Pitchers and catchers report in just two weeks. There’s just one footnote of a football game left. Some major baseball projection systems are already published. Excitement is in the air for Baseball in 2023, and that has us thinking about fantasy.

As fantasy baseball players, we’re extremely lucky. Projections in baseball are by far the best among major American sports. The large sample sizes, long history and one-at-a-time nature of baseball lends itself well to projections, and there is a whole industry of smart, hardworking baseball nerds constantly improving on the science. That’s great, but if you’re planning to use projections to draft your team, one question remains: which projections should I use?

That’s where we come in. At FantasyPros, our goal is to provide useful fantasy advice, and an important part of that is evaluating which rankings and projections are most accurate. Today, we’re measuring the accuracy of pre-season projections for the 2022 baseball season. Before we get to the results, let’s have a quick rundown of our accuracy methodology.

Methodology

Fantasy is just one of many reasons to be interested in baseball projections, so we’ve developed an accuracy methodology that measures their value as fantasy advice. We do it in three steps:

1. Isolate the right data: Since we’re concerned with accuracy as it pertains to fantasy baseball, we look at the standard 5×5 roto categories: runs scored (R), home runs (HR), runs batted in (RBI), stolen bases (SB) and batting average (AVG) for hitters, and wins (W), saves (SV), strikeouts (K), earned run average (ERA), and walks + hits per inning pitched (WHIP) for pitchers. Additionally, we also evaluate at-bats (AB) for hitters and innings pitched (IP) for pitchers, so that we can separate playing time and quality-of-play projections. To do this, we evaluate other counting stats on a per-AB or per-IP basis. For example, last year Trevor Story hit 16 home runs in 357 at-bats. If you projected him to hit 25 homers on 550 at-bats, that’s a big miss for at-bats, but right on the money for home runs, since you nailed his 4.5% HR/AB rate.
2. Evaluate each projection against the field: For a given player and stat, we start by measuring the error – the difference between the projection and actual result – for each projection source. We then calculate the z-score for each projection. That is, we judge each projection by how many standard deviations better or worse than the average (among projection sources) it was. For example, the projection systems we looked at had an average error of 202.6 for Trevor Story’s at-bats, with a standard deviation of 33.2. If you projected 550 at-bats, that’s an error of 193 (because he actually had 357). That’s a bit better than the average error; 0.289 standard deviations better, to be exact. Thus, you get a z-score of -0.289 for Story’s at-bats (z-scores are negative when better than the average, and positive when worse). We do this conversion to z-scores so that we can compare and combine scores for stats that have different scales in their raw forms.
3. Aggregate by stat, and overall: now that we have z-scores for every projection made by every system, we average together the z-scores across the player pool to get a score in every stat for each projection system. We add these together across categories to get hitter and pitcher scores, and add those together for our final, overall accuracy score.

2022 Projections Accuracy Results

Remember that for the scores, lower is better, since they represent a projection system’s “error” in that category. Now for the results:

If you’re a regular reader of these articles, this is a familiar headline: Ariel Cohen‘s Average Total Cost (ATC) projections were the most accurate projections for fantasy baseball in 2022! Incredibly, this is ATC’s fourth win in a row (2021, 2020, 2019). This level of consistent accuracy is a powerful endorsement for Cohen’s consensus methodology. ATC is based on a combination of historical stats and several input systems, but a very smart combination. ATC goes further than a pure average like our own Zeile consensus projections – ATC uses different weighting of its inputs for each stat, to find the best possible combinations. ATC is firmly established as the best in the business.

Consensus systems like ATC might produce the most accurate projections possible, but their success is contingent on an industry of “original” projections, made from scratch using historic data and a heaping helping of ingenuity. We have a clear winner in that category too: Derek Carty‘s The Bat and The Bat X! Carty has been publishing The Bat since 2013, and in 2020 he introduced The Bat X for hitters. The Bat X adds another layer on top of traditional sabermetric projections by leveraging statcast data in a variety of ways. I encourage you to check out this article for more detail on what makes the Bat X special. It’s hard to understate how successful this upgrade has been – we’ve now seen The Bat X for three years, and it has been the most accurate original (that is, not a consensus of other projections) projection system all three years.

A congratulations is also in order for Mike Podhorzer for his Pod Projections, which earned the distinction as the most accurate original projection system for pitchers, in addition to typing for third place overall. This is another chapter in a tradition of success for Pod Projections, which have been top-5 every year we’ve measured them, and were hot on The BAT’s heels as the #2 original projection system last year.

Category Results

Here is the breakdown showing each system’s accuracy for each category that we studied, hitters first:

And pitchers:

—

Jacob Herlin is a Data Scientist for FantasyPros. You can find him on twitter @jacoblawherlin.