agt-the-walker / arimaa-utils

Small utilities to download archived games from http://arimaa.com/arimaa/, print normalized setups...

Purpose

fetch-new-games

This script downloads new archived games from Arimaa into $HOME/.arimaa-utils (this location can be changed in lib/common.sh). It also builds a games database in the same location, which will be used by the following scripts.

good-human-games

This script only prints good human games from the games database built by fetch-new-games. Each human player should be rated at least 1800 and the game, rated, should last 10+ moves.

normalized-setups

This script prints normalized setups for both players from the games database built by fetch-new-games. It can also act as a filter (see example below). By convention, the Gold elephant is always placed on the right (from Gold's point of view).

positions

This script prints FEN positions (as described in Appendix C of Beginning Arimaa) from the games database built by fetch-new-games. It can also act as a filter (see example below). It accepts the following flags:

• --normalize (or -n): print normalized positions. Vertical symmetry of the Arimaa board is taken into account, and so are missing piece types. For instance, a position where only rabbits, dogs and elephants remain will be normalized to a position where only rabbits, cats and dogs remain.
• --skip_duplicates (or -s): don't print repeated positions (with the same player to move) for a given game

Requirements

• Bash
• GNU Coreutils
• GNU Tar
• GNU Wget
• Lynx
• Ruby 2.0+

Usage

$ export PATH=~/src/git/arimaa-utils:$PATH  # adapt accordingly

$ fetch-new-games  # takes a while, especially the first time

# the following command prints the most common setups in all games
$ normalized-setups | sort | uniq -c | sort -rn | head
   7362 1w RRRDDRRR/RHCMECHR  1b RHCEMCHR/RRRDDRRR
   5798 1w RRRRRRRR/DHCMECHD  1b DHCEMCHD/RRRRRRRR
   5256 1w RRRCCRRR/RHDMEDHR  1b RHCEMCHR/RRRDDRRR
   4821 1w RRRDDRRR/RHCMECHR  1b RHDEMDHR/RRRCCRRR
   4403 1w RRRRRRRR/DHCMECHD  1b RHCEMCHR/RRRDDRRR
   3989 1w RRRCCRRR/RHDMEDHR  1b RHDEMDHR/RRRCCRRR
   3982 1w RRRRRRRR/DHCMECHD  1b RHDEMDHR/RRRCCRRR
   3741 1w RRRDDRRR/RHCMECHR  1b DHCEMCHD/RRRRRRRR
   3703 1w RRRCCRRR/RHDMEDHR  1b DHCEMCHD/RRRRRRRR
   1843 1w RRRRRRRR/DHCMECHD  1b DHCMECHD/RRRRRRRR

# number of good human games (see definition above) in the database
$ good-human-games | sed 1d | wc -l
3971

# the following command prints the most common setups in good human games
$ good-human-games | normalized-setups | sort | uniq -c | sort -rn | head
    277 1w RRRDDRRR/RHCMECHR  1b RHCEMCHR/RRRDDRRR
    190 1w RRRDDRRR/RHCMECHR  1b RHDEMDHR/RRRCCRRR
    166 1w RRRCCRRR/RHDMEDHR  1b RHCEMCHR/RRRDDRRR
    114 1w RRRCCRRR/RHDMEDHR  1b RHDEMDHR/RRRCCRRR
     62 1w RRRDDRRR/RHCMECHR  1b RHCDEDMH/RRRRRCRR
     42 1w RRRDDRRR/RHCMECHR  1b HMCEDCHR/RRRDRRRR
     41 1w RRRDDRRR/RHCMECHR  1b RHCEMDHR/RRRDCRRR
     37 1w RRRCDRRR/RHDMECHR  1b RHCEMCHR/RRRDDRRR
     36 1w RRRDDRRR/RHCMECHR  1b RHCDECMH/RRRRDRRR
     29 1w RRRRRRRR/DHCMECHD  1b RHCEMCHR/RRRDDRRR

# print the most common Gold setups in good human games
$ good-human-games | normalized-setups | cut -d' ' -f1-2 | \
  sort | uniq -c | sort -rn | head
   1272 1w RRRDDRRR/RHCMECHR
    634 1w RRRCCRRR/RHDMEDHR
    165 1w RCRDDRCR/RHRMERHR
    159 1w RRRCDRRR/RHDMECHR
    155 1w RRRRRRRR/DHCMECHD
    108 1w RRRRRCRR/RHCDEDMH
     63 1w RDRRDRRR/RHCMECHR
     60 1w RDRCCRDR/RHRMERHR
     53 1w RRRCCRRR/DHRMERHD
     50 1w RRRDCRRR/RHCMEDHR

# print the most common Silver setups in good human games if Gold chooses
# the most common setup
$ good-human-games | normalized-setups | \
  grep -Po '(?<=^1w RRRDDRRR/RHCMECHR  ).*' | \
  sort | uniq -c | sort -rn | head
    277 1b RHCEMCHR/RRRDDRRR
    190 1b RHDEMDHR/RRRCCRRR
     62 1b RHCDEDMH/RRRRRCRR
     42 1b HMCEDCHR/RRRDRRRR
     41 1b RHCEMDHR/RRRDCRRR
     36 1b RHCDECMH/RRRRDRRR
     29 1b RHREMRHR/RCRDDRCR
     27 1b RHCEDCMH/RRRDRRRR
     24 1b DHCEMCHD/RRRRRRRR
     18 1b RHCEMCHR/RRRDRRDR

