The Z Files: Closing Time

We've all heard the narrative closers on bad teams can get saves. However, there's a decent correlation between team wins and team saves, as demonstrated by this 10-year lookback.

By means of a quick review, r is the correlation coefficient, in this case between team wins and saves for the given year. A coefficient of 1 is indicative of a strong positive relationship, -1 is strong negative relationship while 0 means there's no relationship, the numbers are essentially random.

In this instance, it's logical to assign team saves in direct proportion to team wins. For those wondering, the correlation between wins and save opportunities is much less. The average coefficient for saves is .64 compared to .40 for opportunities.

Now the question is, how many saves? The following shows the percentage of saves compared to wins over the same ten-year span.

Interestingly, the number is on the rise, likely a result of fewer complete games. There's also an economic component as the better teams not only have the better closers, but also better setup men, and as just revealed, saves are in proportion to wins.

In terms of the actual projection, a logical starting point for 2017 was projecting team saves for 53 percent of the team's projected wins. Those saves are then subjectively allocated among the club's relievers.

Following this historical data is fine, but part of what we need to do as prognosticators is identifying players with a pathway to veer from the above boundaries. Another narrative surrounding saves is teams tending to play close games are more likely to generate save opportunities. Further, the better a team's pitching, the more likely they are to play close games. On the other hand, teams with strong offenses win more games by a greater than three-run margin, decreasing save chances.

To test these hypotheses, let's first correlate runs allowed to saves.

The negative correlation reflects allowing fewer runs is better. Keeping that in mind, an argument can be tendered that teams with lower expected ERAs may garner saves on more than 53 percent of their wins. Here's the correlation between runs scored and saves:

The takeaway message here isn't that high-scoring teams generate fewer save chances, but rather there isn't any correlation between strength of offense and saves.

Runs scored and runs allowed are obviously only part of the story. Run differential should be relevant. One can construct all sorts of narratives, but the most logical one is teams with smaller run differentials are likely to have a higher percentage of games saved. Note, this isn't raw saves, but percentage of wins resulting in saves. To test this, let's look at the differential for all 300 teams in the sample, then parse the data into positive and negative differential. The correlation will be done in total, not season-by-season.

Absolute value was used in the differential, since we're measuring a proxy for how close the games were. There's some degree of correlation with teams scoring more runs than allowed differential, but it's not strong. Of course, run differential doesn't have to accurately reflect the number of close games, which could be accounting for some of the lack of correlation. Still, it seems there's a large degree of randomness with respect to save chances.

Flipping this discussion to in-season considerations, projecting rest-of-season saves is part of what I do. The conundrum is what to do with closers significantly ahead or lagging behind initial expectations. We're not quite there yet, but for the ease of this exercise, let's say we're at the midpoint of the season and Greg Holland has 28 saves (he has 25 as of this writing). Do I project another 25? Do I figure he'll end with 40 and give him only 12 more?

How about someone like A.J. Ramos, originally projected for 41 saves, but likely to have low double-digits halfway through? Will he get 30 or so to bring his total where initially expected?

Obviously, how a closer is pitching affects the answer, but the larger consideration is how the team is performing relative to preseason expectations. There's no one-size-fits-all manner to adjust, context is key. As such, let's look at Holland and Ramos, along with a couple other intriguing ninth-inning stalwarts.

Greg Holland, Colorado Rockies

Colorado has 47 wins with 27 saves, a 57 percent clip. This is higher than the average, but supported by an ERA better than league average. The club has 25 road wins with 16 saves, a 64 percent mark. This is very high, but keep in mind the Rockies staff totes the second lowest road ERA in the league. According to the above data, teams with a lower ERA can have a high percentage of wins saved, though 64 percent feels excessive.

What's done is done. We need to surmise what's likely to transpire over the second half and how that will influence Holland's saves. Something to keep in mind is the Rockies staff is young, so their innings may be managed. On the other hand, they're deep and have Jon Gray returning. Maybe there will be some tailing off, but it certainly appears Colorado will be in the playoff picture so I'd expect the brass to distribute innings among the masses, perhaps taking a page out of the Dodgers playbook and using the 10-day disabled list as their 26th man.

