Brandon Marsh and the Anatomy of an Outlier

This week both Fangraphs and Baseball Prospectus put out there season standings predictions, and both came out fairly similarly on the Phillies predicting 85 and 84 wins respectively. It is dangerously close to the Mets and a large distance behind the Braves. For a team that won 90 games, changed pretty much nothing, and has made the NLCS the last two seasons, this does feel like a slight against them, even if projection systems are notoriously conservative. I want to dive into the projections more later, but there is one player whose projection is at the heart of both of these, and is a good way to talk about projection systems and outliers, and that is Brandon Marsh.

Before going into Marsh, I want to talk about baseball statistics and really my taxonomy of how to view a statistic. In general I view them as falling into 3 broad categories; stats that tell you what happened, stats that tell you what should have happened, and stats that say what should happen in the future. Since we are talking about a hitter, a stat that tells you what happened is your base triple slash line or a composite stat like a OPS+ or wRC+. Some might have different weightings and factors on how they value certain events, but ultimately the actual event will drive the number. What should have happened are the expected stats on Baseball Savant or as will matter for our discussion DRC+ which is Deserved Runs Created, a stat from Baseball Prospectus that tries to control for the various factors of an event and then say what should have happened. What should happen in the future is often applied to stats applying what should have happened in the past, but really what I am referring to here are stats such as PECOTA, ZiPS, Fangraphs projections, Steamer, or other season projection stat. These may incorporate our other stats as measures for audience consumption, but ultimately they are a projection.

On the surface, Brandon Marsh’s stat line for the 2023 season looks great. The 25 year old outfielder hit .277/.372/.458 in 472 plate appearances. He has a 125 wRC+ and was worth 3.4 fWAR according to Fangraphs, and Baseball Reference had him at 127 OPS+ and 3.4 bWAR. If you went one level deeper, alarm bells likely went off. Marsh had a 30.5% strikeout rate, which was 10th highest among 168 batters with at least 450 plate appearances (I will be using 450 PAs as my criteria for all leaderboards mentioned unless stated otherwise). He had a .397 BABIP, which was 1st among batters, 27 points ahead of Freddie Freeman. This combination of strikeouts and perceived luck on balls in play, would be his .277 batting average into strong question. This is where DRC+ comes in. Due to those numbers Baseball Prospectus’s metric saw Marsh with DRC+ of just 89, driving his WARP to 0.5.

Given that projections are based on past events for the player, include historical trends, and tend to normalize luck, it is unsurprising that Brandon Marsh comes out poorly on projections. PECOTA thinks that maybe it was slightly too harsh on him last year and comes in with a projected DRC+ of 91, a slash line of .227/.316/.374, and a total of 1.2 WARP in 477 plate appearances, with a projected BABIP of just .310. ZiPS is slightly more optimistic, coming out at 1.7 fWAR with a 94 wRC+ and slash line of .245/.324/.390 as it regresses his BABIP all the way to .338.

Whenever you have a data model, there is likely to be outliers. Brandon Marsh may not be the most outlying of outliers, and he is likely to regress, but there are some things about Brandon Marsh’s numbers that indicate he is not normal.

We can start with the strikeout rate. Historically, a high strikeout rate has been a proxy for bad decisions at the plate, poor ability to make contact, and swinging away. Marsh falls much more into the modern three true outcomes model, except we will get to the batted ball part later. He walked at a 12.5% rate, so he definitely was not up at the plate hacking. To put some of this into context, here are his swing related metrics in relations to the other left handed hitters in the starting lineup. (Stats from Fangraphs leaderboard using Pitch Info’s tracking, 168 qualifying).

	Marsh Rank	Brandon Marsh	Bryson Stott	Bryce Harper	Kyle Schwarber
O-Swing%	70th	27.1%	30.9%	33.8%	22.6%
Z-Swing%	18th	57.9%	56.1%	75.1%	53.3%
Swing %	25th	42.2%	44.5%	51.1%	37.6%
Z-Contact%	76th	86.7%	90.9%	81.8%	78.9%

There is a whole discussion on Bryce Harper, that Rob Orr of Baseball Prospectus already wrote about, so let’s just say that Brandon Marsh and Bryson Stott should swings at strikes more. Overall, we see a swing profile that has some more zone expansion than Schwarber, but more contact ability. This tracks with the walk and strikeout rates we see for Marsh. His plate discipline looks like it is a plan and not a true deficiency. This tracks with his progression. He swung at pitches 5% less in 2023, both in zone and out of zone, and saw a 5% rise in his zone contact rate as well. If he keeps the same approach and does not increase his zone swing rate, he is likely to have a high strikeout rate again, but as we can see with the Harper and Stott examples here, a higher zone swing rate or high zone contact rate could see a dramatic drop in overall strikeout rate (also likely a drop in walk rate). In general, I don’t see a big problem with the projections on the strikeout rate, but I do think the walk rate of 2023 is mostly real. But more so, this leads to the big actual outlier for Marsh, his contact profile.

BABIP is a volatile metric, but contrary to some earlier assertions, it is controllable by the player via their contact profile. There are certain types of balls in play that are more likely to be outs or hits than others. A pop up or any soft contact is likely to be an out. Line drives are much more likely to be hits. Fly balls have a high tendency to be outs if they aren’t home runs. Opposite field ground balls are more likely to be a hit than a pulled ground ball.

Brandon Marsh’s batted ball outcomes are geared towards hits. By Fangraph’s Soft% of balls in play, Marsh’s rate of 6.4% was the lowest. Only Ryan McMahon (1.7%) and Luis Arraez (0.6%) had lower infield fly ball rates than Marsh (2.2%). He runs a middle of the pack line drive and ground ball rate, while keeping a fairly even split of pull-center-oppo. He hits the ball hard by Statcast metrics too. His average exit velocity last season was 91.3 mph, which was 39th and equal with Jorge Soler and Paul Goldschmidt. His average launch angle was 12.4 degrees which was middle of the pack and speaks to his lack of pop ups and balanced approach. By Statcast’s hard hit measure he was 28th at 48.7%, just behind Austin Riley and just ahead of Kyle Schwarber.

This also isn’t an outlier for Marsh. In 1193 major league plate appearances he has a BABIP of .384. Because of the pandemic and injuries, he doesn’t have a very long minor league track record, but his two full years in 2018 and 2019 he had a BABIP of .356 and .384. A BABIP of .397 is almost certainly unsustainably high, but the batted ball profile is geared towards hits, and the swing profile indicates that much of this is purposeful.

Given history, it is easy to see why projections heavily disregard Marsh, but given he should also add Gold Glove caliber defense in left field, there is a very good chance that a healthy Marsh beats his projections, even if he does have some batted ball regression.