I have nothing to add in regard to Dustin May that hasn’t already been said. A plan and direction seem to be emerging. I will continue to watch, turn the pages and see how the story unfolds. In the meantime, I’ve tried to assemble more mental floss as the off-season slowly unfolds.
This is a continuation of a series of Roster Construction exploration articles I’ve been developing. Here is the original, if you missed it. This is part three and is a deeper exploration of the roster construction approach
I will think of as “Roster Efficiency”. That is just a name, but it portrays an underlying narrative that certain teams (typically small market) are finding success by using all parts of their roster more efficiently, particularly with smart match-up platoons and the strategic use of versatile players to maximize the skills the roster has.
“Roster Efficiency” is ostensibly a way to create a competitive team without having to have 9-10 “everyday” Starz players, and without having to invest untold millions (billions?) of dollars in acquiring the unicorn players (Ohtani, Judge, et. al.) that is beyond the reach of many teams. And, of course, we have the Dodgers, who appear to be applying all 3 roster construction approaches simultaneously (acquire the unicorns, acquire a bunch of StarZ, AND use the roster efficiently to fill gaps and offset weak spots).
In the reader comments from the first two articles, commenters proposed that the 2025 Brewers and the 2021 Giants teams would be good ones to start with as candidates to explore this “Roster Efficiency” approach. Canaries in the coal mine, if you will. So I did. I used their story to try to identify the breadcrumb trail that would help me find other teams statistically through queries.
Let me remind myself of the parameters set in the first two articles:
- A solidly built roster will have at least 8-9 players that can reasonably be expected to deliver “Starz” level performance. This roster can be projected to win 90 games. It’s not guaranteed, but this appears to be the threshold where team roster construction moves beyond “crapshoot” into one that is truly “competitive”.
- The mix of pitchers to hitters that are StarZ can vary, team by team, so how a team builds to 8-9-10 Starz is flexible.
- StarZ players range from minimum 2.7 WAR and up, and as a group must average 3.8 WAR.
- StarZ pitchers range from minimum 1.6 WAR and up, and as a group must average 2.7 WAR.
- On a 90 or more-win team, these Starz players will produce, on average, ~ 75% of the teams’ total WAR. Here, the 80/20 rule is more like the 75/25 rule.
I used these parameters when looking at the canaries and discovered some interesting tidbits. Some of the tidbits expand and strengthen the StarZ model with new rules of thumb that seem obvious in retrospect but are now quantified. Some of the tidbits just leave me scratching my jaw. There is something here, but I haven’t grasped it yet.
Upon closer examination, the 2021 SF Giant indeed stick out. At first glance, they looked like a team that fit the StarZ model. 9 players that produced top 20th percentile results. There were (5 (!) StarZ pitchers (Webb, Gausman, Cueto, DeSclafani and Wood). While the rotation would have looked solid at the outset, I did have to ask myself if the Giants Front Office personnel could reasonably have expected 5 Starz performances here. DeSclafani and Wood both produced StarZ level output 4 or more times in their career, so the talent was there, but health was probably an unexpected outcome. On the player side, Posey, Crawford, Belt and Ruf produced StarZ performance. Posey, Crawford and Belt were still solid, if aging, players. Could anyone reasonably have projected Darin Ruf to produce virtually all of his career WAR at a StarZ level (2.9 WAR) in a single age-34 season? Indeed, at season outset, this roster construction would have looked like 6-7-8 StarZ players, would be 50-50 on play-off expectations and fit neatly into the “Crapshoot category”. Yet they went on to win 107 (!) games. How? First, a solid rotation that experiences no health problems is a great start. Difficult to anticipate, but it happens. Second, this team made highly effective use of platoons, leveraging the skills of an aging Posey, Longoria and Belt in ways that maximized their effectiveness. Guys like Wade, Ruf, Estrada and Solano filled in gaps capably, all producing 1 or more WAR. In total, this team only produced 63% of its’ WAR from the StarZ players, more than 1 standard deviation off the mean of 76%. That gave me my first clue on how to look for “Roster Efficiency” teams. Who else won 90 or more games and finished with a StarZ WAR contribution significantly less than average?
A few more of these teams appear in the last five years (2021-2025) as I see eleven teams in 5 years. In the prior eight years (2012-2019), I see eleven teams over the eight year span. A trend? Or is Milwaukee just an outlier organization over the last five year? I don’t know.
