Let me say it again: I love projections. They’re a lot of fun, and I think making more-and-more accurate projection systems is a pretty cool goal. But, from a decision-making perspective, their utility is a little limited. Sure, they’re helpful for assessing player value and production when it comes to potential free agent signings, or trades. But the array of players available for acquisition is generally not huge, especially when viewed in the context of a team’s available resources. A team is, for better or worse, stuck with many of its players. Knowing the projection for those players is nice, but it’s not too helpful, unless they’re going to be used in a trade, or pushed aside for other candidates. But, there’s one place where I think projections really shine: team-level decision-making. Especially in this era of tanking and superteams, team projections really help put numerical context around the decision to “go for it” and not endure another lost season, versus leaving the current roster to fend for itself en route to an unsatisfying win total.

The main challenge with the above, though, is that again, most projections are still fed to us as point estimates, without good public documentation of the range of those estimates. Past research has indicated that teams finish with about three wins of their point estimate win totals about half the time, and within five wins of their point estimate win totals about two-thirds of the time. Note that these ranges are less wide and scary than they may seem at first, because they still center around zero. If a team is as likely to finish with 90 wins as 84 wins around a central point estimate of 87 wins, that actually reinforces the 87-win central estimate more than it dilutes it, in my view.

So, here’s what we know right now about various Atlanta Braves 2018 win total estimates:

You’ve got some high, some low, and some stuff in the middle. Seems reasonable. The average ends up being 76 wins, as does the median. But all we have for these are point estimates. The question remains: what would happen if we actually received, or were able to understand, the distributions behind these win totals? That would surely help with planning. It’s one thing to say, “This team is projected to win 76 games, and all we have to go on is that it has a 50% chance of winning 73-79 games, and a 67% chance of winning 71-81 games, and those two ranges are just based on the historical relationship between projections and final records.” It’s another thing to look at a composite distribution built upon the inherent potential variation of the projections for the players that make up that team. And that, that is why I’ve presented distributions of my projections to this point: because I think it helps make the following more intuitive.

Let’s think about position players first. The easy way to set up position players is that every position will get 700 PAs, pitchers will get 330 PAs, and 250 additional PAs will be used on PH/DH duties. (Note: I’ve already done the research to determine the reasonableness of these numbers, so you don’t have to, but be advised that the median NL team had 6,200 PAs last year, so 6,180 seems reasonable. You can use the Fangraphs Depth Charts as a comparison: they have 5,840 PAs without pitchers, which is essentially the same estimate once you throw in pitcher PAs; the only difference is that they spread out catcher PAs more across the bench, but in reality, the Braves probably won’t want to pinch-hit for their catchers too often.)

At this point, I’m just going to transition to only discussing IWAG, because I don’t have distributions for any other projection system. So, if you only care about what Steamer/ZiPS/other systems say, feel free to stop reading here. The table below presents a very optimistic playing time scenario: every player gets 600 PAs or more (catchers aside). This is a best-case, best-foot-forward type scenario. Note that the players below can appear multiple times at multiple positions, just to give an understanding of how each position is being filled.

Again, I want to re-emphasize that these are unrealistically high projections, which use pretty much no bench resources. But, sadly, even under these assumptions, you get a position player group that would only be something like ninth-best in baseball. Of course, that’s not really realistic. (And, you may recall that IWAG also tends to be a slight overestimate on a by-player basis.)

So, what if we change it up, and instead use the IWAG point estimate for “likely” PAs based on injury history and the like, forcing the team to backfill appearances by guys much lower down on the depth chart? Well, that looks like this:

That’s a three-win swing downward, taking the team’s pie-in-the-sky placement of around ninth in MLB, and dropping it to something like 12th or 13th. Still decent, but not nearly as exciting. But, these are still point estimates. We wanted to talk about distributions. So, the below is what happens when you take all the distributions in the position player posts, and run a Monte Carlo analysis with them across these two different ways of allotting playing time.

The labeled values are the percentile estimates at each decile; in other words, this distribution figures the Braves will have a 90 percent chance (again, based just on IWAG projections) of having 17.5 fWAR or more from position players in the realistic playing time scenario. The Braves actually finished with 16.6 fWAR among position players last year, so that’s a pretty nice outcome. Unfortunately, the band of variation is not too wide: last year, 17.5 would have finished 16th in MLB, while 24.5 would have finished 9th among all MLB teams for position player value. (One of the reasons why the band is narrow is because the Monte Carlo assumes player variation is independent, so that most outcomes are a mix of the good and the bad. The very few scenarios where everything goes poorly are instead around nine total wins; the very few scenarios where everything goes super-well are around 33 total wins. Neither is likely to occur, though I do think it’s hilarious that even in this perfect world scenario, the Braves would still end up with less position player fWAR than the Astros actually accumulated last year.)

