So, according to forecasts supplied by Yahoo, your fantasy team is projected to win 125-100 against its archrival this week. Sounds pretty promising, right? But just how confident should you be in that prediction?
Table 1 provides favorites’ winning percentage as a function of spread for both actual NFL games (based on 1994-2012 data from The Wizard of Odds) and fantasy matchups (using a representative sample of 2013 Yahoo! Fantasy Football data).
Note that while actual spread data is measured in points, fantasy spread is represented in percentage terms to account for league-specific differences in scoring settings. That is, a projected 125-100 winner would be a 25 percent favorite. So would a projected 1,250-1,000 winner (if your league prefers to go big with 50-point touchdowns!).
As anyone who's played fantasy football can attest, it's a bit of an unpredictable ride. Unlike in an actual game, those garbage-time yards and points (you know, the ones racked up during the waning moments of a 38-10 game) can often be decisive in a fantasy setting.
Also, fantasy football lines aren't being created by professional Vegas bookmakers or tweaked/adjusted based on millions of dollars worth of crowd-sourced feedback in the form of bets. Finally, individual performance (even when aggregated) is inherently more volatile than team performance – especially when considering the always-fickle individual touchdown.
As such, it shouldn't be surprising that actual NFL favorites win more frequently (66.2 percent) than fantasy favorites (61.2 percent). Of course, once Automated Insights starts doing more and more predictive work for fantasy football, we humbly expect those fantasy favorites to become safer and safer picks.
Table 2 reveals another interesting insight: there are projected to be many more near coin-flips in fantasy football along with far fewer forecasted blowouts.
Spreads in which the favorite wins at least 80 percent of the time occur 10 percent of the time in actual games vs. only 5.4 percent of the time in fantasy contests. Conversely, spreads in which the favorite wins less than 60 percent of the time happen in 60.3 percent of fantasy games, but just 36.5 percent of the time in real competition.
All that being said, both actual and fantasy favorites share a similarly (and predictably) shaped curve when mapping winning percentage as a function of spread.
A 2.5-to-3-point favorite in an actual contest is comparable to a 7-to-10 percent favorite in fantasy (with each winning about 60 percent of the time). That is, a 24-21 projected win in an NFL game should be treated with a similar degree of confidence as a 110-100 projection in your fantasy league. Likewise, a 6-point NFL favorite is roughly similar to a 15-to-20 percent projected fantasy winner (each wins about two-thirds of the time), and both 10-point NFL favorites and 30-to-40 percent fantasy favorites win about 80 percent of their games.
So, as you're trying to determine the level of smack you should talk to your best friend or significant other, consult the data. Going back to our theoretical 125-100 forecast, feel about 70 percent confident in bringing home a "W" for that week – as safe as you might feel if your favorite NFL team was predicted to win by a score of 27-20.
Adrian Atkinson is a data scientist at Automated Insights (Ai), the company that powers Yahoo Fantasy Football's draft recaps and matchup reports. Ai's software thinks and writes like a human, turning raw data into plain English.