This is part of part of the Z-Book, an ongoing compilation of new and refreshed pieces. It's part of the 2020 Mastersball Platinum subscription, available for just $39.95, featuring the industry's earliest and most comprehensive set of player projections.
Welcome to the final installment of the Mastersball projection process, The Pitching Zystem. The general process hasn’t changed much since the last refresh of the method, however the projection for the individual skills has been improved.
Like with hitting, a three-year weighted average serves as the foundation, with MLEs (major league equivalencies) acting as a surrogate for Double-A and Triple-A performance. Unfortunately, the minor league data necessary for some of the skills regression of pitching isn’t as robust, so pertinent skills are regressed to league mean.
Hitting projections work off plate appearances, pitching revolves around batters faced (BF). Projecting walks, hit by pitch, sacrifices against and sacrifice files yield at bats. There’s science behind projecting walks; the others are based off history and have wide error bars, but the raw numbers are small, so the effect is minimal.
The four primary skills (homers, strikeouts, walks and hits) are all projected by using regressing actual to expected, with the default lever setting at 50% unless noted. This is then neutralized by applying park and aging factors.
HOME RUNS
The method is analogous to hitters, using average fly ball distance to discern expected homers. Actual homers allowed are regressed to expected, with the lever set at 75% actual, 25% expected to account for park factors. The projected homers are then neutralized using park and aging factors.
STRIKEOUTS
Apologies, I’m not sure who popularized this, but at this point it’s considered mainstream analysis. Swinging strike rate correlates well to strikeout rate. It’s not perfect, but it serves to unearth hurlers with the chance of increasing, or in some cases deceasing their punch outs. Each pitcher’s expected strikeouts are determined and regressed to actual, then neutralized.
WALKS
The excellent research staff at BaseballHQ.com found a similar relationship between the percent of balls thrown and walk percent. This is employed to calculate expected walks, regressed with actual and neutralized.
HITS
Hits are determined in an analogous manner to hitting, extrapolated from xBABIP (expected batting average on balls in play). Statcast data includes xBA (expected batting average) so xHits can be derived and plugged into the standard BABIP, using other xStats. Statcast data is neutralized by nature, so in order to regress it to actual BABIP, the components of BABIP must first be neutralized. The neutralized version is regressed to expected. From this, hits can be projected.
WHIP and ERA
WHIP can be determined directly from the projected stats after re-applying park factors and aging.
Even though there are some very elegant ERA estimators improving on the old school Gill and Reeve formula, I still use the antiquated method for a couple of reasons. Primarily, the inputs are hits, strikeouts, walks and homers, all projectable within reason, especially with the improved regression levers described above. The newer estimators utilize inputs with wider error bars, so any increased correlation is mitigated with added variance. Plus, the final ERA isn’t the expected ERA, but regressed to it based on the previous relationship between the pitcher’s actual ERA and Gill-Reeve xERA. At some point, I’d like to transition to a different xERA, but I’m not yet comfortable projecting some of the necessary components. Or maybe better said, the degree of accuracy of a new method isn’t sufficient to move off the simplicity of Gill-Reeve.
In lieu of a regression lever, final ERA is calculated using an index, comparing neutralized runs to expected runs (derived from Gill-Reeve). The projected ERA emanating from Gill Reeve using projected homers, strikeouts, walks and hits is then adjusted via the index. This bakes in allowances for regression and the ability for certain pitchers to over/underachieve their peripherals.
After each pitcher is projected, their LOB% (left on base percent) is compared to previous seasons. Some project LOB% and use that to yield ERA, I prefer to use LOB% as an eyeball litmus test. League average LOB% is around 72%. However, not all pitchers should be expected to regress towards that mean as some can outpitch the expectation. Ace starter and reliever can sport high LOB% levels. Starters can post marks in the 78% range while relievers can register levels in the mid-80s. I’ll look at the projected LOB%. If it doesn’t regress at least a little towards 72% (or higher for the elite), I’ll examine the projection and likely tweak the most logical regression lever, so the final projection is more sensical.
WINS/LOSSES
I’m seeing an increase in usage of iterations of this method, I wish I published it many moons ago when I first started employing it. Bill James developed a method to approximate team winning percentage using runs scored and runs allowed, termed Pythagorean Baseball Theorem since the formula resembles a2+b2=c2 for right triangles. It’s not perfect, but starter decisions correlate with innings so based on games started, the associated decisions can be secured. Using the pitcher’s ERA, adjusted to include unearned runs, his winning percentage can be found by including projected run support. Knowing winning percentage and decisions renders wins and losses. The same formula works or many relievers, however those used in higher leverage scenarios don’t enjoy the same chance for a win since they’re often entering the game with a lead, so the best they can do is log a save or hold. I’ve developed a tweak to the formula to account for this, applying it to all relievers projected for at least five combo saves and holds.
SAVES
Saves are not the crapshoot most believe. I’ve conducted studies showing the number of save opportunities correlates to team wins and ERA. Further, between 45%-55% of a team’s wins yield a save opportunity. Ergo, the starting point for save opportunities is 50% of team wins, estimated by the Pythagorean Theorem. Opportunities are then massaged in accordance with how the team’s projected ERA compares to league average. Relievers are assigned a percentage of save chances with converted saves projected using projected success rate.
HOLDS
I’ve looked, but to date I can’t find a method to scientifically project saves, so admittedly it’s a guesstimate based on team and player history.
MISCELLANEOUS
Ancillary stats such as wild pitches, pickoffs, hit batsmen, etc. are simply carried through the three-year weighted average without neutralization.
Thus completes the Mastersball Projection process. Please feel free to comment, critique or ask for clarification on the site message forum.{jcomments on}
Todd Zola is the Primary Owner and Lead Content Provider for Mastersball. He’s the defending Great Fantasy Baseball Invitational champion, besting 314 of the industry’s finest. Todd is a former Tout Wars and LABR champion as well as a multi-time NFBC league winner.