowledgment that any team’s final record is heavily dependent on factors that simply cannot be forecast ahead of time.
Let’s just say, for instance, that we had a perfect player projection system. Dan Szymborski, Jared Cross, and Brian Cartwright all got together and found the holy grail of baseball, figuring out precisely how to identify each player’s true talent level, and they perfected aging curves precisely for every kind of player. So, heading into 2013, we knew exactly what each player was going to produce in terms of hits, walks, strikeouts, home runs, etc…
From that, we could accurately forecast each team’s projected wOBA, both for and against, and quickly identify which teams were better than the others. However, to get from wOBA to wins, you need two things: the sequencing of individual events that lead to runs, and the distribution of when those runs occur within certain games.
For instance, last year, both the Reds and the Giants had team wOBAs of .314. The Giants scored 718 runs while the Reds score 669, a difference of 49 runs. The A’s and Royals both had .311 wOBAs, but Oakland outscored Kansas City 713 to 676, a 37 run gap. The Braves and Indians both had .310 wOBAs, and yet Atlanta scored 710 runs, while Cleveland managed just 667. In some of those cases (ATL/CLE especially), the differences can be partially explained by baserunning — which we could theoretically account for with our Perfect Projection System (TM) — but even after removing that aspect and equalizing the number of plate appearances, there are still 30 to 50 run gaps between teams that were essentially offensive equals based on on their raw totals. These gaps are mostly about the distribution of when those hits occur, and that’s just not something we can project.
The same, of course, is true on the pitching side of things. In the American League last year, the average team allowed a .311 wOBA with the bases empty, .324 with men on base, and .318 with runners in scoring position. The Indians pitching staff was basically league average with no one (.311 wOBA against), but was the worst in the league with men on base (.357 wOBA against) and with runners in scoring position (.347 wOBA against).
Now, compare that distribution to the Blue Jays: .331 wOBA against with the bases empty, .341 with men on base, and .331 with runners in scoring position. Overall, opposing hitters actually did better against Toronto’s pitchers than against Cleveland’s, but because of when those hits occurred, the Indians allowed 845 runs, while the Blue Jays gave up only 784. Again, even adjusting for total batters faced and baserunning, we’re left with a 50 run margin in results that simply couldn’t have been forecast. There’s no way we can know in advance how well teams are going to cluster their hits together, or the timing of the hits they’re going to allow. And these sequencing patterns have a significant impact on runs scored and runs allowed.
And that only gets us from wOBA to RS/RA. That still doesn’t bridge the gap from runs to wins, which gives us another layer where sequencing can have a huge impact but essentially can’t be forecast. Most forecasting systems use some version of the pythagorean expectation to get from runs to wins, but as anyone who has watched baseball knows, team records often vary from their pythag record by significant amounts. Last year, the spread just on pythag was 16 wins, ranging from the Orioles (11 wins more than expected) to the Cardinals, Red Sox, Diamondbacks, and Rockies (five wins less than expected). The Orioles were an historic outlier, but the spread itself wasn’t abnormal – the average of the 2009-2011 spreads was 15 wins per season.
Just one standard deviation on pythag last year was +/- four wins, and that variance is based on knowing runs scored and runs allowed. When you factor in the sequencing gap on those events, your standard deviation grows to something closer to eight wins. And that assumes that you have perfect forecasts, both in terms of performance and playing time.
Which, of course, no one has. We can make pretty decent guesses about player performance and slightly less decent guesses about player health, but these forecasts aren’t anywhere close to being perfect, and one team isn’t a large enough sample for all the missed forecasts to come out in the wash. Some teams just over-perform their true talent levels for six months. Other teams are destroyed by injuries or sell off their best players at the trade deadline, and these are things we basically can’t predict in advance.
Realistically, when you add up the uncertainty around the projections themselves, the wide variability in team health, and then the effects of mostly random sequencing on events that lead to runs and wins, we’re left with a team forecast that can’t really be any more precise than some projected mean, plus or minus at least 10 wins, and quite possibly more. If we think the Marlins were an 80 win team before the big trade last week, then what we were really saying is that they were likely to finish with somewhere between 70 and 90 wins, and maybe more like 65 and 95.
Of course, some results are more likely than others, and teams shouldn’t see possibility and probability as the same thing, but if we’re going to suggest that teams build their off-season plans around an honest assessment of their expectations for the following year, the most honest assessment is that no one really has much of a clue what’s going to happen next year.
