Playoff odds calculator

Concept

A website that keeps track of official and live (accounting for in-progress games) standings for a variety of sports/leagues, with various playoff outcomes calculated based on each. Data is tracked and stored at the game level for easy real-time tracking and fast calculations of live standings, as well as (eventually) more granular predictive/data science-y coolness.

User-facing calculations

• Playoff spot odds
• Winning/Losing magic numbers
• Tiebreakers
• Strength of schedule
• Most important games
• What if?
• Clinching options
• Elimination options

Objects

Season

• Standings rules (tiebreakers)
• Some leagues use coin flips
• Games (schedule/records)
• Teams (conferences/divisions)
• Playoff rules

Team

• Name
• Statistics represented in standings, derived from Games
• Currently playing (to look up for live standings/odds)
• Data source ID

Game

• State
  • Scheduled
    • Start time
  • In Progress
    • Game clock
  • Complete
  • Postponed/rescheduled/suspended
  • Intermission/halftime?
• Score
• Home team
• Away team
• Sport/league-specific, standings-relevant stats
  • MLS
    • Home/away goals for/against
    • Disciplinary points
  • NHL
    • Goals for/against
• Data source ID

Mechanics

Preseason

• Pull in schedules once they're live, as Games

During season

Tracking games

• Around midnight each day, use Game Start Time to pull list of IDs for day's games
• Shortly before each Game's Start Time, spawn a worker to periodically ping for current game state until state becomes Complete/Final
  • Pull values we care about (keeping track of in Game) and compare to known state
    • If changes, update Game
  • Rate limit based on number of live Games/workers
• Periodically update Live Standings based on known Game states

Calculating playoff odds

• Method 1: Naive tallying of outcomes
  • Each game's possible outcome (eg A def B, B def A, tie, A def B (OT), B def A (OT), A def B (SO), B def A (SO)) is simulated
  • Final standings of each path calculated
  • Probabilities calculated as number of final states with Team X in Position >= Y
• Method 2: Weighting of outcome type
  • Same as above, but eg shootouts are weighted to historical probabilities with equal odds of victory
  • Could also take into account strength of schedule, historic matchups, etc
• Method 3: Full weighting of potential outcomes
  • Same as above, with added odds of victory based on team performance, loosely defined
  • Requires a good deal of planning/analysis to decide on weighting algo
  • To consider: Messi effects, etc

Gameplan

1. Pick a league or two to start with (done; NHL, MLS)
2. Create Game/Team/Season objects, relationships to get from games to standings (done)
3. Create methods to pull in real historical data (done)
4. Create methods to pull in and parse live game data into official/live calculations (in progress)
   1. standardize game statuses (scheduled, live, final, postponed/suspended, etc) (NHL done, MLS todo)
   2. NHL: how to handle tie games in live standings? (don't count game or 1 point each until update? latter doesn't help tiebreakers - would need to positively choose OTL/SOL to make sense. also, handle ties in regulation differently from OT/SO having already started?)
5. Create methods to calculate standings, including tiebreakers where possible (done)
   1. add s/x/y/e, etc to relevant teams
6. Create methods to persist game data (done)
7. Create methods to simulate unplayed games (in progress)
8. Create methods for additional probabilistic calculations listed above
9. Create systems for polling of live games, intelligent handling of off-days to minimize requests, etc
10. Create wrappers for league data sources
11. Create test suite
12. Create front-end

Down-the-road cool stuff

1. Let user assign outcome odds or pick winners/losers, see how it changes outcomes (would probably require some JS magic, might not be worth pursuing)
2. Feed outcomes into wagering bot?