Throwing the Cutter and Beating the Market

For those of you hoping to build that game-changing model and get that free money betting the big leagues this year, rest assured there are few inefficiencies in the MLB betting market these days, especially near game time. However, there is one notable angle that has popped up since 2016. Betting pitchers who throw a cutter more than 20% often in games in the last two months compared to the other pitcher, in non-bullpen games (which I define as games where both starting pitchers have averaged more than 18 hitters faced per start in the current + last season, with one proxy start of 15 hitters faced added), has profited 36.1 units over 1187 games.

Even better, if one limits their sample to pitchers who have either started throwing the cutter 20% more in games in the last two months compared to the last season, relative to the other pitcher, or rookie pitchers who throw the cutter at least 20% more often, the bettor would have profited 27.2 units over 436 games against Pinnacle closing lines. The theory behind this angle is that pitchers who have picked up the cutter might be undervalued compared to the year before, as well as rookie pitchers who are undervalued compared to the typical rookie who does not throw a cutter. The theory is especially attractive because the cutter is "sneaky good" in a way that say, a pitcher who massively improves his strikeout rate, a statistic rightfully highly valued by the market, is not. In fact pitchers who add the cutter may actually strike out fewer batters than before and walk more batters but be better pitchers due to the unique properties of the pitch.

The counter theory is that these pitchers just were randomly good in these games, or that if you had picked one pitch out of the several pitch types that pitchers could have started throwing year over year, you were bound to find one that outperformed in this way.

Do Cutter Throwers Beat their Projections as Well as the Market?

Either way, we can look at why this happened and whether it is likely to be real by looking at how pitchers who have started throwing more cutters performed compared to their Steamer projections. If these pitchers really did beat their projections, that would make it somewhat more likely that this is a real trend, as opposed to the alternative, where these pitchers didn't really pitch better, but their teams won every one-run game and lost every blowout, and the outperformance of the angle was all luck. Since we don't know how many cutters rookie pitchers threw last year, we will limit our sample only to those pitchers with at least 150 BF in the current season and the year before.

Due to having the best out of sample performance in predicting future runs allowed compared to the other Steamer metrics I tried, we will use Steamer SIERA projected as a base for each pitcher's projection. However as we are more interested in outcome than components, especially since we are studying a pitching trait that is focused on preventing hard contact at the expense of strikeouts, meaning a component ERA like SIERA might be systemically wrong, we will instead use wOBA allowed as our measurement and use the following best fit equation to project "Steamer wOBA allowed" in each season based on the Steamer SIERA projection over the years 2016-2019:

Season wOBA allowed = .1459 + log(Steamer SIERA projected)  * .1365

The log is weird, but it was quite a bit better so we will go with it here.

We can now compare three groups of pitchers. The first group is made up of pitchers who threw 20% more cutters as a percentage of pitches than they did the year before, the second are those within plus or minus 20%, and the final group are those pitchers who threw the cutter 20% less.

wOBA (Steamer) wOBA (Actual) Difference Number of Pitchers
>20% More Cutters than LY 0.326 0.312 -0.014 43
-20 to 20 0.326 0.326 0 1116
>20% Less Cutters than LY 0.329 0.325 -0.004 15

Only 43 pitchers, who qualified by having 150 BF two years in a row, added the cutter to their arsenal in the 2016-2019 and across this group, the average pitcher outperformed their (implied) Steamer projection by 0.014 wOBA. 15 pitchers who dumped the cutter also improved slightly, while everyone else was in line.

Moving on, since the cutter is a contact-suppression pitch, it is possible that the Steamer SIERA projections fail to account for these qualities. Does Steamer SIERA under-rate cutter pitchers as a whole?

All Pitchers with >20% or More Cutters Thrown 0.323 0.328 -0.005 221
All Others 0.328 0.328 0 953

Given this set of 221 pitchers includes the 43 who started throwing the cutter and beat by 0.014, there would be seem to be little systemic error in the Steamer projections, at most 0.003 wOBA. So the beat here would appear to be due to these pitchers adding the cutter and getting better as a result.

Of the pitchers who began throwing the cutter, we can investigate if their improvement was due to the cutter, or if they simply started pitching better for other reasons. It could be that these were worse pitchers that had more upside, or perhaps were pitchers that were motivated to improve due to a contract year or some other reason, and the addition of the cutter is simply an expression of this motivation. We can see how the pitch performed using this framework, that analyzes the change in expected wOBA given what would have been predicted before and after the pitch. This allows us to study each pitch type on a per-pitch basis. We begin with the group of all other pitchers - those who did not change their cutter usage by 20% or more:

wOBA per Pitch, Control Group
Current Season Previous Season Difference
Fastball Performance 0.0021 0 0.0021
Cutter Performance 0.0011 -0.0016 0.0027
Cutter + Fastball Performance 0.0028 0 0.0028
Other Pitch Performance -0.0033 -0.0057 0.0024

Across all pitch types, the control group got worse by 0.0021 to 0.0028 wOBA per pitch. Given there an average of about four pitches in a plate appearance, this implies the control group pitched worse by about 0.010 wOBA per PA. While aging played a slight role, keep in mind these are season over season comparisons and we are using the same Statcast wOBA scale each year. Comparing 2016 to 2015, 2017 to 2016, 2018 to 2017, and 2019 to 2018, any set of pitchers would get slightly worse on average due to the ball becoming more juiced in three out of four of those years.

wOBA per Pitch, Pitchers who Started Throwing 20% More Cutters
Current Season Previous Season Difference
Fastball Performance 0.0025 0.0034 -0.0009
Cutter Performance -0.0086 0.0028 -0.0114
Cutter + Fastball Performance -0.0026 0.0022 -0.0048
Other Pitch Performance -0.0017 -0.0032 0.0015

Moving on, the pitchers who added the cutter improved more or less solely due to their cutter. Their cutter performed at -0.0086 wOBA per pitch, compared to their performance across all fastballs of 0.0022 wOBA the year before. These new cutter-throwers often threw the pitch the year before, and may have upped their usage due to an increased mastery of the pitch, as their cutter also performed much better than their cutter the previous season. Their performance on non-fastball pitches did not change significantly.

Overall, there may be a solid early-season angle around betting pitchers who picked up the cutter in spring training and look like they know what they are doing with it, as well as betting rookies with mediocre stats that utilize the cutter heavily. These pitchers are likely to be undervalued by the market.