Applying Hitter Volatility Research to Fantasy by Brad Johnson February 10, 2014 Over the years, Bill Petti has done some fascinating research on hitter volatility. He recently updated that work at the new Hardball Times. I strongly recommend taking a look. You can follow Petti on Twitter @BillPetti. Hopefully he’ll now forgive me for shamelessly stealing his work for fantasy purposes. Before we dive into Petti’s work and how it applies to fantasy baseball, let’s define volatility. In finance, volatility refers to variation in price over a period of time. A stock that sees frequent fluctuation of price is considered to be more volatile than a stock with very little price fluctuation. In fantasy baseball, we can replace price with category production or points. Player A is more volatile than Player B if his day-to-day performance is more varied. Petti’s research is in day-to-day volatility, which differs from streakiness. As fantasy owners, we know that Jay Bruce is one streaky mofo guy. He’ll put up 20 excellent games followed by 30 atrocious games and he’s been doing it since the day he entered the majors. By Petti’s measure, Bruce is only four percent more volatile than league average over the last three years. That’s (probably) because he’s not measuring these long term fluctuations. Generally speaking, power hitters are more volatile than contact hitters, with some notable exceptions. Petti’s stat, indexed volatility, is about as predictive year-to-year as batting average or batting average on balls in play, so one year of data has to be heavily regressed towards league average. For more detail, I once again recommend that you read his research linked above. So how do we use this research to our advantage in fantasy baseball? In roto and points formats, volatility isn’t much of an issue. Hypothetically, if you know a hitter is volatile and he produces a few good games, you can try to sell high. But day-to-day volatility won’t affect your draft price or preseason projection. Generally, you don’t care how the player gets to his final numbers as long as they are good. It’s a different story in head-to-head formats. A player’s volatility could affect a team’s performance. Returning to the example of Bruce, he’s capable of sealing a victory in his best weeks, but he can also ruin otherwise good performance when he’s playing poorly. Depending on the timing of Bruce’s streaks relative to the performance of your other players, streakiness could be a very good or very bad thing. Basically, your entire roster takes on a higher range of possible results due to one player’s volatility. While Petti’s volatility stat doesn’t account for this streakiness factor, it makes sense that volatile players are probably more prone to streaks. It’s long been theorized that volatility is a bad thing in H2H leagues. I am unaware of any specific research proving this assumption. As I noted above, volatility probably increases the range of expected outcomes, but not the mean expected outcome. Depending on the construction of your roster, that could be good or bad. That’s my disclaimer, now let’s get into Petti’s results. Here is a link to Petti’s spreadsheet with his VOL- statistic for the past three seasons. VOL- is simply volatility with league average set at 100, just like FIP-. So a 95 VOL- is five percent less volatile than league average. We can use Petti’s research as is to get a general sense of who is and isn’t volatile. However, it’s not very useful to us as fantasy owners to know that Marco Scutaro, Skip Schumaker, and Alberto Callaspo are not volatile. It would be better if we interacted performance with volatility. That way, we would have an idea of the most stable players who are also fantasy assets. After all, a down day for Miguel Cabrera is still better than many of Scutaro’s best days, so there must be a point where raw performance overwhelms volatility. I have taken Petti’s tables and added a column called Sharpe Ratio, which divides weighted Runs Created plus by VOL-. Credit for the idea to use a Sharpe Ratio goes to commenter Brandon. The Sharpe Ratio lets us find the best combination of high performance and low risk/volatility. You can find the full data in the updated spreadsheet here. Nothing has been done to adjust the Sharpe Ratio to a friendly scale. Higher numbers are better than low. Below is a table of the top 30 hitters over the past three seasons by Sharpe Ratio. These players have featured the best combination of performance and volatility. Three year data is ostensibly more predictive than a single season of data, although I have done no testing to confirm this with VOL-. Hitter VOL- STD VOL- Sharpe Ratio Miguel Cabrera 94% 0.10 189 Joey Votto 91% 0.02 178 Matt Holliday 92% 0.03 161 David Ortiz 100% 0.07 158 Shin-Soo Choo 85% 0.03 157 Prince Fielder 94% 0.04 156 Jose Bautista 99% 0.03 155 Andrew McCutchen 97% 0.06 151 Chase Headley 86% 0.03 149 David Wright 93% 0.03 148 Aramis Ramirez 94% 0.01 144 Giancarlo Stanton 100% 0.02 144 Robinson Cano 99% 0.03 143 Joe Mauer 93% 0.03 142 Jacoby Ellsbury 88% 0.03 141 Dustin Pedroia 86% 0.04 140 Mike Napoli 101% 0.08 139 Carlos Beltran 99% 0.06 139 Edwin Encarnacion 100% 0.07 137 Jose Reyes 89% 0.05 137 Alex Gordon 89% 0.01 136 Evan Longoria 101% 0.04 136 Josh Willingham 91% 0.06 136 Carlos Quentin 101% 0.07 136 Carlos Santana 95% 0.06 134 Yadier Molina 100% 0.02 133 Freddie Freeman 97% 0.03 133 Chase Utley 89% 0.05 133 Adrian Beltre 103% 0.00 133 Jayson Werth 96% 0.06 132 Unsurprisingly, Cabrera tops the list as the best