Projections-Fueled Top 25 Pitching Prospects Entering 2025

Continuing an annual tradition, this article ranks the top 25 starting pitching prospects by peak projected MLB ERA heading into 2025 (skip to the bottom for the ranking!)

Like the other projection systems at FanGraphs, my projections capture the usual ingredients: past performance weighted by recenecy, regression to the mean that accounts for a player’s probability of making the major leagues, major league equivalencies to adjust for minor league difficulty, aging, park effects, and league scoring environment. The peak projections make use of aging curves to translate a player’s forecast to a late-20s peak forecast. Regardless of where the “true” peak age is, there is broad consensus that most growth happens in the teen years and early-20s, however.

You may have seen the redraft version of my projections that incorporates StuffPlus published by Eno Sarris over at The Athletic, or peak projections át Scout the Statline or Prospects Live. 2024 was another good season for my pitching projections in terms of forecast accuracy, particularly for the rookie class.

Notwithstanding, there are always improvements to be made in the offseason, in addition to more routine updates. I have now finally added in average fastball velocity to the peak projections. I had been holding off because velocity data had been harder to find for minor leaguers, but it isn’t so hard to find anymore. Velocity data is sourced from The Board at FanGraphs, Baseball Savant, and, in only a few instances, MLB.com. I also added in a velocity aging curve, and updated the regression amounts, major league equivalencies, aging curves, park factors, and league scoring environments. I now account for league scoring environment by subtracting league average (and then adding back the league average I am rescaling the projections to) instead of dividing by league average–both of these methods are similar, but I found the additive approach results in slightly better forecasts. Finally, I made substantial changes to the recency weights, generally weighing more recent performance more heavily compared to in the past. I had previously used Marcel’s 3/2/1 weights for 2024/2023/2022. For K%, I’m now using 3/1.2/.8, a weighting that minimized forecast error on historical data covering 2002-2024. The K% forecast weighs recency more aggressively than the other components, which are closer to Marcel.

To find an aging curve for average fastball velocity, I used the delta method, with data from a couple of different sources. First, FanGraphs has velocity data from Sports Info Solutions going back to 2002. Second, Baseball Savant offers velocity data for various minor leagues dating back to 2021. As there are very few early-20s pitchers in the majors, where most aging growth happens, the minor league velocity data from Baseball Savant is an important new data source for increasing the sample size of young pitchers with velocity data. I also cautiously considered velocity data from The Board given the data set is less systematically constructed.

To briefly summarize the delta method here, I subtracted a player’s average fastball velocity in their age 22 season from their average fastball velocity in their age 23 season to find the average change for all 22-year-olds (with at least 10 pitches in each season). I did this for every age, then chained the ages together to find a general aging curve. Fastball velocity is one of the most reliable baseball statistics, so a 10 pitch minimum was enough to generate a reasonably smooth aging curve (for other noisier statistics more subject to selection bias, a more complex version of the delta method might be necessary).

The figure below shows a general aging curve for average fastball velocity using 2021 to 2024 minor and major league data from Baseball Savant. If you chain the changes together, an 18-year-old gains around two miles per hour in average fastball velocity at his peak, and the peak occurs around 25. This curve captures data from 11 18-year-olds, and 27 19-year-olds, while there were only 3 19-year-olds in the MLB from 2002 to 2021–the minor league data offers a big increase in sample size at the younger ages. As we get more minor league data, we should have an even clearer understanding of how velocity typically changes in the teenage years (note, the graph captures data from 49 20-year-olds, 99 21-year-olds, and at least 100 players at each of the other ages).

That is enough stalling and exposition for now. The top 25 pitching prospects by peak projected MLB ERA are shown in table below. The projections assume a neutral park in the 2024 MLB environment. Each of the projections is adjusted to assume 20 total batters faced (TBF) per game, or about 5 IP/G–this makes it easier to compare pitchers with bigger differences in TBF/G. Relative to in the past, I am happy with the improvements as the list is looking closer to the scouting-based lists, particularly at the very top (Painter, Matthews, Rocker, and Rosario!)–an important source of triangulation for me.