Dataset Preview
Duplicate
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
Job has been terminated due to a temporary spike in resource usage and may be restarted later.
Error code:   JobManagerCrashedError

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

table_id
int64
player_id
int64
player_name
string
faction
string
bid_starting_vp
int64
is_winning_bid
bool
100,002,164
31,653,829
fyt3029888
darklings
25
false
100,002,164
33,421,869
insertcoin1
mermaids
30
true
100,002,164
84,548,886
okborn
darklings
24
true
100,002,164
85,444,399
Ginooo
cultists
24
false
100,002,164
31,653,829
fyt3029888
cultists
23
true
100,002,164
85,444,399
Ginooo
engineers
30
true
100,002,615
85,784,708
Linhp
halflings
26
true
100,002,615
85,134,274
jobunchuang
engineers
17
false
100,002,615
85,065,629
sexykodo
engineers
16
true
100,002,615
85,604,495
Vali1590
engineers
20
false
100,002,615
85,134,274
jobunchuang
darklings
21
false
100,002,615
85,065,629
sexykodo
darklings
25
false
100,002,615
85,604,495
Vali1590
darklings
20
true
100,002,615
85,134,274
jobunchuang
halflings
27
false
100,002,615
85,134,274
jobunchuang
swarmlings
30
true
100,017,428
85,438,799
Nitroustorm
dwarves
30
true
100,017,428
87,200,623
haydeezz
auren
29
true
100,017,428
84,947,359
Divino Otelma
darklings
20
false
100,017,428
85,395,108
SHIPxIT
darklings
19
true
100,017,428
84,947,359
Divino Otelma
halflings
30
true
100,020,299
86,063,751
Ghostly 1
witches
29
true
100,020,299
84,775,213
kimsan
cultists
30
true
100,020,299
84,309,199
bananafish23
fakirs
30
true
100,020,299
85,065,629
sexykodo
mermaids
30
true
100,022,956
84,943,102
flower2886
nomads
30
true
100,022,956
87,042,230
lactia
swarmlings
28
true
100,022,956
84,813,650
rafaelhfuryu
dwarves
30
true
100,022,956
85,380,490
wasgibts
swarmlings
29
false
100,022,956
85,380,490
wasgibts
alchemists
30
true
100,033,838
84,156,873
AndersonSousa
swarmlings
14
true
100,033,838
87,435,107
Myoblast
dwarves
22
true
100,033,838
27,568,577
kevin980585
cultists
20
true
100,033,838
84,962,776
riedmann
swarmlings
15
false
100,033,838
84,962,776
riedmann
dwarves
23
false
100,033,838
87,435,107
Myoblast
cultists
21
false
100,033,838
27,568,577
kevin980585
swarmlings
21
false
100,033,838
84,962,776
riedmann
nomads
30
true
100,058,797
82,120,832
545747544574
dwarves
24
true
100,058,797
85,604,495
Vali1590
darklings
18
false
100,058,797
87,435,107
Myoblast
darklings
11
false
100,058,797
86,063,570
haski33
darklings
10
true
100,058,797
85,604,495
Vali1590
dwarves
25
false
100,058,797
82,120,832
545747544574
cultists
18
false
100,058,797
86,063,570
haski33
cultists
17
false
100,058,797
85,604,495
Vali1590
cultists
16
true
100,058,797
87,435,107
Myoblast
swarmlings
30
true
100,071,989
85,444,399
Ginooo
engineers
8
true
100,071,989
85,605,063
simonbe
darklings
14
false
100,071,989
84,792,460
Captain_Hannibal
fakirs
30
true
100,071,989
85,302,633
Mertz
engineers
9
false
100,071,989
85,302,633
Mertz
darklings
13
true
100,071,989
85,605,063
simonbe
engineers
17
false
100,071,989
85,605,063
simonbe
swarmlings
30
true
100,159,275
84,240,656
tttomleung
engineers
14
true
100,159,275
84,060,015
Nachsville
engineers
15
false
100,159,275
85,425,468
kab_c
engineers
21
false
100,159,275
84,156,873
AndersonSousa
cultists
15
true
100,159,275
85,425,468
kab_c
cultists
24
false
100,159,275
85,425,468
kab_c
swarmlings
26
false
100,159,275
84,240,656
tttomleung
cultists
16
false
100,159,275
84,060,015
Nachsville
swarmlings
25
true
100,159,275
85,425,468
kab_c
darklings
30
true
100,181,158
86,432,732
Kongpop1
cultists
15
false
100,181,158
85,605,063
simonbe
darklings
12
false
100,181,158
85,604,495
Vali1590
darklings
11
true
100,181,158
85,134,274
jobunchuang
cultists
19
false
100,181,158
86,432,732
Kongpop1
darklings
16
false
100,181,158
85,605,063
simonbe
cultists
14
true
100,181,158
85,134,274
jobunchuang
engineers
20
true
100,181,158
86,432,732
Kongpop1
engineers
21
false
100,181,158
86,432,732
Kongpop1
witches
30
true
100,182,677
85,257,822
zitcob
engineers
23
false
100,182,677
84,947,359
Divino Otelma
darklings
16
false
100,182,677
85,107,462
tac
engineers
17
false
100,182,677
84,230,883
peeman
engineers
16
true
100,182,677
85,107,462
tac
darklings
15
true
100,182,677
85,257,822
zitcob
darklings
25
false
100,182,677
84,230,883
peeman
mermaids
28
false
100,182,677
85,257,822
zitcob
mermaids
27
true
100,182,677
84,947,359
Divino Otelma
chaosmagicians
30
true
100,185,597
85,251,103
Neramoor
halflings
30
true
100,185,597
86,125,352
Fatpanda1985
darklings
12
true
100,185,597
85,604,495
Vali1590
darklings
13
false
100,185,597
85,106,786
blue_sky27
giants
30
true
100,185,597
85,604,495
Vali1590
engineers
30
true
100,190,140
56,035,322
Hefaistos
engineers
23
false
100,190,140
84,363,860
Vignesh85
engineers
14
false
100,190,140
85,192,206
daisakuikeda
engineers
13
true
100,190,140
87,042,230
lactia
cultists
20
true
100,190,140
56,035,322
Hefaistos
cultists
21
false
100,190,140
87,042,230
lactia
darklings
24
false
100,190,140
84,363,860
Vignesh85
darklings
20
true
100,190,140
56,035,322
Hefaistos
swarmlings
30
true
100,190,389
84,548,886
okborn
darklings
22
false
100,190,389
85,020,830
strandbaer
engineers
16
true
100,190,389
87,435,107
Myoblast
darklings
11
true
100,190,389
83,866,793
volrath
engineers
17
false
100,190,389
85,020,830
strandbaer
darklings
12
false
100,190,389
84,548,886
okborn
cultists
16
true
100,190,389
83,866,793
volrath
cultists
17
false
End of preview.