combination of performance and volatility over the past three seasons. In volatility, he’s just six percent below league average, but the sheer weight of his performance outdoes his closest competition, which is Votto. As we can see, players like Ortiz, Stanton, and Encarnacion rank as some of the best values despite league average volatility. Beltre is three percent above league average. The take away is that overall production probably can overwhelm the value of consistency. Freeman’s inclusion on the back end of the list caught my attention. Excluding 24 plate appearances in 2010, Freeman is the only player who made the list in his first three big league seasons. The Braves recently signed Freeman to an eight-year, $135 million extension. Given the financial constraints facing the club, this extension might mean that Jason Heyward, Craig Kimbrel, and Justin Upton will reach free agency. Research shows that steady production is good for real-world teams. Whether consciously or unconsciously, Freeman’s combination of non-volatility and production may have been a contributing factor in the decision to extend him. He’s currently being drafted as the 24th pick and sixth first baseman off the board this season, which seems about right for strong four category production. Owners looking for steady production may want to target him if they miss out on the top tier of first baseman. For H2H league owners who want to build a lineup that features consistent production week-to-week, this list is probably a great place to start. Every position is represented, and generally speaking the players range from the second (Cabrera) to 318th pick (Quentin). Expand the list a bit past 30 and you can probably build an entire roster of players who combine production with stability. What about players who didn’t qualify for the above list? I have also included the 2013 season on a separate page of the spreadsheet, since only 154 batters qualified for the three-year table. I’ve limited the list to only players with at least 300 plate appearances and have presented the top 30 below. Bear in mind, this is much like a list of hitter BABIP. Just because Hanley Ramirez topped the list last season doesn’t mean he’ll do the same next season. VOL is Petti’s raw measure of volatility, which is used to derive VOL-. Hitter VOL- VOL Sharpe Ratio Hanley Ramirez 91% 0.432 210 Mike Trout 86% 0.406 205 Miguel Cabrera 99% 0.468 194 Shin-Soo Choo 85% 0.405 177 Andrew McCutchen 89% 0.423 174 Joey Votto 90% 0.426 174 Matt Carpenter 86% 0.409 170 Matt Holliday 88% 0.418 168 Paul Goldschmidt 97% 0.458 161 David Ortiz 94% 0.447 161 Yasiel Puig 99% 0.471 161 Freddie Freeman 94% 0.444 160 David Wright 97% 0.461 159 Carlos Quentin 91% 0.432 157 Michael Cuddyer 89% 0.423 157 Marlon Byrd 87% 0.411 157 Jayson Werth 102% 0.484 157 Joe Mauer 92% 0.436 157 Chris Davis 111% 0.524 151 Josh Donaldson 100% 0.472 149 Carlos Santana 91% 0.431 148 Allen Craig 92% 0.436 147 Chase Utley 86% 0.407 147 Starling Marte 82% 0.391 147 Jhonny Peralta 85% 0.402 145 Edwin Encarnacion 101% 0.480 143 Carlos Beltran 92% 0.437 143 Brandon Belt 97% 0.462 143 Aramis Ramirez 93% 0.443 141 Dustin Pedroia 81% 0.386 141 We see some new names appear like Trout, Carpenter, Goldschmidt, and Donaldson to name a few. These are players that we can probably expect to make the three year list once they qualify. By restricting ourselves to one season, we’re cutting into the reliability of the VOL- statistic. We’re also seeing just one season of wRC+ performance, which is less informative. The one year sheet is probably best used to form a list of new players who might eventually find a ranking on the three-year list, like Trout. I did not expect to find Byrd on the 2013 list with a VOL- 13 percent below league average (87% VOL-). If I had to bet on any one player on this list being a fluke, it would be Byrd. As a point in his favor, he was even less volatile in 2012. Of course, that was part of a miserable 153 plate appearance campaign that included a PED suspension. Last season included career highs in home runs, isolated slugging, and swinging strike rate. It would seem natural for a high swinging strike rate to coincide with high volatility. His 2011 season featured roughly league average volatility, which is more in line with what we probably should expect from him. In short, I wouldn’t add him to my short list of low volatility players based on his 2013 season. Take It One Step Further Enterprising owners in points leagues (especially points H2H formats) can further customize these lists. Substitute projected points for wRC+ and then divide by VOL-. The resulting list should give you a ROUGH idea of which players have the best combination of performance and stability in your points format. Just remember, the Sharpe Ratio you calculate will look like a points total, but it’s not. It can only be used to form a rough list. Statistically, there’s all kinds of things wrong with this approach, but the resulting output should be close to correct. If you’re very enterprising, you could also try to project VOL and then calculate your own projected VOL-. Petti described the formula in his article, but it’s basically the standard deviation of daily wOBA divided by yearly wOBA raised to the .52 power. You could try taking a player’s past standard deviation regressed some amount to league average and then divide by his wOBA projection. Then take projected points divided by projected VOL-. That sounds terrible to me, I would never attempt it, but it should be possible.