Blog - Reedy For the Big Time: Rookie Wide Receiver Jayden Reed Ought to be a Serious Consideration for Your Fantasy Football Team

Although Fantasy Football is highly popular among sports fans, I personally have never taken up much of an interest in it. That being said, as I have steadily improved my sports analytics skills over the past several months, it seemed like a relevant area of application. Thus, the prospect of implementing machine learning to predict fantasy points that will be accumulated by rookie wide receivers this upcoming season was rather enticing.

This year’s NFL draft included 31 wide receivers in total with 3 of them being first round picks. Due to this, it would be sensible to assume that those 3 should accumulate the most fantasy points this upcoming season. Is this really the case though? Well, in order to investigate, we must determine which factors we deem important to predict fantasy scores as they pertain to wide receivers. Personally, I took these to be pre-draft rating, total yards, and total yards per perception.

Pre-draft rating makes sense to use as it takes into account various factors that suggest a player’s potential. So while it is not a concrete stat per say, it hints at the kind of impact a player could make. As for total yards, this is relevant because receivers that accumulate large amounts of yardage are able to impact their teams in a positive way. Adding context to yardage, I thought it was worthwhile to implement total yards per reception as this essentially shows how impact a receiver tends to be each time he catches a pass. In other words, it is a measure of a efficiency in some sense.

In addition to integrating these aspects into my model, I used originally planned to use the xgboost algorithm as it very popular among data scientists for solving various problems and is also has the capacity to handle new data relatively well. However, due to difficulties I settle for the the lightgbm algorithm, and in doing so, I was able to create a predictive model and ultimately visualize data pertaining to rookie wide receivers Fantasy Points as shown below.

Code

library(tidyverse)
library(tidymodels)
library(cfbfastR)
library(janitor)
library(bonsai)
library(gt)

wrstats <- read_csv("https://mattwaite.github.io/sportsdatafiles/wrdraftedstatswithratings.csv") %>% mutate(FantPt = case_when(FantPt == 0 ~ NA, TRUE ~ FantPt))

wrselected <- wrstats %>%

  select(
    name,
    year,
    college_team,
    nfl_team,
    overall,
    pre_draft_grade,
    total_yards,
    total_ypr,
    FantPt
  ) %>% na.omit()

player_split <- initial_split(wrselected, prop = .8)

player_train <- training(player_split)
player_test <- testing(player_split)

player_recipe <- 
  recipe(FantPt ~ ., data = player_train) %>%
  update_role(name, year, college_team, nfl_team, new_role = "ID")

lightgbm_mod <- 
  boost_tree() %>%
  set_engine("lightgbm") %>%
  set_mode(mode = "regression")

lightgbm_workflow <- 
  workflow() %>% 
  add_model(lightgbm_mod) %>% 
  add_recipe(player_recipe)

lightgbm_fit <- 
  lightgbm_workflow %>% 
  fit(data = player_train)

lightgbmpredict <- 
  lightgbm_fit %>% 
  predict(new_data = player_train) %>%
  bind_cols(player_train) 

lightgbmpredict <- 
  lightgbm_fit %>% 
  predict(new_data = player_test) %>%
  bind_cols(player_test) 

currentwrstats <- read_csv("https://mattwaite.github.io/sportsdatafiles/currentwrstats.csv")

currentwrselected <- currentwrstats %>%
  select(
    name,
    year,
    college_team,
    nfl_team,
    overall,
    pre_draft_grade,
    total_yards,
    total_ypr
  ) %>% na.omit()

currentlightgbmpredict <- 
  lightgbm_fit %>% 
  predict(new_data = currentwrselected) %>%
  bind_cols(currentwrselected) 

currentlightgbmpredict %>% 
  top_n(5, wt=.pred) %>%
  select(name, nfl_team, overall, .pred) %>%
  arrange(desc(.pred)) %>%
  gt() %>%
  cols_label(
    name = "Player",
    nfl_team = "Team",
    overall = "Draft Pick",
    .pred = "Fantasy Points"
  ) %>%
  tab_header(
    title = "Predicted Fantasy Points for Rookie Wide Receivers",
    subtitle = "Jayden Reed outperforms his status as the 50th overall pick."
  ) %>%  
  tab_source_note(
    source_note = md("by: Jacob Spooner")
  ) %>% 
  tab_style(
    style = cell_text(color = "black", weight = "bold", align = "left"),
    locations = cells_title("title")
  ) %>% 
  tab_style(
    style = cell_text(color = "black", align = "left"),
    locations = cells_title("subtitle")
  )

Player	Team	Draft Pick	Fantasy Points
Predicted Fantasy Points for Rookie Wide Receivers
Jayden Reed outperforms his status as the 50th overall pick.
Jayden Reed	Green Bay	50	105.78890
Jordan Addison	Minnesota	23	105.26333
Jaxon Smith-Njigba	Seattle	20	101.90746
Jalin Hyatt	New York	73	94.15025
Quentin Johnston	Los Angeles	21	81.03259
by: Jacob Spooner

Observing our top 5, there is one name that particularly sticks out: former MSU star Jayden Reed who is predicted to more than 100 Fantasy Points. Looking at last season, this would put him near Chris Godwin (who had 119 Fantasy Points), which is far from bad company considering that Godwin was a reliable number two receiver for Tampa Bay. Couple this with the fact that Reed was the 6th wide receiver drafted, and it becomes clear as to why he would be a solid Fantasy pick.

All in all, based on this model, I think it is fair to conclude that Jayden Reed will be a potential steal for those who will be playing Fantasy Football this upcoming season. While it is yet to be determined whether or not I be among those taking part, in the case that I do, Jayden Reed will most definitely be a part of my team.