fantasy-baseball

Fantasy baseball analysis

These scripts help value and rank players in ESPN and Yahoo fantasy baseball head to head categories league. League settings are configurable, including the categories.

See the introduction for details on the valuation model.

The category valuations are applied against rest-of-season projections from Fangraphs to calculate a cumulative winning probability added for the players.

Before each season

Fangraphs

Fangraphs has changed their form data in the past. Verify that the existing form data works for the current season.

• Fangraphs form data should be saved as config/projections.viewstate and config/leaderboard.viewstate.
• Note: if updating this form data is too time-consuming, just save response data into the projections directory.

Yahoo

1. Update config.json in Yahoo league data directory.

ESPN

1. Update config.json in ESPN league data directory.
2. Update configuration file by making a copy of config.py.sample, renaming to config.py, and filling out the configuration.
   • Update:
     • CURRENT_YEAR
     • season_final_day
     • BATTER_BUDGET_RATIO
       • Traditionally, this is set to around 0.66 (i.e. 33% to pitchers)
       • In 2018, the actual ratio was 0.68

Preparing for a draft

1. scrape_fangraphs.py --draft
   • Scrape the pre-season projections from Fangraphs. The projection system used is configurable.
2. Update player information in each league.
   • ESPN: Call save_eligibilities method after instantiating ESPN league.
3. map_players.py
   • TODO: update this script

Yahoo

ESPN

1. Update keepers list under data/keepers.csv.
2. python batter_valuation.py --draft
3. python pitcher_valuation.py --draft

Each week / scoring period

To gather fresh data:

1. scrape_playing_time.py
   • Scrape the number of PA / IP for each player in the past 14 days.
2. scrape_fangraphs.py
   • Scrape the rest-of-season projections from Fangraphs. The projection system used is configurable.
3. map_players.py
   • Each data source uses different names and ids for players. This creates a mapping table that translates between different ids from different systems.

In general, we want to update league specific information and valuations next.

1. Update rosters.
   • We want to have current rosters for each team, and positional eligiblities to make positional adjustments for players. This is different for each league, so we store this in a league-specific data directory.
2. Scrape the most recent scores.
   • After each scoring period, we'll have additional information on the relative value of each category. (For example, if SB have been low this season, the value of an individual SB will likely increase.)
3. Update the category valuations.
4. Update player valuations.

Yahoo

1. For Yahoo leagues, to update positional eligibility and team rosters:

   yahoo = Yahoo("example_league_data_directory")
   yahoo.refresh_rosters()
   yahoo.refresh_eligibilities()

2. Call refresh_scores with the week number you want scores for.

   yahoo.refresh_scores(week_number)

3. Update category valuations.
   • TODO: Implement this.
4. Update player valuations.

   yahoo.load_players()
   yahoo.output_valuations()

ESPN

1. For ESPN leagues, to update positional eligibility and team rosters:

   espn = league.Espn("example_league_data_directory")
   espn.save_eligibilities()

2. Call save_scores which will save all scores up to the current week.

   espn.save_scores()

3. prob_added.py
   • Update the logistic regression.
4. Update player valuations.
   • batter_valuation.py
   • pitcher_valuation.py