There are exceptions at the extreme – the Astros are almost certainly going to be a losing baseball team next year, and now that the Marlins have torn down their roster, they’re probably going to lose more than they win too. Betting against either one to make the playoffs is probably a safe call at this point, though of course the 2012 Orioles and Athletics should remind you to not even be too sure in assumptions that look relatively safe.
But, for the large majority of teams, we simply don’t have enough information to say that a team is definitively going to have a winning season or not. We certainly didn’t have enough information to know that the Marlins were going to be lousy again next year, and that goes for probably 25 of the teams in Major League Baseball. If a team is going to punt the present in an effort to get future wins, they have to acknowledge that they’re giving up some real chance at a playoff run and the revenues that go along with playing meaningful baseball in September, and they’ll have to get enough of a long term reward to justify losing on purpose.
And, contrary to popular thinking, the calculation isn’t as simple as “playoffs or no playoffs”, with every win in between having zero value. The greater population of fans simply don’t see teams in that light, especially on the losing side of that spectrum. The Astros decision to blow up the team and start over was almost certainly correct, given the talent base they were starting from, but the result was a loss of 5,700 fans per game in attendance, and it’s not like the 2011 Astros were any good either.
From 2007 to 2010, the Astros averaged 77 wins per season and never finished higher than third in the standings. During those four years, their attendance dwindled from 3.0 million fans per season to 2.3 million fans per season. In the last two seasons, where they’ve averaged 55 wins per season, their attendance went from 2.3 million down to 1.6 million. In other words, they lost as many fans in two years of being atrocious as they did in four years of being mediocre.
There’s a real price to be paid for being awful. The casual fan wants to know that his team at least has a fighting chance to put on a good show, or else he’ll simply substitute into some other kind of entertainment until the product improves. In terms of selling tickets, there is a demonstrable difference between 55 wins, 65 wins, and 75 wins. And the more tickets you sell, the more revenue the team generates, the more money the baseball operations department has to spend on the roster.
I’m not advocating for roster construction strategies that refuse to admit the obvious, throwing good money at mediocre veterans in order to try and prop up a roster that has too many leaks to plug at once. That’s what got the Astros into their current mess to begin with. There is a point at which it makes sense to just blow the roster up and start over. But that point is when you get enough long term value in exchange for your short term assets to make the trade-off worth it. You shouldn’t just implode your team on purpose in an effort to be bad simply for the sake of earning a higher draft pick.
That works in the NBA, where a bad season can land you Tim Duncan or LeBron James, changing the entire outlook of your franchise with one single draft pick. That works to some degree in the NFL, where a top pick can make an instant impact on the field in the very next season, and high draft selections can be flipped for valuable contributors who can overhaul a roster in short order. In baseball, the effects of a high draft pick aren’t seen for years, and the likelihood of landing a franchise player with a high pick is much lower than it is in other sports. The value of simply picking near the top of the draft in baseball simply isn’t as large as it in basketball or football, and it’s not a good enough reason to blow up a mediocre baseball team.
I see this argument a lot as it pertains to teams like the Mets, with valuable walk-year veterans in R.A. Dickey and David Wright. We saw it last year when the A’s signed Coco Crisp to play center field. The perceived value of putting a respectable team on the field is quite low, but as the A’s showed last year, the actual value of doing just that can be extremely high. We simply don’t know enough about the future to say that Dickey and Wright aren’t going to be part of the next competitive Mets team. We do not have the forecasting capabilities to look at a 75 win team and tell them that they can’t be a 90 win team in the following year.
There’s too much variation in baseball for teams to simply accept their most recent record as evidence of their short term future. There’s too many things that simply can’t be projected — and too much uncertainty around the things we do know — for more than one or two teams per year to simply punt the entire season and lose on purpose. Trading from the present to improve the future is one thing; trading from the present simply because we see no future is another thing entirely, and requires a level of certainty in forecasting that we simply don’t have.
If our forecast for a team is 65 to 90 wins, then making smart moves to improve the roster and increase the likelihood of getting towards that 90 win part of the bell curve could very well be a better move for the future of the franchise than blowing up the roster and accepting the ramifications of a 65 win season. Rebuilding can be the right path to take, especially if you get the kind of offer for your veterans that can inject a real talent boost into the organization. Just dumping good players because there’s no point in having veterans on a team with a mean forecast of 78 wins, though? That’s just putting too much faith in what we know. If we’re going to stress honesty, let’s honestly admit that we don’t know enough to suggest that a 78 win team should give up hope and lose on purpose.