Impact of Weather on Optimal Lineup Setting

Impact of Weather on Optimal Lineup Setting

Article Image

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

rotogrinders misc weather small

I recently interviewed a professional meteorologist for a player profile on Fanduel.com. In the course of the interview, he mentioned that a lot of the paranoia that many of us have regarding the impact of weather on daily fantasy contests is overblown, since barely more than 1% of the 2,700 major league baseball games that are played each year get rained out on average.

I’ll start by saying that I think that’s a misleading statistic. Most of our paranoia is directed towards the games with a substantial chance of rain in the forecast…let’s say 30% or higher. How many of the 2,700 games meet that threshold? I haven’t done the research (yet), but I would estimate that it averages about three for each day when all teams play. So that’s 20%, or 540. So out of that ‘universe’ of games we’re talking about 27 rainouts out of 540 potentially problematic days. Still only 5%, but that seems frequent enough to be an issue, especially when you consider that the percentage certainly gets higher as the forecasted chance of rain gets higher.

Article Image Before getting into the specifics of the impact weather has on daily fantasy lineup selection, it’s worth thinking about how most weather forecasts take place. When you visit a site like Weather.com or Yahoo.com that provides weather forecasts for all U.S. cities and towns, the forecast is 100% computer generated. Programmers (presumably directed by meteorologists) have created a statistical model to forecast the chance of rain, temperature, wind speed, wind direction, and more. The first thing to remember is that those models handle most common weather features moderately well, but there are some more unusual conditions that they may not ‘grasp’. The second thing to remember is that the models are not ‘refreshed’ on a continuous basis. The most popular models are run every 6 hours…and some of the data that is used as an input for their calculations may only be collected every 12 hours. Add to that the time it takes for the complex calculations to be run (about an hour as far as I can tell) and you’re looking at a forecast that may be based on data from as much as 13 hours ago. Another thing to keep in mind is that the verbal descriptions used in those forecasts are generated by computers too. For example, on Yahoo weather, during the summer a 30% chance of thunderstorms is always described as ‘isolated thunderstorms’, while a 40% chance of thunderstorms is always described as ‘scattered thunderstorms’. For the most part, the computer generated text doesn’t really add any information.
In addition to the fact that computer generated forecasts are almost always somewhat ‘stale’, you should be aware of how the site you’re using ‘consolidates’ information. When the main forecast for a city from Weather.com shows a 70% chance of rain, that simply refers to the chance of rain at some point during that day. To get any real value from the forecast, you’ll need to drill down to the hour by hour forecast. That can be pretty time consuming to do one game a time, so I’d recommend a daily visit to my site which shows the hour by hour forecasts for all games on a single web page.

Article Image You can sometimes get better forecasts from the weather men on television, but not always. To some extent, it depends on the specific weather man. Some are meteorologists, others aren’t. Some tend to simply report on what they see on the computer models (often adding a few hours more to the lag from data collection until you see the forecast), while others apply more independent analysis to what they’re seeing in the models, on radar, and other data. In general, the closer you get to the time they’re forecasting for, and the more unusual the weather situation, the more likely that the weather men are going to add some value. A 6am forecast for that night’s weather is probably taken almost verbatim from the computer models. A 6pm forecast for game time when there’s an approaching storm almost certainly includes some real analysis based on live or near-live data.

So how does all this impact your player selections? For me, it boils down to two things. I want to avoid all players whose games have a substantial chance of being postponed. And I want to also avoid any starting pitchers whose games may be interrupted by rain, shortening the pitcher’s outing. While ‘isolated thunderstorms’ wouldn’t scare me away from a hitter, it definitely would make me think twice about selecting a starting pitcher. Another factor that people take into account is wind strength and wind direction. To some extent, using this data effectively depends on a knowledge of each specific ballpark. For some of them, you can assume that the wind direction and strength will be fairly similar to what is being reported in the forecast for the area. Due to the layout of the ballparks, others are more quirky and often have wind patterns that don’t match those of the surrounding area. Without a substantial amount of further research, it’s hard to know exactly how much to take wind into account. But you can bet that when the wind is blowing out at Wrigley field, a lot of your opponents will stack their lineups with every Cubs player with a pulse.

About the Author