DFS MLB: Ignore Recent Performance With Hitters
“Probably the most important and hardest part of understanding sports stats – accepting how much luck there really is.”
– Steamer’s twitter account
At the moment I am writing this article, Bryce Harper very nearly leads the league in wOBA at a whopping .491, an extraordinary number. Harper is most probably the best hitter in baseball, and may be the most talented prospect in MLB history, so the idea of Harper sustaining these numbers may not seem far fetched. However, every MLB projection system thinks Harper’s rest of season pace will be far below not only his production level so far this season, but even his production level last season. Harper is a young, extremely talented player, hitting super stardom, so why is everyone so bearish on his production? The simple answer is luck, but there are a lot of facets to concept that will be good to understand for Daily Fantasy MLB on FanDuel and DraftKings.
A baseball players performance, although based firmly on skill, will have a large deal of randomness from at bat to at bat. For example, when a player hits a HR, there was a considerable amount of luck that went into it happening. With a HR that barely clears the wall, even a small gust of wind would have brought it back in the park for an out, or blown a warning track out over the wall. Similarly, a groundout has a lot of luck involved, sometimes a few feet left or right gets through the infield for a single or double, other times it’s hit right towards a fielder for an out or a double play. Every event in baseball has some luck aspect to it, and some more than others. And even over very long periods of time with most events, luck can still be a dominant factor.
It may be tempting to think Daniel Murphy, who has started the season with a torrid .491 wOBA, is a must play in daily fantasy baseball right now. There is a strong consensus from projections posted on Fangraphs that Murphy will have a wOBA of around .33-.34 for the rest of the season. This makes perfect sense. Murphy is 31 and has tons of history, and his wOBA for the past several seasons has hovered around that .33-.34 number. To assume that a .491 wOBA is a reasonable projection for Murphy, you would have to assume Murphy, who is 31 years old, has drastically improved his skill. For a young player or a player with less history, a big skill increase could be occurring, although not likely on this massive scale. But for someone like Murphy, with so much history, the obvious conclusion is he has not gotten drastically better, and a .340 wOBA should still be a good estimate.
Instinctually, it can be hard to decide fade players who are doing well and play hitters who have had terrible seasons. I even find myself being hesitant when picking an ice cold hitter or fading a hot hitter whose ownership is going to be very high. But the math backs that strategy up. In “The Book”, the best baseball statistics book you can buy, author Tom Tango does some simple analysis on whether players who have had large or small wOBA’s over short stretches of time causes the hitter to outperform their historical wOBA over the next short stretch of time. What did the data show? Extremely little effect. I’m not saying recent hitter performance does not impact daily results (I do believe there are some aspects of recent performance that can help predict daily performance, which I will touch on in a future article), but it does suggest you can’t just look at fantasy points or wOBA in recent games and expect that trend to continue.
It is dangerous to assume every hitter who has performed way above his projection or career statistics has done so because of a skill increase. The reason is a statistical concept referred to as the multiple comparisons problem, also known as the Texas Sharpshooter Fallacy. If you have read my articles in the past, you have probably heard about me talk about it before, because I find it to be one of the fascinating concepts in statistics.
I can’t explain the multiple comparisons problem as well as I can give an example. Suppose you are flipping 500 different quarters ten times each in a “most heads flipped” concept. The expected amount of heads or tails for each quarter is 5 for each outcome. In reality, the results of each 10 quarter flips will vary wildly. It is likely, because we are flipping so many coins, that a few coins will have 9 or even 10 heads flipped. Now does this mean those coins that flipped heads a ton are more heads-prone coins? Of course not. We would expect at least some fair coins in a set of 500 to have these results, those coins just happened to be the ones that got lucky.
This problem is not so different than looking at the short term results of 500 MLB hitters. While grounded in skill, unlike a coin, short term results are still governed largely by luck. We expect out of every hitter in baseball in the short term, some of them are going to have extremely good results, others are going to be extremely bad. And that’s precisely what we see. Even Manny Machado, who is an amazing prospect and currently leads the league in wOBA, is not going to be keep up .a 514 wOBA. Similarly, Erick Aybar is not going to be as bad as his league worst .154 wOBA.
Picking hot hitters and avoiding cold hitters may seem instinctually like the right thing to do. However, because luck is so prevalent in baseball, we should expect hitters to regress to their career performance or performance of an average MLB hitter. Most players who perform well or terribly at the beginning of the season have not done so because of any meaningful change in skill, and therefore we shouldn’t expect them to be any better or worse than their projection as made by Steamer, ZIPS, or any other solid projection service. Do not fall into the amateur trap of overvaluing recent results for hitters.View all posts by Daniel Steinberg