Datasets:
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
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.
episode_id string | split_group_id string | source_id string | source_file string | source_episode int64 | source_line int64 | generation int64 | policy string | scenario string | opponent string | step int64 | action_id int64 | action_index int64 | action_kind string | card_id string | action_json string | reward float64 | terminated bool | truncated bool | episode_outcome string | state_json string | next_state_json string | legal_actions_json string | teacher_label_usable bool | original_record_json string | obs list | next_obs list |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
historical_0146ca0d32d20f17bbf8aec8330abb7200d3fe884c908e5b509309f0d18b4eda | 0146ca0d32d20f17bbf8aec8330abb7200d3fe884c908e5b509309f0d18b4eda | 14bb812decbb86c5b04adcbd98189f67800ebe574610b29bb005877056f275d9 | consequence_teacher_argmax_v32_combined.jsonl | 5 | 1,461 | 1 | search_teacher | hunter_core_m14_fullhand | fast_bot | 0 | 4,261 | 185 | card | pawns_path | "{\"capture_value\":0,\"card_id\":\"pawns_path\",\"card_index\":3,\"card_name\":\"Pawns Path\",\"com(...TRUNCATED) | 17 | false | false | win | "{\"board\":[{\"piece\":\"pawn\",\"side\":\"friendly\",\"square\":\"a2\"},{\"piece\":\"pawn\",\"side(...TRUNCATED) | "{\"board\":[{\"piece\":\"pawn\",\"side\":\"friendly\",\"square\":\"a2\"},{\"piece\":\"pawn\",\"side(...TRUNCATED) | "[{\"capture_value\":0,\"from\":[0,6],\"id\":3104,\"is_capture\":false,\"is_castle\":false,\"is_king(...TRUNCATED) | false | "{\"episode\":5,\"step\":0,\"obs\":[0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,(...TRUNCATED) | [0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0(...TRUNCATED) | [0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0(...TRUNCATED) |
historical_0146ca0d32d20f17bbf8aec8330abb7200d3fe884c908e5b509309f0d18b4eda | 0146ca0d32d20f17bbf8aec8330abb7200d3fe884c908e5b509309f0d18b4eda | 14bb812decbb86c5b04adcbd98189f67800ebe574610b29bb005877056f275d9 | consequence_teacher_argmax_v32_combined.jsonl | 5 | 1,462 | 1 | search_teacher | hunter_core_m14_fullhand | fast_bot | 1 | 4,269 | 193 | card | wall_bounce | "{\"capture_value\":0,\"card_id\":\"wall_bounce\",\"card_index\":2,\"card_name\":\"Wall Bounce\",\"c(...TRUNCATED) | -1 | false | false | win | "{\"board\":[{\"piece\":\"pawn\",\"side\":\"friendly\",\"square\":\"a2\"},{\"piece\":\"pawn\",\"side(...TRUNCATED) | "{\"board\":[{\"piece\":\"pawn\",\"side\":\"friendly\",\"square\":\"a2\"},{\"piece\":\"pawn\",\"side(...TRUNCATED) | "[{\"capture_value\":0,\"from\":[0,6],\"id\":3104,\"is_capture\":false,\"is_castle\":false,\"is_king(...TRUNCATED) | false | "{\"episode\":5,\"step\":1,\"obs\":[0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,(...TRUNCATED) | [0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0(...TRUNCATED) | [0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0(...TRUNCATED) |
historical_0146ca0d32d20f17bbf8aec8330abb7200d3fe884c908e5b509309f0d18b4eda | 0146ca0d32d20f17bbf8aec8330abb7200d3fe884c908e5b509309f0d18b4eda | 14bb812decbb86c5b04adcbd98189f67800ebe574610b29bb005877056f275d9 | consequence_teacher_argmax_v32_combined.jsonl | 5 | 1,463 | 1 | search_teacher | hunter_core_m14_fullhand | fast_bot | 2 | 4,252 | 176 | card | omni_queen | "{\"capture_value\":0,\"card_id\":\"omni_queen\",\"card_index\":0,\"card_name\":\"Omni Queen\",\"com(...TRUNCATED) | -1 | false | false | win | "{\"board\":[{\"piece\":\"pawn\",\"side\":\"friendly\",\"square\":\"a2\"},{\"piece\":\"pawn\",\"side(...TRUNCATED) | "{\"board\":[{\"piece\":\"pawn\",\"side\":\"friendly\",\"square\":\"a2\"},{\"piece\":\"pawn\",\"side(...TRUNCATED) | "[{\"capture_value\":0,\"from\":[0,6],\"id\":3104,\"is_capture\":false,\"is_castle\":false,\"is_king(...TRUNCATED) | false | "{\"episode\":5,\"step\":2,\"obs\":[0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,(...TRUNCATED) | [0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0(...TRUNCATED) | [0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0(...TRUNCATED) |
