This dataset contains historical playing data from Major League Baseball (tm), disaggregated at various levels. These are based on the files produced by Retrosheet.
There are two main groupings of files:
- daybyday: Game-by-game records for players and teams, for all seasons available from Retrosheet.
- splits: Various batting and pitching splits generated from event-level data, that is, which cannot be computed from the game-by-game records. This covers 1969 through present, for which Retrosheet has full play-by-play coverage.
The daybyday files include records from Retrosheet event files, deduced event files, and boxscore files. If Retrosheet offers a game in more than one of these types of file, then there will be multiple records for the players and teams in that game. There is a column 'game.source' which indicates the source of the record:
- evt: event file
- ded: deduced event file
- box: boxscore file
Retrosheet proofs games closely for offensive and pitching statistics. Defensive statistics, specifically putouts and assists, are less closely proofed, as these are known to be much less reliably kept, and because scorecards often have missing or inaccurate data for these. Generally boxscore files reflect official totals, while event files will reflect the limitations of the known sources of play-by-play for the game in question.
Deduced event files are play-by-play files which are reconstructed from partial information and context, rather than complete or almost complete game accounts. Some statistical categories may therefore be unreliable in these accounts.
We include the statistical totals derived from each of these sources because they have relative strengths; for example, plate appearances are not included in boxscore files, but in most cases can be computed accurately from event files. Users will need to choose which source(s) to use for their own work.
A player may also have more than one line for a game from a source. In the case of courtesy runners (pre-1950) or the occasional illegal substitution, a player may appear in more than one slot in the batting order. In this case, a player will have one line for each slot in which he appeared in a game. The statistics reported in that line are for his activity when he was in that lineup slot.
The player day-by-days contain two date fields: game.date and appear.date. The first of these is the date the game started. The second date is the date on which the player appeared in the game. The second field can differ from the first in the case of suspended games which are resumed on a subsequent date. For example, CHA201607230 was suspended before the top of the 9th inning due to the field becoming unplayable due to rain. It was resumed the following day, 2016-07-24. In the bottom of the 9th, Justin Wilson (wilsj004) came in to pitch for Detroit. His day-by-day entry for this game has 2016-07-23 for game.date and 2016-07-24 for appear.date. The appearance date is useful for correctly inferring player stints. There are a few dozen instances of a player entering a resumed game which started before the player joined the club. The official convention is that his statistics are credited on the day the game started. However, for some purposes one wants to compute the range of dates on which he actually performed with the club; appear.date can be used to get a correct first appearance date for each player on each club.
The underlying data is obtained free of charge from and is copyrighted by Retrosheet. Interested parties may contact Retrosheet at www.retrosheet.org.
This dataset is made available under the Open Database License: http://opendatacommons.org/licenses/odbl/1.0/. Any rights in individual contents of the database are licensed under the Database Contents License: http://opendatacommons.org/licenses/dbcl/1.0/.
The maintainer of this dataset is Dr Theodore Turocy, Chadwick Baseball Bureau; ted.turocy@gmail.com / http://www.chadwick-bureau.com.