A few weeks back I posted an article looking at the biggest values of the 2016 fantasy season through week 13. Now that the season is over, let’s rerun the analysis and see what happens.

The methodology is the same as before. Simply taking the ratio of the average draft position (ADP) to the player’s Value Over Replacement (VOR) ranking. The methodology isn’t perfect of course, it doesn’t consider consistency or playoff performance among other things, but with parsimony in mind we’ll stick to this simple formula. Throughout the off-season I’ll explore other measures.

Thanks to some feedback from reddit user jitit who suggested having a cut off ADP, so guys drafted in the 500’s don’t pollute the rankings (which of course I should have done in the first place), this time around I limited the ranking to players with an ADP of 200 or less. Players with an ADP over 200 were put into the free agency rankings.

All values are based on FantasyPros PPR scoring and include weeks 1 through 16 because if your championship goes into week 17, you need to reevaluate your life. The R code to reproduce this is at the bottom of the article.

Biggest Values

Jordan Howard ended the season with the highest overall draft value and for good reason. Drafted as RB62, he ended the season as RB10, outscoring guys like Todd Gurley (drafted as RB1, finished RB18) and Lamar Miller (RB5, RB16). Matt Ryan (QB19, QB2)  and DeMarco Murray (RB18, RB5) were close behind.

Before the playoffs started, Melvin Gordon (RB23, RB7), was by far the best draft value of the year. Through week 13, Melvin Gordon was RB3, behind only David Johnson (RB2, RB1) and Zeke Elliott (RB3, RB2) consensus first round picks. But going down with an injury in week 14 dropped him to a still impressive RB7.

Biggest Busts

The bust list (sort the draft value column ascending) is pretty much unchanged from the previous (week 13) bust list. Adrian Peterson (RB4, RB120) and Keenan Allen (WR8, WR 155) take the top. But a special award needs to be given to DeAndre Hopkins (WR4, WR34) for not missing a single game and still being a bigger bust than Jamaal Charles (RB8, RB113) who had a grand total of 12 carries the entire 2016 season. Hats off the you DeAndre, here’s hoping that Tom Savage will bring you back to superstar status.

Best Free Agent Picks Ups

Which players were free agents at the start of the year will vary widely by league, so to be consistent and stay in line with the cut off from above, I defined free agents here as players with an ADP of over 200. Since value doesn’t really make sense for free agents… they’re free, I used VOR to rank them. Below is a table of the top 50 free agents based on full season VOR.

Because the drop off of TE scoring is so steep (that will be another post), Kyle Rudolph (TE30, TE3) and Cameron Brate (TE36, TE6) finished the season with the highest VOR of any free agent players, 48.2 and 38.9 points, respectively. Probably who most people would consider the best free agent pick up of the year, Davante Adams (WR82, WR12), finished directly behind Rudolph and Brate with an end of season VOR of 35.7 points.

Best Second Half Free Agents

Considering the entire season however might not tell the whole story as some players don’t come on until later in the season. Looking at the entire season washes out the value that players like Tyreek Hill (WR153, WR26) and Adam Thielen (WR142, WR23) added to your team in the second half of the season. In the second half of the season, Tyreek Hill finished as WR7, outscoring the likes of Brandin Cooks (WR14, WR7) and T.Y. Hilton (WR15, WR5) in weeks 8-16. In the same time frame, Adam Thielen finished as WR9. In the table below all values are calculated using only the games from weeks 8-16, thus the Pos Rank column is the rank of that player in weeks 8-16 only, not the entire season as in the other tables and the parenthetical notations throughout the article.

Bonus: They Are Who We Thought They Were!

A total of 8 players ended up finishing the season at exactly same positional ranking as they were drafted.

Antonio Brown PIT (WR1, WR1)
Odell Beckham Jr. NYG (WR3, WR3)
Mark Ingram NO (RB9, RB9)
Jameis Winston TB (QB15, QB15)
Zach Miller CHI (TE18, TE18)
Phil Dawson SF (K26, K26)
DeAndre Washington OAK (RB52, RB52)
Blaine Gabbert SF (QB33, QB33)

Say what you want about San Francisco, but at least they’re predictably bad.

