This is a description of the new records created to support "Game Files".
The purpose of Game Files is to provide information on Negro League games for
which box-score level data are not available.

Game Files are marked by a file name extension of ".gf"

Game Files are modeled after Box-Score Event Files (*.eb?) with some additional
records introduced to convey information that may be available which cannot be
calculated directly from box-score level data which may be incomplete. A
description of box-score event files can be found at Retrosheet's website
(https://retrosheet.org/boxfile.txt).

In addition to the introduction of some new lines, which are described below,
unknown or uncertain information is identified somewhat differently in game
files. Data which are unknown are identified in game files by a ? (unknown
information is identified by the number -1 in box-score files). In addition, if
there are any data for which the number is known to be greater than zero but not
known with certainty (e.g., a batter had at least one hit in the game), the
number of known instances will be followed by an i (e.g., "1i" would indicate
at least one of the relevant statistic).

As with box-score and event files, each game record will begin with a line
identifying the game id.
id,{GameID}

The second line will identify the file version
version,5

This distinguishes game files (version 5) from earlier file types.

As with box-score event files, the record types within a game file may appear
in any order. In addition, all sequence information within record types is 
explicitly described in each record and is not implied by its order in the file.

The new record types:

info,gametype,

This identifies the type of game. Game types will take one of the following values:

  Ex -> Exhibition game
  AS -> All-Star game
  PS -> Playoff game
  IL -> Inter-League Exhibition game
  IR -> Inter-racial game
  NAL -> Official Negro American League game (1937-48)
  NN1 -> Official Negro National League game (original NNL, 1920-31)
  NN2 -> Official Negro National League game (second NNL, 1933-48)
  ANL -> Official American Negro League game (1929)
  ECL -> Official Eastern Colored League game (1923-28)
  NSL -> Official Negro Southern League game (1932)

A series of flags have been added to game files to identify whether or not
box-score information was available for the game. There are four flags as
follows. In all cases, flag values are either "Y" or "N".

flag,box,box1,box2
flag,box_o,b_ab,b_r,b_rbi
flag,box_d,d_po,d_a,d_e
flag,box_p,p_ip,p_h,p_r,p_er,p_w,p_k

Where:

  box1  - indicates whether a box-score exists (Y) or not (N)
  box2  - indicates whether the box score contains additional information beyond
          a simple box identifying players and their basic statistics
  b_ab  - indicates whether the box score indicates batter at bats
  b_r   - indicates whether the box score identifies runs scored by batter
  b_rbi - indicates whether the box score identifies RBIs by batter
  d_po  - indicates whether the box score identifies putouts by fielder
  d_a   - indicates whether the box score identifies assists by fielder
  d_e   - indicates whether the box score identifies errors by fielder
  p_ip  - indicates whether the box score identifies IP by pitcher
  p_h   - indicates whether the box score identifies hits allowed by pitcher
  p_r   - indicates whether the box score identifies runs allowed by pitcher
  p_er  - indicates whether the box score identifies earned runs allowed by
          pitcher
  p_w   - indicates whether the box score identifies walks allowed by pitcher
  p_k   - indicates whether the box score identifies strikeouts by pitcher

stat,bline - a record describing a player's batting and base-running data
for the game.

For each of the basic player lines in box-score event files - bline, dline, and
pline - game files also include a parallel line for each team.

These stat lines are identified as tmbline, tmdline, and tmpline, respectively
and contain the following information.

stat,tmbline,TeamID,side,,,ab,r,h,2b,3b,hr,rbi,sh,sf,hbp,bb,ibb,k,sb,cs,gidp,int

The columns associated with "pos" and "seq" in player blines (related to the
player's lineup position) are left blank here as these make no sense at the
team level. They are left blank, rather than removed, however, so that the 
team batting line (tmbline) lines up with player batting lines (bline) in all
other regards.

stat,tmdline,TeamID,side,,,if*3,po,a,e,dp,tp,pb
stat,tmpline,TeamID,side,,ip*3,no-out,bfp,h,2b,3b,hr,r,er,bb,ibb,k,hbp,wp,balk,sh,sf

As with tmblines, the tmdline and tmpline include blanks for various items which
are not relevant at the team level (defensive position, the sequence of the
players).

The final new piece of information provided in game files are a new type of
event file which provides whatever information may be known about the runs
scored in the game.

The syntax for these lines is as follows:

event,rline,side,inning,bat_run,bat_rbi,pitch_r,pitch_rbi,{seq of events}

Where:

  bat_run -   is the player who scored the run
  bat_rbi -   is the player who recorded the RBI on the play
  pitch_r -   is the pitcher to whom the run was charged
  pitch_rbi - is the pitcher who was on the mound when the run scored
 {seq of events} then details the plays involved in producing the run,
              to the extent these are known

Example:

event,rline,1,1,soueh101,broww101,clarv101,clarv101,W,1/G.1-2,S1.2-H

Herb Souell (soueh101) was driven in by Willard Brown (broww101) in the bottom
of the first inning with Vibert Clarke (clarv101) pitching. The run-scoring
sequence began with a walk to Souell who advanced to second base on a ground
out to Clarke and was driven in by Brown on a single which was also fielded
by Clarke.