# print the first ten positions of game 132386
$ good-human-games | grep -w ^132386 | positions | head
132386  1s 8/8/8/8/8/8/RHDMEDHR/RRRCCRRR
132386  2g rrrdrrcr/hhcedrmr/8/8/8/8/RHDMEDHR/RRRCCRRR
132386  2s rrrdrrcr/hhcedrmr/8/4E3/8/1H6/R1DM1DHR/RRRCCRRR
132386  3g rrr1rrcr/h1cddrmr/1h6/3eE3/8/1H6/R1DM1DHR/RRRCCRRR
132386  3s rrr1rrcr/h1cddrmr/1h6/3eE3/8/1H4H1/1RDMCDR1/RRRC1RRR
132386  4g rrr1rrcr/2cddrmr/hh6/4E3/1e6/1H4H1/1RDMCDR1/RRRC1RRR
132386  4s rrr1rrcr/2cddrmr/hh6/1E6/1e6/1H4H1/R1DMCDR1/RRRC1RRR
132386  5g rrr1rrcr/2cddrmr/hh6/1E6/He6/6H1/R1DMCDR1/RRRC1RRR
132386  5s rrr1rrcr/2cddrmr/hh6/1E6/1e6/1H4H1/R1DMCDR1/RRRC1RRR
132386  6g rrr1rrcr/2cddrmr/hh6/1E6/He6/6H1/R1DMCDR1/RRRC1RRR

# print the corresponding normalized positions (due to vertical symmetry)
$ good-human-games | grep -w ^132386 | positions -n | head
132386  1s 8/8/8/8/8/8/RHDEMDHR/RRRCCRRR
132386  2g rcrrdrrr/rmrdechh/8/8/8/8/RHDEMDHR/RRRCCRRR
132386  2s rrrdrrcr/hhcedrmr/8/4E3/8/1H6/R1DM1DHR/RRRCCRRR
132386  3g rrr1rrcr/h1cddrmr/1h6/3eE3/8/1H6/R1DM1DHR/RRRCCRRR
132386  3s rcrr1rrr/rmrddc1h/6h1/3Ee3/8/1H4H1/1RDCMDR1/RRR1CRRR
132386  4g rcrr1rrr/rmrddc2/6hh/3E4/6e1/1H4H1/1RDCMDR1/RRR1CRRR
132386  4s rcrr1rrr/rmrddc2/6hh/6E1/6e1/1H4H1/1RDCMD1R/RRR1CRRR
132386  5g rcrr1rrr/rmrddc2/6hh/6E1/6eH/1H6/1RDCMD1R/RRR1CRRR
132386  5s rcrr1rrr/rmrddc2/6hh/6E1/6e1/1H4H1/1RDCMD1R/RRR1CRRR
132386  6g rcrr1rrr/rmrddc2/6hh/6E1/6eH/1H6/1RDCMD1R/RRR1CRRR

# print the first ten non-repeated positions of game 132386
$ good-human-games | grep -w ^132386 | positions -s | head
132386  1s 8/8/8/8/8/8/RHDMEDHR/RRRCCRRR
132386  2g rrrdrrcr/hhcedrmr/8/8/8/8/RHDMEDHR/RRRCCRRR
132386  2s rrrdrrcr/hhcedrmr/8/4E3/8/1H6/R1DM1DHR/RRRCCRRR
132386  3g rrr1rrcr/h1cddrmr/1h6/3eE3/8/1H6/R1DM1DHR/RRRCCRRR
132386  3s rrr1rrcr/h1cddrmr/1h6/3eE3/8/1H4H1/1RDMCDR1/RRRC1RRR
132386  4g rrr1rrcr/2cddrmr/hh6/4E3/1e6/1H4H1/1RDMCDR1/RRRC1RRR
132386  4s rrr1rrcr/2cddrmr/hh6/1E6/1e6/1H4H1/R1DMCDR1/RRRC1RRR
132386  5g rrr1rrcr/2cddrmr/hh6/1E6/He6/6H1/R1DMCDR1/RRRC1RRR
132386  6s rrr1rrcr/2cddrmr/hh6/1E6/1e6/H2M2H1/R1D1CDR1/RRRC1RRR
132386  7g rrr1rr1r/2cddrcr/1h4m1/hE6/8/He1M2H1/R1D1CDR1/RRRC1RRR

# print the final position of the longest good human game
$ good-human-games | positions | sort -k2 -rn | head -1
53484   136g 8/5c2/1H1rE1d1/8/5De1/1C2D3/8/7r

# print the corresponding normalized position (since camels were missing)
$ good-human-games | grep -w ^53484 | positions -n | tail -1
53484   136g 8/5c2/1H1rM1d1/8/5Dm1/1C2D3/8/7r

Limitations

fetch-new-games is very slow when it has to rebuild the games database. I should probably use SQLite instead, as in Pastebin example.