The R code to reproduce all of this is below

library(dplyr)
library(rvest)

### functions -----
getPoints <- function(year = 2016, weeks, PPR = TRUE, name.with.team = TRUE) {
  
  format_url <- ifelse(PPR, "ppr.php", "")
  
  week_min <- min(weeks)
  week_max <- max(weeks)
  week_url <- paste0(week_min, ":", week_max)
  
  points_url <- paste0("https://www.fantasypros.com/nfl/reports/leaders/",
                       format_url, "?year=", year, "&week=", week_url)
  
  fp <- html(points_url) %>% html_table() %>% .[[1]] 
  
  fp$starting_week <- week_min
  fp$ending_week <- week_max
  
  if (name.with.team) {
    
    fp$Player <- paste(fp$Player, fp$Team)
  
    }

  names(fp) <- tolower(names(fp))
  names(fp)[which(names(fp) == "rank")] <- "point_rank"
  
  fp
}


getADP <- function(year = 2016, PPR = TRUE) {
  
  format_url <- ifelse(PPR, "ppr-overall.php", "overall.php")
  
  # need this to get team abbrevations
  points_url <- "https://www.fantasypros.com/nfl/reports/leaders/ppr.php?year=2016&week=1:16"
  fp <- html(points_url) %>% html_table() %>% .[[1]] 

  adp_url <- paste0("https://www.fantasypros.com/nfl/adp/",
                       format_url, "?year=", year)
  
  adp <- html(adp_url) %>% html_table() %>% .[[1]] 
  
  
  ## ADP Table -----
  teams <- unique(fp$Team)
  teams <- c(teams, "DST", "FA")
  
  # current free agents do not have a team or FA indicator
  # find them and add FA
  fa_index <- which(!grepl(paste0(teams, collapse = "|"), adp$`Player Team (Bye)`))
  fa_names <- adp$`Player Team (Bye)`[fa_index]
  fa_new_names <- paste(fa_names, "FA")
  adp[fa_index, "Player Team (Bye)"] <- fa_new_names
  
  adp_name <- sapply(strsplit(adp$`Player Team (Bye)`, " "), FUN = function(x) {
    
    team_vec <-  x %in% teams
    team_index <- which(team_vec)
    
    if (sum(team_index > 0)) {
      
      name_vec <- x[1:(team_index)]
      out <- paste0(name_vec, collapse = " ")
      
    } else {
      
      out <- paste0(x, collapse = " ")
    }
    
    out
  }
  )
  
  adp$Player <- adp_name
  adp$apd_rank <- adp$Rank
  
  adp <- adp %>% mutate(player = adp_name,
                        adp_rank = Rank,
                        adp_avg = AVG,
                        adp_pos = POS) %>%
    select(player, adp_rank, adp_avg, adp_pos)
  
  adp
}
### functions -----



fp <- getPoints(weeks = 1:16, PPR = TRUE)
adp <- getADP(PPR = TRUE)

### join the tables and do some clean up -----
df <- right_join(adp, fp, by = "player")

### players that were not in the adp fill in the rank and avg -----
df[is.na(df$adp_rank), "adp_rank"] <- max(df$adp_rank, na.rm = TRUE) + 1
df[is.na(df$adp_avg), "adp_avg"] <- max(df$adp_avg, na.rm = TRUE) + 1