historical_0146ca0d32d20f17bbf8aec8330abb7200d3fe884c908e5b509309f0d18b4eda | 0146ca0d32d20f17bbf8aec8330abb7200d3fe884c908e5b509309f0d18b4eda | 14bb812decbb86c5b04adcbd98189f67800ebe574610b29bb005877056f275d9 | consequence_teacher_argmax_v32_combined.jsonl | 5 | 1,464 | 1 | search_teacher | hunter_core_m14_fullhand | fast_bot | 3 | 3,779 | 21 | move | "{\"capture_value\":90,\"from\":[3,7],\"id\":3779,\"is_capture\":true,\"is_castle\":false,\"is_king_(...TRUNCATED) | 369 | false | false | win | "{\"board\":[{\"piece\":\"pawn\",\"side\":\"friendly\",\"square\":\"a2\"},{\"piece\":\"pawn\",\"side(...TRUNCATED) | "{\"board\":[{\"piece\":\"pawn\",\"side\":\"friendly\",\"square\":\"a2\"},{\"piece\":\"pawn\",\"side(...TRUNCATED) | "[{\"capture_value\":0,\"from\":[0,6],\"id\":3104,\"is_capture\":false,\"is_castle\":false,\"is_king(...TRUNCATED) | false | "{\"episode\":5,\"step\":3,\"obs\":[0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,(...TRUNCATED) | [0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0(...TRUNCATED) | [0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0(...TRUNCATED) | |
historical_0146ca0d32d20f17bbf8aec8330abb7200d3fe884c908e5b509309f0d18b4eda | 0146ca0d32d20f17bbf8aec8330abb7200d3fe884c908e5b509309f0d18b4eda | 14bb812decbb86c5b04adcbd98189f67800ebe574610b29bb005877056f275d9 | consequence_teacher_argmax_v32_combined.jsonl | 5 | 1,465 | 1 | search_teacher | hunter_core_m14_fullhand | fast_bot | 4 | 4,109 | 13 | card | law_of_mid_exchange | "{\"capture_value\":40,\"card_id\":\"law_of_mid_exchange\",\"card_index\":0,\"card_name\":\"Law of M(...TRUNCATED) | 64 | false | false | win | "{\"board\":[{\"piece\":\"pawn\",\"side\":\"friendly\",\"square\":\"a2\"},{\"piece\":\"pawn\",\"side(...TRUNCATED) | "{\"board\":[{\"piece\":\"pawn\",\"side\":\"friendly\",\"square\":\"a2\"},{\"piece\":\"pawn\",\"side(...TRUNCATED) | "[{\"id\":4096,\"kind\":\"end_turn\",\"label\":\"end_turn\"},{\"capture_value\":50,\"card_id\":\"law(...TRUNCATED) | false | "{\"episode\":5,\"step\":4,\"obs\":[0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,(...TRUNCATED) | [0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0(...TRUNCATED) | [0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0(...TRUNCATED) |
historical_0146ca0d32d20f17bbf8aec8330abb7200d3fe884c908e5b509309f0d18b4eda | 0146ca0d32d20f17bbf8aec8330abb7200d3fe884c908e5b509309f0d18b4eda | 14bb812decbb86c5b04adcbd98189f67800ebe574610b29bb005877056f275d9 | consequence_teacher_argmax_v32_combined.jsonl | 5 | 1,466 | 1 | search_teacher | hunter_core_m14_fullhand | fast_bot | 5 | 4,098 | 2 | card | unholy_attack | "{\"capture_value\":0,\"card_id\":\"unholy_attack\",\"card_index\":0,\"card_name\":\"Unholy Attack\"(...TRUNCATED) | -11 | false | false | win | "{\"board\":[{\"piece\":\"pawn\",\"side\":\"friendly\",\"square\":\"a2\"},{\"piece\":\"pawn\",\"side(...TRUNCATED) | "{\"board\":[{\"piece\":\"pawn\",\"side\":\"friendly\",\"square\":\"a2\"},{\"piece\":\"pawn\",\"side(...TRUNCATED) | "[{\"id\":4096,\"kind\":\"end_turn\",\"label\":\"end_turn\"},{\"capture_value\":0,\"card_id\":\"unho(...TRUNCATED) | false | "{\"episode\":5,\"step\":5,\"obs\":[0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,(...TRUNCATED) | [0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0(...TRUNCATED) | [0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0(...TRUNCATED) |