BGA Replays

Parsed play-by-play data from BoardGameArena replays for Terra Mystica and Tokaido. One row per player per game (results), one row per player per round (Terra Mystica only), one row per scoring/economy action (moves), and, for Terra Mystica only, one row per power-leech offer and one row per player per faction bid in the setup auction.

Generated by bga-replay-parser, which scrapes raw replays from BGA and extracts structured CSVs.

Contents

terra_mystica/
  player_results.csv     # one row per player per game (63 cols)
  round_summaries.csv    # one row per player per round (55 cols)
  moves.csv              # one row per action (35 cols)
  leech_offers.csv       # one row per leech offer per offered player, incl. declines (13 cols)
  auction_bids.csv       # one row per player per faction bid in auction games (6 cols)
  setup.csv              # one row per game (game settings + final outcomes, 25 cols)

tokaido/
  player_results_base.csv     # one row per player per base game (42 cols)
  player_results_xroads.csv   # one row per player per Crossroads game (61 cols)
  moves_base.csv              # one row per action in base games
  moves_xroads.csv            # one row per action in Crossroads games

examples/
  terra_mystica_quickstart.ipynb   # load CSVs, faction win rate, plot
  tokaido_quickstart.ipynb         # load CSVs, traveler win rate, plot

Tokaido is split by base / Crossroads expansion because the two formats differ enough that mixing them adds noise (19 Crossroads-only columns would be 0 in every base row).

Stats (current snapshot)

Game Games Player-rows Round-rows Move-rows
Terra Mystica 30,582 122,328 733,956 15,897,975
Tokaido (base) 48,861 195,444 — 9,355,887
Tokaido (Crossroads) 18,526 74,104 — 4,066,793

Terra Mystica leech_offers.csv: 2,297,442 offer-rows (every game has some). auction_bids.csv: 253,099 bid-rows across 28,597 auction games.

Source: game histories of top Arena-leaderboard players on BoardGameArena (Arena and Normal mode tables). All games are 4-player. Terra Mystica games span 2020-01 to 2026-10; Tokaido games span 2016-05 to 2026-05.