### calculate VOR and draft value -----
vorp_df <- df %>% group_by(position) %>% arrange(desc(points)) %>% 
  mutate(pos_rank = 1:n(),
         replace_rank = ifelse(position == "QB", 13, 30),
         replace_rank = ifelse(position == "RB", 25, replace_rank),
         replace_rank = ifelse(position == "WR", 25, replace_rank),
         replace_rank = ifelse(position == "TE", 13, replace_rank),
         replace_rank = ifelse(position == "K", 13, replace_rank),
         replace_rank = ifelse(position == "DST", 13, replace_rank),
         replace_value = nth(points, median(replace_rank)),
         vor = points - replace_value) %>% ungroup() %>% 
  arrange(desc(vor)) %>% 
  mutate(vor_rank = 1:n(),
         draft_value = adp_avg / vor_rank) %>% ungroup() %>% 
  mutate(pos_rank = paste0(position, pos_rank))

# filter out DST and K b/c their ADPs are so low
# that they throw off the rankings
vorp_df <- vorp_df %>% filter(!position %in% c("DST", "K"))


### draft value table -----
draft_df <- vorp_df %>% select(player, position, adp_avg, 
                               adp_pos, pos_rank, points, 
                               avg, vor, draft_value) %>% 
  arrange(desc(draft_value)) %>% filter(adp_avg <= 200)

draft_df$draft_value <- round(draft_df$draft_value, 3)
draft_df$vor <- round(draft_df$vor, 1)

names(draft_df) <- c("Player", "Pos", "ADP", "Drafted As", "Pos Rank", 
                     "Points", "PPG", "VOR", "Draft Value")

readr::write_csv(draft_df, "data/draft_value.csv")



### top free agent pick ups avg adp > 200 -----
fa_df <- vorp_df %>% filter(adp_avg > 200)

fa_df <- fa_df %>% select(player, position, adp_avg, 
                          adp_pos, pos_rank, points, 
                          avg, vor) %>% 
  arrange(desc(vor)) %>% top_n(50, vor)

fa_df$vor <- round(fa_df$vor, 1)

names(fa_df) <- c("Player", "Pos", "ADP", "Drafted As", "Pos Rank", 
                  "Points", "PPG", "VOR")

readr::write_csv(fa_df, "data/free_agent_value.csv")



### second half of the season best FA -----
fp_2nd <- getPoints(weeks = 8:16, PPR = TRUE)

df_2nd <- right_join(adp, fp_2nd, by = "player")

### players that were not in the adp fill in the rank and avg -----
df_2nd [is.na(df_2nd $adp_rank), "adp_rank"] <- max(df_2nd $adp_rank, na.rm = TRUE) + 1
df_2nd [is.na(df_2nd $adp_avg), "adp_avg"] <- max(df_2nd $adp_avg, na.rm = TRUE) + 1


### calculate VOR and draft value -----
fa_df_2nd <- df_2nd  %>% group_by(position) %>% arrange(desc(points)) %>% 
  mutate(pos_rank = 1:n(),
         replace_rank = ifelse(position == "QB", 13, 30),
         replace_rank = ifelse(position == "RB", 25, replace_rank),
         replace_rank = ifelse(position == "WR", 25, replace_rank),
         replace_rank = ifelse(position == "TE", 13, replace_rank),
         replace_rank = ifelse(position == "K", 13, replace_rank),
         replace_rank = ifelse(position == "DST", 13, replace_rank),
         replace_value = nth(points, median(replace_rank)),
         vor = points - replace_value) %>% ungroup() %>% 
  arrange(desc(vor)) %>% 
  mutate(vor_rank = 1:n()) %>% ungroup() %>% 
  mutate(pos_rank = paste0(position, pos_rank)) %>%
  filter(adp_avg > 200, !position %in% c("DST", "K"))

fa_df_2nd <- fa_df_2nd %>% select(player, position, adp_avg, 
                          adp_pos, pos_rank, points, 
                          avg, vor) %>% 
  arrange(desc(vor)) %>% top_n(50, vor)

fa_df_2nd$vor <- round(fa_df_2nd$vor, 1)

names(fa_df_2nd) <- c("Player", "Pos", "ADP", "Drafted As", "Pos Rank", 
                  "Points", "PPG", "VOR")

readr::write_csv(fa_df_2nd, "data/free_agent_2nd_half_value.csv")