historical_0146ca0d32d20f17bbf8aec8330abb7200d3fe884c908e5b509309f0d18b4eda | 0146ca0d32d20f17bbf8aec8330abb7200d3fe884c908e5b509309f0d18b4eda | 14bb812decbb86c5b04adcbd98189f67800ebe574610b29bb005877056f275d9 | consequence_teacher_argmax_v32_combined.jsonl | 5 | 1,467 | 1 | search_teacher | hunter_core_m14_fullhand | fast_bot | 6 | 3,724 | 0 | move | "{\"capture_value\":10,\"from\":[2,7],\"id\":3724,\"is_capture\":true,\"is_castle\":false,\"is_king_(...TRUNCATED) | 19 | false | false | win | "{\"board\":[{\"piece\":\"pawn\",\"side\":\"friendly\",\"square\":\"a2\"},{\"piece\":\"pawn\",\"side(...TRUNCATED) | "{\"board\":[{\"piece\":\"pawn\",\"side\":\"friendly\",\"square\":\"a2\"},{\"piece\":\"pawn\",\"side(...TRUNCATED) | "[{\"capture_value\":10,\"from\":[2,7],\"id\":3724,\"is_capture\":true,\"is_castle\":false,\"is_king(...TRUNCATED) | false | "{\"episode\":5,\"step\":6,\"obs\":[0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,(...TRUNCATED) | [0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0(...TRUNCATED) | [0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0(...TRUNCATED) | |
historical_0146ca0d32d20f17bbf8aec8330abb7200d3fe884c908e5b509309f0d18b4eda | 0146ca0d32d20f17bbf8aec8330abb7200d3fe884c908e5b509309f0d18b4eda | 14bb812decbb86c5b04adcbd98189f67800ebe574610b29bb005877056f275d9 | consequence_teacher_argmax_v32_combined.jsonl | 5 | 1,468 | 1 | search_teacher | hunter_core_m14_fullhand | fast_bot | 7 | 4,096 | 0 | end_turn | {"id":4096,"kind":"end_turn","label":"end_turn"} | -41 | false | false | win | "{\"board\":[{\"piece\":\"pawn\",\"side\":\"friendly\",\"square\":\"a2\"},{\"piece\":\"pawn\",\"side(...TRUNCATED) | "{\"board\":[{\"piece\":\"pawn\",\"side\":\"friendly\",\"square\":\"a2\"},{\"piece\":\"pawn\",\"side(...TRUNCATED) | [{"id":4096,"kind":"end_turn","label":"end_turn"}] | false | "{\"episode\":5,\"step\":7,\"obs\":[0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,(...TRUNCATED) | [0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0(...TRUNCATED) | [0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0(...TRUNCATED) | |
historical_0146ca0d32d20f17bbf8aec8330abb7200d3fe884c908e5b509309f0d18b4eda | 0146ca0d32d20f17bbf8aec8330abb7200d3fe884c908e5b509309f0d18b4eda | 14bb812decbb86c5b04adcbd98189f67800ebe574610b29bb005877056f275d9 | consequence_teacher_argmax_v32_combined.jsonl | 5 | 1,469 | 1 | search_teacher | hunter_core_m14_fullhand | fast_bot | 8 | 196 | 2 | move | "{\"capture_value\":0,\"from\":[3,0],\"id\":196,\"is_capture\":false,\"is_castle\":false,\"is_king_h(...TRUNCATED) | 179 | false | false | win | "{\"board\":[{\"piece\":\"pawn\",\"side\":\"friendly\",\"square\":\"a2\"},{\"piece\":\"pawn\",\"side(...TRUNCATED) | "{\"board\":[{\"piece\":\"pawn\",\"side\":\"friendly\",\"square\":\"a2\"},{\"piece\":\"pawn\",\"side(...TRUNCATED) | "[{\"capture_value\":30,\"from\":[3,0],\"id\":193,\"is_capture\":true,\"is_castle\":false,\"is_king_(...TRUNCATED) | false | "{\"episode\":5,\"step\":8,\"obs\":[0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,(...TRUNCATED) | [0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0(...TRUNCATED) | [0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0(...TRUNCATED) | |
