使用 Stattleship Sports API 和 Neo4j 的 NHL 球队排名模型

MATCH (n)
RETURN n
LIMIT 15

MATCH (t1:Team {nickname:'Bruins'})-[r1]->(g1)<-[r2]-(t2)
RETURN t1, r1, g1, r2, t2

// win loss percentage for each team
MATCH (n:Team)-[r]->(g:Game)
WITH n.nickname as team, sum(r.win) as wins, count(r) as games
RETURN team, wins, games, (wins*1.0/games*1.0) as wpc

MATCH (t:Team)-[r]->(x:Game)
WITH t.nickname as team, CASE TYPE(r) WHEN 'AWAY_TEAM' THEN 1 ELSE 0 END as away_game, CASE TYPE(r) WHEN 'HOME_TEAM' THEN 1 ELSE 0 END as home_game, r
RETURN team, sum(away_game) as away_games, sum(home_game) as home_games, count(r) as total_games
ORDER BY total_games DESC

MATCH (t1:Team)-[r1]->(g1:Game)<-[r2]-(opponent:Team)
WITH t1.nickname as team, sum(r1.goals) as score, sum(r2.goals) AS opp_score
RETURN team, score, opp_score, score-opp_score as score_diff
ORDER BY score_diff DESC

MATCH (t1:Team)-[r1]->(g1:Game)<-[r2]-(opponent:Team)-[r3]->(g2:Game)<-[r4]-(t2:Team)-[r5]->(g3:Game)<-[r6]-(t3:Team)
WHERE NOT t1=t2 AND NOT t1=t3
WITH t1, sum(r6.goals) AS oop_gf, sum(r5.goals) as oop_ga
// goal differentials for team, and the teams they played (OPP)
MATCH (t1:Team)-[r1]->(g1:Game)<-[r2]-(opponent:Team)-[r3]->(g2:Game)<-[r4]-(t2:Team)
WHERE NOT t1=t2
WITH t1, sum(r3.goals) AS opp_gf, sum(r4.goals) as opp_ga, oop_gf, oop_ga
// goal differential for the team's opponents
MATCH (t1)-[r1]->(g1:Game)<-[r2]-(opponent:Team)
WHERE NOT t1=opponent
WITH t1.nickname as team,
     sum(CASE WHEN type(r1) = "HOME_TEAM" THEN r1.goals ELSE 0 END) as home_gf,
     sum(CASE WHEN type(r1) = "HOME_TEAM" THEN r2.goals ELSE 0 END) as home_ga,
     sum(CASE WHEN type(r1) = "AWAY_TEAM" THEN r1.goals ELSE 0 END) as away_gf,
     sum(CASE WHEN type(r1) = "AWAY_TEAM" THEN r2.goals ELSE 0 END) as away_ga,
     sum(r1.goals) as gf,
     sum(r2.goals) as ga,
     opp_gf,
     opp_ga,
     oop_gf,
     oop_ga
WITH team,
     home_gf,
     home_ga,
     away_gf,
     away_ga,
     gf,
     ga,
     opp_gf,
     opp_ga,
     home_gf-home_ga as HGD,
     away_gf-away_ga as AGD,
     gf-ga as GD,
     opp_gf-opp_ga as OGD,
     oop_gf,
     oop_ga,
     oop_gf-oop_ga as OOGD
RETURN team, .25*(1.4*AGD + .6*HGD) + .5*(OGD) + .25*(OOGD) as rpi
ORDER BY rpi DESC

###############################################################################
## Use Stattleship and Neo4j to Rank NHL Teams for the 2015-16 Season
## @brocktibert
###############################################################################

## factors are the devil
options(stringsAsFactors = FALSE)

## packages -- if errors, install with ?install.packages
library(stattleshipR)
library(dplyr)
library(lubridate)
library(stringr)

## set the token from an environment variable
set_token("STATTLE_TOKEN_HERE")

## parse out entries from ss_get_result when walk=T and length > 1
parse_stattle <- function(stattle_list, entry) {
  x <- do.call("rbind", lapply(stattle_list, function(x) x[[entry]]))
  stopifnot(is.data.frame(x))
  return(x)
}

## The Atlantic Division id
atlantic_div <- "15ca4e46-2b49-4f37-84ea-befb62de28c5"

## get the NHL teams in the atlantic division and keep just the key fields
teams <- hockey_teams() %>%
  filter(division_id == atlantic_div) %>%
  select(id, nickname, slug)

## get the finished games, keep only those between atlantic division teams
all_games <- hockey_games(team_id="")
games <- filter(all_games,
                !is.na(ended_at) &
                  home_team_id %in% teams$id &
                  away_team_id %in% teams$id)
games <- select(games,
                id,
                started_at,
                scoreline,
                home_team_id,
                away_team_id,
                winning_team_id,
                attendance,
                duration,
                home_team_score,
                away_team_score,
                score_differential,
                home_team_outcome,
                away_team_outcome)

## extract dateparts from started date
games <- transform(games,
                   start_date = strptime(started_at, "%Y-%m-%dT%H:%M:%S"))
games <- transform(games,
                   year = year(start_date),
                   month = month(start_date),
                   day = day(start_date))
games$start_date <- NULL
games$started_at <- NULL

## save the csvs for import -- put on google drive for public access
write.table(teams, file="teams.csv", sep=",", row.names=F)
write.table(games, file="games.csv", sep=",", row.names=F)