Using Bill James' Pythagorean theorem, the Rockies are about four wins ahead of what's expected. That is, if they score the same number of runs while giving up the same number, chances are they'll win a couple fewer games.

It seems to me Holland's pace of saves should wane a bit over the second half. Barring an unlikely collapse from their staff, the team should continue to play well, producing a lot of save chances. They're pacing for 103 wins. Let's be conservative and say they finish with 95, meaning 48 more victories. Now let's drop the percentage to 55 percent of wins, still higher than the league average, but a tick below the current 57 percent mark. That yields 26 more saves for the club.

To this point, Holland has 25 of the 27 team saves. Allowing for a similar ratio the rest of the season, projecting Holland for 24 more saves seems reasonable. This makes the leap of faith he remains healthy, not at all a foregone conclusion.

A.J. Ramos, Miami Marlins

The Marlins won 78 games last season, I expected a similar total in 2017. I projected their pitching staff to be a little better than average, so the math was 78 x .54 = 42 saves, 41 of which were assigned to Ramos. For what it's worth, Ramos has earned all the Marlins saves so far.

Despite not pitching as well as expected, Ramos low save total isn't a result of blown saves since he's 10-of-11 for the season. He's just not getting chances. To wit, 32 percent of Miami wins have been saved. Putting that in perspective, of the 300 data points the past 10 years, only one club saved a fewer percentage of wins – the 2006 Cleveland Indians. In fact, only 16 times was there a mark below 40 percent. The bottom line is even if the Fish continue the same pace, Ramos should see considerably more save chances.

The Marlins are pacing for 74 wins, which seems reasonable. That's 42 more. Using the league average of 53 percent, Ramos can be projected for 22 saves. However, if he doesn't improve his walk and home run rate, he's likely to blow a higher percentage. The result will be a team mark below 53 percent, but that's fine. Projecting Ramos for 18 more saves feels right.

Craig Kimbrel, Boston Red Sox

To date, Boston is 40-32. The curious thing is they haven't played their best ball yet. This isn't just a member of Red Sox Nation opining, but the consensus around baseball. They're on pace for 90 wins so earmarking the squad for 95 is within reason. That entails 55 more victories.

Of the current 40 wins, 21 have been saved, a tidy 53 percent. Assuming the same rate going forward, the Red Sox should garner 29 more saves. Kimbrel has recorded 20 of the 21 saves, the lone wolf going to Ben Taylor and was of the three-inning variety. Allowing for something similar, or even just being unable once or twice the next three-plus months, projecting 27 more saves for Kimbrel makes sense.

Kenley Jansen, Los Angeles Dodgers

Jansen is coming off a 47-save campaign. His ratios are better than ever and he's currently working on a record-setting streak, walking zero, zilch, nary a hitter this season, while punching out 50 in 29 2/3 innings. All this is well and good, and fantasy useful, but he's only saved 15 contests, equating to 32 percent of the club's wins.

The Dodgers check in with the league's best ERA, which should lead to more save chances. However, they're top five in runs per game. The data presented earlier suggested a correlation, albeit not strong, between run differential and percent of games saved. Still, the Dodgers own the largest positive run margin in the league, so chances are this has cost Jansen some save opportunities.

When you're at the extremes like the Dodgers sit so far, there will likely be some movement towards the middle. That said, there's nothing fluky about their hitting or pitching so Jansen's saves should continue to lag initial projections.

The squad is looking at 104 wins at the current clip. Let's temper that to a still-impressive 98, yielding 51 more victories. If Jansen saves 45 percent, he'll add 23 more saves to his ledger.

With many closers, the procedure entails considering injury risk and job security. However, the major point is there's a great deal of method to the madness. Will the results be perfectly accurate? No, of course not, but that shouldn't sway one from applying logic to the process.

Before calling it a day, holds are a story unto themselves. Yes, there's correlation between wins and holds, but not nearly the extent as saves. Some reasons are bullpen usage and quality of starting pitching. Some managers play the platoon matchups more than others. Clubs with pitchers working deeper into games avail fewer opportunities for holds. Projecting holds isn't throwing darts, but there's certainly more subjective analysis within team context.