historical_0146ca0d32d20f17bbf8aec8330abb7200d3fe884c908e5b509309f0d18b4eda | 0146ca0d32d20f17bbf8aec8330abb7200d3fe884c908e5b509309f0d18b4eda | 14bb812decbb86c5b04adcbd98189f67800ebe574610b29bb005877056f275d9 | consequence_teacher_argmax_v32_combined.jsonl | 5 | 1,470 | 1 | search_teacher | hunter_core_m14_fullhand | fast_bot | 9 | 4,097 | 1 | card | extended_range | "{\"capture_value\":0,\"card_id\":\"extended_range\",\"card_index\":0,\"card_name\":\"Extended Range(...TRUNCATED) | -1 | false | false | win | "{\"board\":[{\"piece\":\"pawn\",\"side\":\"friendly\",\"square\":\"a2\"},{\"piece\":\"pawn\",\"side(...TRUNCATED) | "{\"board\":[{\"piece\":\"pawn\",\"side\":\"friendly\",\"square\":\"a2\"},{\"piece\":\"pawn\",\"side(...TRUNCATED) | "[{\"id\":4096,\"kind\":\"end_turn\",\"label\":\"end_turn\"},{\"capture_value\":0,\"card_id\":\"exte(...TRUNCATED) | false | "{\"episode\":5,\"step\":9,\"obs\":[0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,(...TRUNCATED) | [0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0(...TRUNCATED) | [0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0,0.0(...TRUNCATED) |
Chessovania historical 8x8 card gameplay
This release contains historical experience produced by earlier Chessovania bot experiments. The collection includes imperfect behavior and losses. Recorded actions are observations of behavior, and their strategic quality has not been established. New experience generated by a frozen Qwen agent should be published with its own provenance and dataset version.
The export contains 13,402 unique episodes, 741,371 transitions, and 193,395 recorded card actions.
Every retained episode starts at step zero, ends in termination, contains at least one selected legal card action, and passes the exporter's structural checks. Export with --include-truncated also retains explicitly truncated episodes. Entire episodes are retained, including ordinary moves and end-turn decisions. Games with zero card actions, unsupported board sizes, missing action metadata, malformed records, invalid actions, safety caps, or broken observation continuity are excluded.
Loading
from datasets import load_dataset
data = load_dataset("YOUR_USERNAME/chessovania-historical-8x8-cards", "transitions")
episodes = load_dataset("YOUR_USERNAME/chessovania-historical-8x8-cards", "episodes")
record = data["train"][0]
import json
state = json.loads(record["state_json"])
original = json.loads(record["original_record_json"])
For a local export use load_dataset('json', data_files={'train': 'data/transitions/train/*.jsonl.gz'}). The README features declare nullable numeric columns explicitly for Hub loading.
Representation and traceability
The transitions configuration has one row per decision. The episodes configuration has one row per game and joins through episode_id. source_id is the source file SHA-256. source_file and source_line identify the original local record. original_record_json preserves the input record including policy values, search targets, teacher feedback, and diagnostic fields when available. Arbitrary metadata is stored in JSON strings to keep the tabular schema consistent across runs.
obs and next_obs preserve the 803-value numeric observations. state_json decodes board pieces, 16 scalars, five hand slots, and the two encoded effect slots. Squares use a8=(0,0) and h1=(7,7). Friendly/enemy piece channels retain the acting player's perspective. Hand card IDs are resolved only from legal metadata for that slot. Other IDs remain null because historical numeric registry indices cannot safely be mapped through today's card registry. Encoded playable flags and legal card-action availability are recorded separately.