How to load

Direct from the dataset over HTTPS:

import pandas as pd
URL = "https://huggingface.co/datasets/liamdj/bga-replays/resolve/main"
tm_players  = pd.read_csv(f"{URL}/terra_mystica/player_results.csv")
tok_players = pd.read_csv(f"{URL}/tokaido/player_results_base.csv")

Or git clone the dataset and read files locally (faster for repeated reads):

git clone https://huggingface.co/datasets/liamdj/bga-replays

examples/ contains two runnable quickstart notebooks (load CSVs, compute a basic win-rate aggregate, plot). Terra Mystica moves.csv is 2.2 GB (15.9M rows): read it with the category/int16/float32 dtypes shown in the Terra Mystica notebook (about 1.3 GB in memory), since pandas' default dtypes need well over 10 GB.

Schema notes

Terra Mystica player_results.csv includes a full VP breakdown (vp_round_scoring, vp_favor_tiles, vp_bonus_cards, vp_towns, vp_faction_ability, …), structures built, cult positions, total resource flow, and the tiles/cards held. starting_vp plus the vp_* components (excluding vp_generated/vp_spent/vp_converted) equals final_vp (all but 2 player-rows). vp_power_leaching is ≤ 0.

Terra Mystica round_summaries.csv captures per-round income, cult bonuses received, actions taken, power gained by source, resource flow (coins/workers/priests/power _generated, _spent, _converted), end-of-round state, and pass order. The per-round _converted columns sum to the game totals in player_results.csv.

Terra Mystica moves.csv has one row per logged action, including non-turn rows (income, scoring, leech gains, free conversions). move_number is a row sequence, not a turn index. turn_number is the turn within the round, counted across all players and restarting at 1 each round, so (table_id, round_number, turn_number) identifies one turn; every row produced during that turn shares it (the active player's conversions, action and follow-ups, and other players' leech gains). It is blank outside the action phase (setup, income, cleanup, final scoring). Free-action conversions are action_type == "convert" rows.

Terra Mystica leech_offers.csv has one row per "power via structures" offer per offered player, including declines (which leave no row in moves.csv). offer_id numbers the leech phases within a game; round_number/turn_number are the builder's turn and join to moves.csv. builder_id built or upgraded, target_id was offered power. offered is the power the target could take, cap is the extra adjacency power cut off because their bowls were too full, and n_offered is the number of targets in that phase with offered > 0. auto = 1 marks a free 1-power leech BGA took without asking. response is accept / decline / none; gained is the power taken and vp_cost the VP paid (positive). Per player, vp_cost sums to -vp_power_leaching in player_results.csv.

Terra Mystica setup.csv has one row per game: board, options, scoring tiles, bonus cards, winner, and ;-joined per-player lists. game_mode is Arena mode or Normal mode. auction_type (fast / slow / empty) is derived from the bids in the log and is the reliable auction indicator; starting_vp_setting is BGA's raw option label and is unreliable before 2023 (many 2020–22 games labelled "Adjusted by Faction" were auctions).

Terra Mystica auction_bids.csv has one row per (player, faction) with bid_starting_vp = the starting VP the player offered to accept for that faction (lower = stronger bid). Fast (sealed) auction: every player bids once on every faction. Slow (standard) auction: the row is the player's last bid on that faction, and only factions they bid on appear. is_winning_bid marks the pair the auction assigned (one per player). Join to setup.csv on table_id for auction_type.

Tokaido player_results_*.csv has a VP breakdown (vp_* columns, including vp_temple_rank) that sums to final_score (all but 9 player-rows); the preparations column flags games using the Preparations expansion. Donation routing quirk: Miko encounter VP and Devotion-amulet VP both go to vp_temple; there is no vp_amulet_used column.

Excluded games: Terra Mystica replays with variable-/volcano-faction games and the Landscapes expansion are skipped at parse time, as are Friendly-mode games. Tokaido games with fewer than 3 players are skipped.

Full data-model docs (every column, every quirk) are in the parser repo's docs/ folder (terra_mystica/parser.md, tokaido/parser.md).

Licensing & terms

The parsing, structuring, and aggregation work is licensed under CC BY-NC 4.0 — free to use and adapt for non-commercial purposes, provided you credit this dataset.

Source data is derived from BoardGameArena replays. Users are responsible for complying with BoardGameArena's Terms of Service when working with this dataset. Raw replays are not redistributed here.

Citation

If you use this dataset in a writeup, please cite:

@misc{bga-replays,
  author       = {Liam},
  title        = {BGA Replays: parsed Terra Mystica and Tokaido game data from BoardGameArena},
  year         = {2026},
  publisher    = {Hugging Face},
  url          = {https://huggingface.co/datasets/liamdj/bga-replays}
}
Downloads last month
67