Using Park Factors In Daily Fantasy Contests
Using Park Factors in Daily Fantasy Contests
In addition to a full time job as a software product manager, Alex Zelvin works part time for Fanduel.com (Zoobird on FanDuel) and co-owns Dailybaseballdata.com
Park factors are a more complicated subject than one would think. The general issue that you’re faced with is determining what is going to have better predictive value – an adjustment based on a very specific situation that has few historical instances, or an adjustment based on a more general situation with a larger sample size.
This trade off is going to affect your decisions on a variety of topics regarding park factors. Is it better to use one year of data or multiple years? Is it better to use separate park factors for right-handed batters and left-handed batters, or to group all hitters together? Is there value to using player-specific historical data?
In general, it’s best to use the most specific situational data that you can, provided that there is a sufficient sample size. Looking at player specific data (for example, how has David Wright hit at Wrigley Field) is generally pointless, particularly for visiting players. They simply don’t have enough at bats in each park to give meaningful data.
On the other hand, looking at specific park factors based on handedness (particularly for hitters) can yield useful results, since many parks are not symmetrical. Sample sizes are reasonably large, especially if you’re able to use adjustments based on multiple years of data. Even better than handedness would be dividing hitters by where they tend to hit the ball. Unfortunately, while the raw data needed to do this is available, I don’t think anyone has actually done the work for us, so it’s going to be an impractical approach for anyone but the truly obsessed.
Another challenge is deciding whether to use component park factors – specific adjustments for each type of event, such as singles, doubles, triples, home runs, and even strikeouts and walks. The alternative is using less granular adjustments…maybe limiting your self to park factors for runs scored and for hits and home runs.
There are other practical considerations too. Since most of us are not going to calculate our own park factors, we need to choose from what’s available. Some sources of park factors (such as ESPN) are freely available. Others (such as the Bill James Handbook) require a small purchase.
The most popular free source of park factor data is ESPN . They list the data year by year, and their formula basically takes each team’s average in a category per game at home and divides it by their average in the same category per game on the road. One drawback of their data is that it would be nice if they allowed you to aggregate the data for multiple years. A second drawback is that they include runs, hits, home runs, doubles triples, and walks…but not singles or strikeouts. Another flaw in their methodology is that by counting the home team players, the park factors are disproportionately affected by those players. If the team has been crafted to include hitters who can take advantage of the specific quirks of a stadium, that’s going to create park factors that make the park appear to be more favorable to hitters than it really is.
If you’re will to spend about $20 each year, you can purchase the Bill James Handbook. The park factors included in the book are better than ESPN, because they show cumulative adjustments based on the past three years, and they include strikeouts.
A variety of other sources of data are available on the internet. I’m hesitant to link to them, because many are incomplete or unreliable. For example, one popular one over the past few years is FirstInning.com. However, when I visited the page, the park factors appeared to be incorrect, with minimal adjustments listed for even the most extreme ballparks.
One other thing to keep in mind is that you need to know what you’re adjusting. Remember that performance projections typically already assume that players have half their games at home (and the other half in an average park). So if your projections show that Adrian Gonzalez will hit .275 with 36 home runs, remember that they’re already assuming that he’ll achieve that level of performance playing in San Diego’s extremely unfavorable home park. If your park factors suggest that his home park suppresses home runs by 30%, don’t adjust his average performance down by 30% when you’re forecasting his performance at home. By doing so, you’d be ‘double counting’ the impact of the home park. First, adjust his season long projection to be ‘park neutral’, then use the full 30% downward adjustment for home games. Or, alternatively, just take his average projected performance and use half of the park adjustment (15% in this case).