Splits and deduplication
Complete observed trajectories are deduplicated across sources using observations, selected action semantics, rewards, outcomes, and collection scenario/opponent. File hashes also eliminate identical input files. These hashes establish observed trajectory identity within the stored representation. Hidden environment state is unavailable. Duplicate origins are recorded in metadata/duplicates.jsonl.gz.
Splits use seed 123, validation fraction 0.1, and test fraction 0.1. The observed trajectory fingerprint determines each whole game's split. Copies across sources and generations are deduplicated before splitting. Different splits can share initial and later positions. These historical splits support held-out episode analysis within the collected conditions. Generalization to new opponents/configurations requires a separate live evaluation protocol. Keep test experience outside memory construction and prompt examples.
Intended use with Qwen
Use training-split trajectories to build external memory, compare retrieved experience with structured lessons, or develop state/action interfaces. Decisions and outcomes are available for controlled downstream experiments. A loss provides game-level feedback. It does not identify every erroneous move. Any derived lesson should retain episode_id and step references. For the proposed frozen-model study, learning occurs through external memory. Parameter training would be a separate experimental condition. The release contains behavioral traces with no invented reasoning or verified optimal-move labels.
Provenance and limitations
metadata/sources.json records available policy evidence, collection settings, and summary hashes. Generation identifies the candidate training iteration. The behavior model hash identifies the parent policy that collected its experience when that hash was recorded. Unknown fields remain explicit. Source collection revisions were not recorded and compatibility with current gameplay rules has not been established by replay. The exporter records the checkout revision and hashes relevant current files as decoding references. The numeric observation is a partial view of the environment and omits some active-effect state and opponent decisions. Structural validation does not certify game-engine correctness or strategic skill. Filtering for card use changes the distribution, so reported statistics describe the selected games.
teacher_label_usable identifies a valid confidence-gated single-action teacher index. Search distributions remain in original_record_json and require their own target validation. Teacher errors and skipped labels are retained as historical diagnostics. Rewards follow historical collection runs and include shaping. Current reward code is included as reference metadata.
Statistics
{
"action_kinds": {
"card": 193395,
"end_turn": 269939,
"move": 278037
},
"card_actions": 193395,
"cards": {
"cursed_teleport": 23195,
"curved_shot": 23911,
"demolished_canon": 7247,
"extended_range": 14387,
"horse_de_mutation": 10758,
"impostor": 10572,
"law_of_mid_exchange": 14048,
"omni_queen": 5190,
"pawns_path": 20463,
"prepared_strike": 24779,
"rewind": 5066,
"temporary_monarch": 10377,
"unholy_attack": 8976,
"wall_bounce": 14426
},
"configuration_rows": {
"episodes/test": 1328,
"episodes/train": 10716,
"episodes/validation": 1358,
"transitions/test": 74141,
"transitions/train": 592160,
"transitions/validation": 75070
},
"episodes": 13402,
"excluded": {
"duplicate_episode": 683,
"missing_action_metadata": 5,
"next_action_mapping": 8,
"no_card_actions": 28
},
"input_records": 784819,
"invalid_json_records": 0,
"opponents": {
"fast_bot": 13328,
"stockfish": 1,
"unknown": 73
},
"outcomes": {
"draw": 284,
"loss": 9620,
"win": 3498
},
"policies": {
"metadata_greedy": 495,
"neural_policy": 12255,
"search_teacher": 574,
"unknown": 78
},
"scenarios": {
"hunter_core_m10": 11653,
"hunter_core_m14_fullhand": 1676,
"unknown": 73
},
"splits": {
"test": 1328,
"train": 10716,
"validation": 1358
},
"teacher_sources": {
"king_hit_metadata": 632,
"skipped_check_state": 11221,
"skipped_no_matching_stockfish_move": 88,
"skipped_no_move_actions": 18923,
"stockfish": 14584,
"teacher_error": 9
},
"transitions": 741371
}
License
This dataset and its accompanying code are released under the MIT License.
- Downloads last month
- 80