We can do the same exercise for starting pitchers. Since I already outlined the difference between the two scenarios, I’ll just collapse the information into the same table.

Note the point estimates here, and don’t worry as much about the innings allotments. (Yes, Matt Wisler gets a lot of innings, but as discussed, IWAG is perhaps too kind to him, so even if you think that McCarthy will get more than 60 frames and Wisler will get far fewer, chances are those two things will relatively wash out.) 10.3, the optimistic, five-man-only scenario, is good for about 15th in MLB. The more realistic scenario with a lot of rotation fluidity yields around nine wins, which is just marginally below average (think 16th or 17th in MLB). 888 innings reflects the median innings pitched by a rotation last year; the Fangraphs Depth Charts have the Atlanta rotation accumulating 936 innings, which seems overly aggressive given that only five rotations actually hit that threshold last year. In any case, the Fangraphs Depth Charts also have the rotation accumulating 10.3 wins in those 936 innings; scaling it back down to 888 innings yields closer to 9.8 wins. So, while IWAG may be ahead of the other projections as far as the 2018 group of Braves position players goes, it’s actually less optimistic as far as rotation production, due in large part to having Gohara and McCarthy get relatively few innings in the “realistic” scenario.

Again, we can throw together a quick distribution chart that reflects the results of Monte Carlo analysis on the starting pitcher-specific IWAG distributions.

The 10th percentile outcome of 5.6 fWAR would put the Braves in the bottom five of MLB rotations, or thereabouts. The 90th percentile outcome of 11.6 fWAR would put the Braves at right around the middle (15th). (Once again, the band is relatively narrow due to the lack of any interpendence in player variation as modeled in the Monte Carlo analysis. The very few scenarios where everything goes poorly give the rotation a historically bad fWAR of under 2 wins below replacement; the very few scenarios where everything goes super-well are around 15 total wins. Again, neither is likely to occur. The Braves probably won’t have the worst rotation pretty much ever, but it’s again telling that even in a very favorable scenario, the rotation is highly unlikely to end up better than 5th or 6th in MLB.

I’m not going to talk about the bullpen much, because it’s the bullpen, but the table and percentile outcome chart are below.

This was discussed in the bullpen projection post, but the range of outcomes for the bullpen is probably highly variable. Still, with everyone healthy, we’re talking a 12th-best bullpen or so in the optimistic scenario, and a 17th-best bullpen or so in the realistic scenario. Notably, the Fangraphs Depth Charts only have this ragtag bunch getting about 2.3 fWAR over 522 innings; if you scale it up to 551 you get around 2.4, so somewhat lower than the realistic scenario via IWAG.

This bullpen could probably use some shoring up, methinks. The range is something like 29th in MLB (10th percentile) to 14th-ish (90th percentile). That actually tracks the rotation pretty well, suggesting that essentially the pitching staff as a whole has some serious downside risk but not that much upside as far as vaunting towards the top of MLB. (The extreme ends of the ranges are -3.5 fWAR to over 6 fWAR, essentially from the worst in MLB to seventh or so. But again, not too likely to happen, yadda yadda.)

Pencils Down, Final Answers

Now that we have the above percentile distributions, we can simply sum them together, and add the adjustment for replacement level to create a likelihood of team wins based on these forecasts.

You can also express the baseline (or realistic) scenario as follows – it might be somewhat easier to interpret, and shows the expected normal-ish distribution of outcomes.

So, I think this is all just a long-winded way of explaining what I currently expect the Braves’ 2018 fortunes to be, in a probabilistic manner. Maybe it’s not as easy as just singling out a point estimate, but hopefully it’s more informative.

I also won’t dwell on this long, but I think this expresses why I’m fairly disappointed that the Braves have chosen to essentially sit out and potentially punt another year. If we call 85 wins the contention threshold, this distribution gives the Braves about a 10 percent chance of hitting that threshold this year. But, let’s say that the team added even three additional wins (with a relatively stable distribution): then we’d be in 30 percent playoff chance mode. Maybe 30 percent still isn’t sufficient to care about investing anything, but it’s still agita-inducing, for me, anyway. Five wins added, and there’s a 50/50 playoff chance based on the above, more or less. Maybe five extra wins this offseason was never really attainable, and I’m sure there’ll be plenty of folks lining up to explain exactly why that’s the case. But with so many players going for bargains and low commitments this year, and some spots clearly in need of upgrade (rotation depth, bullpen, bench, corner outfield, someone to provide depth at SS/3B to cover the potential for adverse variation), it just seems like a shame, is all. That’s all I’ll say here, because it’s clear the Braves have other plans. But if the team ends with 80-82 wins (which has about a one-in-four chance of happening per the above), the lack of tracking down those extra wins this winter has the potential be extra-disappointing.

Anyway, 79 wins in 2018, assuming nothing changes. Book it. Or don’t, because that’s not the point of distributions of outcomes! It’s up to you.