Before getting into Jordan’s intro, we wanted to share that the PLIVE+ 2022 leaderboard and projections are free view and utilize by everyone at the link below. Weekly updates for 2023 will be included on our Patreon at the 60 Tier ($10/mo) level.
PLIVE+ replaces RoboScout in our suite of Dynasty baseball tools at our Patreon. While RoboScout is an excellent tool, it only allowed for comparisions within an MiLB level and was only based on a single year of data. PLIVE+ covers that, but also includes multi-year performance rankings, across-level player comparisons and projections.
This article introduces a new encompassing metric to help evaluate prospects: PLIVE+. It is a kindred spirit—and highly correlated with—Dylan White’s excellent RoboScout and my peak projections that you may have seen on Scout the Statline or at Prospects Live.
In a nutshell, PLIVE+ is a statistic that summarizes peak projections for prospects. For bats, it is based mainly on peak projected wRC+, a comprehensive measure of future offensive talent. It also incorporates peak projected stolen bases, giving it a fantasy tilt, as well as ‘reliability,’ to account for sample size.
Reliability shows the proportion of a player’s projection that comes from their actual historical performance versus the proportion that is comprised of regression (e.g., 80% reliability means 80% of their projection is based on actual historical performance versus 20% based on regression).
More precisely, PLIVE+ is 60% peak projected wRC+, 15% peak projected stolen bases, and 25% reliability. PLIVE+ is scaled to wRC+, except with 100 approximately representing MLB replacement level (80 wRC+) rather than MLB average (100 wRC+). For example, a prospect with a 150 PLIVE+ projects to be 50 percent better than MLB replacement level at peak. About 400 prospects project to be MLB replacement-level or better talents.
These 400 prospects are also given a percentile rank (PLIVE%), with the #1 prospect ranking in the 99.9th percentile and the #360 prospect ranking around the 0 to 10th percentile. PLIVE+ for arms will be detailed more in a future post, but it is based on peak projections for K%, BB%, and fly ball %, plus reliability and innings per appearance.
PLIVE+ incorporates all of the key ingredients found in a typical projection:
Historical performance
Aging curves
Major League equivalencies to account for MiLB difficulty
MiLB scoring environments
Park effects
Regression
The aging curves convert each player’s performance to its peak-equivalent performance. The major league equivalencies convert player performances in different leagues to the same MLB baseline. Regression amounts differ for each component, with stickier measures like K% requiring less regression to the mean, and noisier measures like singles requiring more regression to the mean. Taken together, and accounting for historical performance, park effects, and league scoring environment, these projections allow for an apples-to-apples comparison of prospects in different leagues at different ages (note: DSL performance is ignored). You can find more detail about my process in generating the aging curves and major league equivalencies in my Prospects Live article on projecting the top pitching prospects (the approach to projecting hitters follows the same logic).
“PLIVE+ incorporates all of the key ingredients found in a typical projection: historical performance, aging curves, major league equivalencies to account for minor league difficulty, minor league scoring environments, park effects, and regression. ”
In comparison to my previous work, there are two big additions:
I made use of Baseball America’s indispensable park factors, adding in regression, an oft-forgotten but important step for using park factors
Instead of regressing all prospects to the MLB average, as I’ve done in years past for simplicity, I regressed prospects toward a mean that accounts for their probability of making the majors based on their age relative to their level.
This last step is worth elaborating on: based on historical data, a 21-year-old hitter in Triple-A has an almost 100% chance of making the majors, whereas a 21-year-old hitter in Class A has only a 23% chance of making the majors. In my new approach, the 21-year-old in Triple-A would thus be regressed toward the MLB average, whereas the 21-year-old in Class A would be regressed to be well below the MLB average (23%*MLB average + 77%* MILB average). When testing prospect projections on MLB outcomes, it’s sufficient to regress them all toward the MLB average because the guys that do not make the majors are not included in the test.
However, the goal here is to evaluate prospects with an uncertain chance of making the majors relative to one another and eventually to other major leaguers. Regressing them all toward the MLB average assumes they’re all going to make the majors one day—or at least that they’re MLB-caliber talents—overrating them as a class. Regressing prospects based on their probability of making the majors also gives guys in the upper minors a boost relative to guys in the lower minors. Note that for players with a decent sample of performance, the regression only comprises a small proportion of their projection so the impact of this adjustment is fairly minor.
In comparison to RoboScout as previously featured on the pages of Prospects Live, there are two key additions.
First, PLIVE+ accounts for historical performance rather than just single-year performance (although a single-year version is also included). As a rule, accounting for less than a minimum of three years of performance is going to leave “bills on the sidewalk” when projecting player outcomes, resulting in less than maximum predictive accuracy. PLIVE+ accounts for four years of historical performance for hitters and three years for pitchers, as this was found to maximize predictive power. PLIVE+ uses the well-known MARCEL weights, 5/4/3/2 for bats, 3/2/1 for arms, in each case weighting the most recent year the heaviest and weighting the most distance year the lightest. The recency weightings ensure that the projections adjust quickly to new information and talent changes.
The second divergence is that RoboScout focused on performances within a given level; PLIVE+ makes use of major league equivalencies so prospects at different levels and leagues can all be easily compared to one another.
A future article will take a look at historical data to see how PLIVE+ performed in the past. Offseason PLIVE+ scores for hitters are available now here. Pitchers are soon to follow but in the meantime, my previous article contains a leaderboard of peak projections for arms—PLIVE+ will be tightly correlated to those peak projections. We’ll begin updating the PLIVE+ leaderboards for 2023 as soon as possible. In the meantime, enjoy the lists!