Analysis of Season Win O/Us

gamling_odds_implied_wins.R

library(rvest)
library(stringr)
library(dplyr)
library(ggplot2)
library(tidyr)


min.fun <- function(p, thresh.prob = NULL, thresh = NULL) {
  (thresh.prob - pbinom(thresh, 16, p))^2
}

fit.binomial <- function(thresh.prob, thresh) {
  o <- optimize(min.fun, interval = c(0, 1), thresh.prob = thresh.prob, thresh = thresh)
  o$minimum
}

page <- read_html('https://www.sportsbook.ag/sbk/sportsbook4/nfl-betting/nfl-season-wins.sbk')

teams <- page %>%
  html_nodes('span.team') %>%
  html_text()

## Convert gambling odds into probabilities
df <- page %>%
  html_nodes('div.market') %>%
  {data_frame(txtline = html_text(.), team = rep(teams, each = 2))} %>%
  mutate(bet = str_extract(txtline, '[OU]'),
         games = as.numeric(str_replace(str_extract(txtline, '\\d{1,2}(\\.\\d)?\\('), '\\(', '')),
         line = as.numeric(str_extract(txtline, '[-+]\\d{3}')),
         line = ifelse(is.na(line), 100, line)) %>%
  mutate(prob = ifelse(line < 0,
           abs(line) / (abs(line) + 100),
           100 / (line + 100)))


## The Under prob is equivalent to
## P(wins < K) == P(wins <= K-1)
## P(wins < 8.5) == P(wins <= 8)
## P(wins < 8) == P(wins <= 7)

## The Over prob is equivalent to
## P(wins > K) == 1 - P(wins <= K)
## P(wins > 8) == 1 - P(wins <= 8)
## P(wins > 8.5) == 1 - P({wins <= 8)

df2 <- df %>%
  group_by(team, bet, txtline, line, games) %>% do({
    if(.$bet == 'U') {
      gwp <- fit.binomial(.$prob, floor(.$games - 0.5))
    } else {
      gwp <- fit.binomial(1 - .$prob, floor(.$games))
    }
    data_frame(gwp = gwp)
  })


df2 %>%
  ungroup %>%
  select(team, bet, games, gwp) %>%
  spread(bet, gwp) %>%
  ## Average the O and U implied probs
  mutate(expected.wins.over = 16 * O,
         expected.wins.under = 16 * U,
         expected.wins = 8 * (O + U)) %>%
  select(team, expected.wins, ou.games = games) %>%
  ggplot(aes(x = reorder(team, expected.wins), y = expected.wins)) +
  geom_point() +
  geom_point(aes(y = ou.games), shape = '+') +
  geom_text(aes(label = format(expected.wins, digits = 2, nsmall = 1), y = 4.7), hjust = 'right') + 
  coord_flip() +
  theme_bw() +
  theme(legend.position = 'bottom') +
  scale_y_continuous(breaks = 4:11) +
  xlab('Team') +
  ylab('Implied Number of Wins Implied by Odds')