This isn’t rare, but not common. On average, about 1 or 2 layoff team per season will be one of these “Roster Efficiency” types of construction, although it isn’t that linear. Looking at the win column, these teams don’t just squeak in, though. Note the teams that appear in this table more than once. Tampa Bay and Milwaukee are repeat contributors. Am I surprised? Milwaukee takes the prize with producing what I labelled “deviant WAR” 4 out of the last 5 seasons, in that significantly more WAR comes from their non-StarZ than most playoff teams, most years. Significant as in more than 1 standard deviation from the mean. They truly appear to masters at maximizing the whole roster to produce wins.
I noticed something else. The 2021 Giants team had absolutely zero (ZERO!) players finish in the bottom 20th percentile. Everyone contributed positively. I wondered how rare this is, so I queried for this over the 2012-2025 time frame. Reminder, my original queries excluded the bit players (less than 100 PA’s, less than 20 IP). I charted the outcomes (below) and think I have formulated another rule of thumb on roster construction, which is something like … good teams can’t afford to have more than 1 player who plays more than occasionally AND contributes in the bottom 20th percentile. Even just two of these guys can set a good team back to Crapshoot. Given my relatively low bar for playing time (100 PA’s, 20 IP), it suggests that “Roster Efficiency” team must be super-aggressive in churning out players who aren’t producing. I think of this as the inverse of “runway”.
Notice the bottom right of the graph. Teams with zero or one replacement level player achieved win totals of 87 or more and made the playoffs quite a bit more reliably than everyone else.
After the Giants, I looked at the 2025 Brewers. What do I see? First off, I again see 6 (!) pitchers with StarZ level performance (Peralta, Woodruff, Priester, Patrick, Uribe, Megill). They also had 4 StarZ hitters (Turang, Contreras, Frelick, Churio), plus 2 guys that barely missed (Durbin, Collins). This team won 97 games but got a remarkably low 56% of their WAR contribution from Starz players, almost 2 standard deviations from the mean.
Just because I’m here and I see the data, I will note that even in Milwaukee, the contract plays. Rhys Hoskins was an everyday player (by my standard of playing 80% of available games) in the first half of the season, until getting hurt in early July. He was largely ineffective, but $18m contracts provide lots of runway even in Milwaukee, it seems. Just a sidebar. Back to roster construction.
So, could one have projected this team as a 90-win team heading into 2025? The results were highly dependent not so much on role players/mix-match, but on first full-year break-out performances from Priester, Patrick, Durbin and Collins, and Vaughn replacing Hoskins in the line-up second half of the season (to say nothing about flash-in-the-pan Misiorowski). Otherwise, they had 8 everyday players (using my benchmark of 130 games played, or 80%). They moved them around a bit, but it was the same 8 players for the most part. Vaughn/Hoskins was really the only combo. This wasn’t any advance mix-and-match outcome, it appears due to injury primarily and also performance, more than optimizing match-ups. I’m not overlooking the phenomenon that they are getting great results from non-StarZ players, but noting a reality that their core of StarZ players is quite good. The good bench (ie. and maximized roster) probably carried them from a baseline expectation of 90-92 wins to 97. What a great hedge!
As a side note, I notice that many of their break-out performances (I’ll include Uribe in this group) were from prospects of the older variety (25 and up). Just a note for Cardinal fans who might bemoan the stalled development of players and lack of break-outs such as Walker, Gorman, etc. Milwaukee appears to be quite good at capitalizing on other teams lack of patience with prospects and being patient with their own. It is not difficult to notice that MIL is leveraging players into prominent roles at an age where the Cardinals are looking to deal them away.
I also notice a repeat occurrence. Milwaukee, like San Francisco, has zero bottom 20th contributors in their 97-win season. Two (out of two) isn’t a deep enough data set, but I do wonder if the “Roster Efficiency” model of roster construction means being ruthless about not allowing replacement level performances to linger very long (and having the players that can improve upon that). Something like … if you don’t have that unicorn, or you don’t have quite enough Starz players, you can leverage your 5 bench players to create a competitive team, but absolutely none of the bench players can be replacement level. It’s a theory at least, needing to be proven with more queries, I suppose. I am also curious if this is a new phenomenon, or if I will see similar outcomes in earlier eras of baseball.
