(https://github.com/Kmulla19/Blog/tree/main/Blog3)
In my last post, we began looking at the relationship between the number of Top 100 Fantasy PPR on a team, and that team's total wins.
There were some interesting findings, with a basic Linear Regression modeling explaining 40% of the variance in that team's wins.
So what's the other 60%? What else could potentially further explain the number of wins an NFL team has in a season?
Obviously, there are a number of factors that determine the outcome of each individual game and hence, the total number of wins a team has. What about consistency among those fantasy stars week in and week out?
It's tough to be consistent in the NFL. If you're a Running Back, you can't help it if a star member of your Offensive Line is out one week. You also don't decide who you play each week. The NFL decides who you play based upon where you finished the previous season. Basically, the teams that did well, will play more of the other teams that did well in the following season.
But it's still worth looking into if there is a connection between how consistent a team's Top Fantasy players are in terms of point variance, and how many wins that team had. When I say point variance I mean does the player put up roughly the same numbers each week or do they have a lot of boom and bust weeks that average out to a good points per game average.
First, we need to determine each players points per game. This is easy. We simply take the total number of PPR points they had on the season, and divide it by the number of games they played. Remember not every player, plays all 17 games.
Next, we need to look at how each player did on a weekly basis. Now Pro Football Reference did not have the weekly PPR points for each player on the website but they did have the formula for how PPR is calculated and each player's stats for each week.
Okay that's a little annoying but nothing some Python can't solve. First we need to pull in the weekly stats for each player and then have the stats go through our PPR formula to determine the number of points they had each week. This can be done by using a web-scrape similar to how we pulled in the top players and team standings. We created a for loop that takes the player's name in our Top 100 PPR DataFrame, then pull's in there weekly stats, and puts them through our custom PPR formula.
We tested a few players season total PPR points and found that our formula is usually off by about 1 point in either direction which we are fine with summing up with rounding error. I don't expect this to skew the data much since we are talking about hundreds of total points so it changes the data by less than 1%.
Now that we have each player's weekly stats we can look at the standard deviation in points as a percentage, relative to their weekly average. We need to look at it as a percentage because a 5 point std for a player who averages 20 points per week is different than a 5 point std for a player who averages 12 points.
Stay tuned for the next blog post where we will look at these variance numbers per player, team, and how that affects wins.
Top comments (0)