In znmeb/dfstools: Tidy Data Analytics from Sports Data APIs

ReLoad the saved data from the NB model run, which took 2.9 hours

load("~/DFS/kaggle/slow.RData")
# `WFinalFour` is the teams / ID table
# `nb_model` is the `mvglmmRank` model

Attach libraries

library(dplyr)
library(tibble)
library(mvglmmRank)
library(knitr)

Utility function for parsing a row of mvglmmRank::game.pred output

.get_number <- function(text)
  as.numeric(sub(pattern = "^.*: ", replacement = "", text))

Make a tibble of the six possible games

predictions <- tibble()
for (irow in 1:3) {
  for (jrow in (irow + 1):4) {
    team1_id <- as.character(WFinalFour$TeamID[irow])
    team2_id <- as.character(WFinalFour$TeamID[jrow])
    game_prediction <- capture.output(game.pred(
      nb_model, team1_id, team2_id, neutral.site = TRUE))
    pred_score_1 <- .get_number(game_prediction[2])
    pred_score_2 <- .get_number(game_prediction[3])
    spread_1 <- pred_score_2 - pred_score_1
    spread_2 <- pred_score_1 - pred_score_2
    over_under <- pred_score_1 + pred_score_2
    win_prob_1 <- .get_number(game_prediction[9])
    win_prob_2 <- 1 - win_prob_1

    predictions <- predictions %>% 
      bind_rows(tibble(
        team_1 = WFinalFour$TeamName[irow],
        team_2 = WFinalFour$TeamName[jrow],
        pred_score_1,
        pred_score_2,
        spread_1,
        spread_2,
        over_under,
        win_prob_1,
        win_prob_2
      ))
  }
}

kable(predictions)

Session info

sessionInfo()

znmeb/dfstools documentation built on March 3, 2020, 5:50 p.m.