Command Reference

This document explains each executable command in the ATC project.

Prerequisites

cd /Users/shion/workspace/crypto-auto-trading
uv sync --extra dev

1) `atc` CLI commands

The package exposes the atc entrypoint via:

uv run atc ...

1.1 `atc ingest historical`

Purpose:

Download GMO historical trade CSV gzip files.
Validate schema and data quality.
Write raw files + manifests.
Convert to canonical Parquet events.
If raw data already exists locally, skip network acquisition and reuse local files.

Command:

uv run atc ingest historical --symbol ETH_JPY --start 2026-01-01 --end 2026-01-31

Arguments:

--symbol (required): Symbol name (e.g. ETH_JPY).
--start (required): Start business date in YYYY-MM-DD.
--end (required): End business date in YYYY-MM-DD.
--data-root (optional, default data): Base directory for raw/canonical/derived data.

Typical output (JSON):

symbol
start
end
files_processed
files_downloaded
rows_processed

Data written:

Raw gzip: data/raw/historical_trades/{symbol}/{YYYY}/{MM}/{YYYYMMDD}_{symbol}.csv.gz
Manifest: data/raw/historical_trades/{symbol}/{YYYY}/{MM}/{YYYYMMDD}_{symbol}.manifest.json
Canonical Parquet: data/canonical/events/symbol={symbol}/year={YYYY}/month={M}/day={D}/...

1.2 `atc backtest`

Purpose:

Load canonical events.
Run event-driven strategy simulation.
Apply risk checks and execution simulation.
Simulate execution as limit-only with configurable order latency.
Evaluate alpha metrics and save a report.

Command:

uv run atc backtest --symbol ETH_JPY --start 2022-01-01 --end 2022-12-31 --strategy baseline_momentum_micro

With trade overlay chart generation:

uv run atc backtest --symbol ETH_JPY --strategy baseline_momentum_micro --plot --plot-interval 1m

With both HTML and PNG at once:

uv run atc backtest --symbol ETH_JPY --strategy baseline_momentum_micro --plot --plot-format both --plot-interval 1m

With explicit order latency:

uv run atc backtest --symbol ETH_JPY --strategy baseline_momentum_micro --order-latency-sec 0.3

With slower order cadence (5 minutes between submissions):

uv run atc backtest --symbol ETH_JPY --strategy baseline_momentum_micro --min-order-interval-sec 300

Arguments:

--symbol (required)
--start (optional, default 2021-05-01): business date (JST, 06:00 boundary).
--end (optional, default 2026-01-31): business date (JST, 06:00 boundary).
--strategy (required)
--data-root (optional, default data)
--plot (optional): Generate an overlay chart (price + order markers).
--plot-output (optional): Output chart path (.html or .png).
--plot-format (optional, default html): html, png, or both.
--plot-interval (optional, default 1m): Derived bar interval for plotting.
--order-latency-sec (optional, default 0.3): Minimum delay before a submitted limit order becomes eligible for fill checks.
--min-order-interval-sec (optional, default 300): Minimum interval between new order submissions.
--initial-position-eth (optional, default 0.0): Starting position for the run. Useful for day-by-day chained analysis.
--initial-cash-jpy (optional, default 0.0): Starting cash for the run. Pair with --initial-position-eth to carry portfolio state.
--target-abs-qty-eth (optional): Override strategy target absolute quantity (for strategies that support target_abs_qty).
--max-abs-position-eth (optional): Override risk hard limit max_abs_position_eth.
--benchmark-hold-eth (optional, default 10.0): Fixed buy-and-hold benchmark quantity in ETH for alpha comparison.
--enforce-session-close (optional, default disabled): Enable 05:59-style close constraints.
--no-new-entry-minutes (optional, default 10): Applied only when --enforce-session-close is enabled.
--force-close-start-jst (optional, default 05:55:00): Applied only when --enforce-session-close is enabled.
--close-deadline-jst (optional, default 05:59:50): Applied only when --enforce-session-close is enabled.
--allow-market-fallback / --no-market-fallback (default enabled): Applied only when --enforce-session-close is enabled.

Backtest day boundary:

A single business date YYYY-MM-DD means YYYY-MM-DD 06:00:00 JST to next day 05:59:59.999 JST.
Overlay chart time axis is displayed in JST.

Execution assumptions:

Orders are always limit orders.
Buy limit fill condition: subsequent trade price <= limit_price after latency.
Sell limit fill condition: subsequent trade price >= limit_price after latency.
Limit price selection in trade-only mode: buy uses bid -> mid -> last, sell uses ask -> mid -> last.
New orders are throttled by min_order_interval_sec; strategy evaluation still runs on every event.
Alpha metrics are computed against fixed benchmark PnL: benchmark_hold_eth * (mid(t)-mid(t-1)).
Session-close policy:
- 05:49:00-05:59:59 JST: no new entries.
- 05:55:00-05:59:59 JST: forced close ladder (limit repricing).
- >=05:59:50 JST and fallback enabled: force market close to reach flat.

Available baseline strategies:

baseline_flat
baseline_momentum_micro
baseline_mean_reversion_micro
baseline_inventory_mm
baseline_queue_imbalance_breakout
baseline_sweep_follow
baseline_regime_switch_micro
baseline_flow_imbalance_pulse
baseline_technical_hold_overlay (10ETH hold-bias pseudo-technical overlay: de-risk on downside, rebuild on recovery)
baseline_technical_hold_overlay_slow (lower-turnover conservative variant of technical hold overlay)
ml_direction_linear (requires trained model file)
ml_direction_logistic (requires trained logistic model file)
ml_direction_logistic_swing (smoothed, lower-frequency ML strategy with cooldown/min-hold)
ml_direction_logistic_swing_tokyo (swing model restricted to Tokyo session hours)
ml_direction_logistic_swing_europe (swing model restricted to Europe session hours)
ml_direction_logistic_swing_europe_conservative (Europe session with stricter signal/cadence controls)
ml_direction_logistic_swing_europe_defensive (Europe session, trades only in local bearish micro-regime)
ml_direction_logistic_swing_europe_regime2 (Europe session tuned for stronger trend-following regime)
ml_direction_logistic_swing_europe_regime3 (Europe session tuned from expanded segmented search)
ml_direction_logistic_swing_europe_robust (Europe session + outlier guard for anomalous feature states)
ml_direction_logistic_swing_europe_ultra_conservative (Europe session with very strict low-turnover controls)
ml_direction_logistic_swing_us (swing model restricted to US session hours)
ml_direction_logistic_hold_overlay (10ETH long-bias overlay: hold by default, flatten/short only on strong downside signal)
ml_direction_logistic_hold_overlay_regime_guard (hold overlay with downside/recovery confirmation filters to reduce whipsaw)
ml_direction_logistic_hold_overlay_regime_guard_derisk_only (regime guard variant that de-risks to flat and avoids net-short)
ml_direction_logistic_hold_overlay_regime_guard_conservative (regime guard variant with smaller/stricter short hedges)

Typical output (JSON):

symbol, strategy
events, intents, fills
final_position_eth, final_equity_jpy
metrics.alpha_sharpe
metrics.alpha_max_dd
metrics.turnover, metrics.fill_rate, metrics.adverse_selection
report_path
plot_path (present when --plot is enabled)
plot_html_path / plot_png_path (present when corresponding format is generated)
benchmark_mode, benchmark_hold_eth
session_close_policy
forced_close_orders, forced_close_fills
market_fallback_count
deadline_breach_flag

Report file:

data/derived/reports/backtest_{symbol}_{start}_{end}_{strategy}.json

1.2b Chart Legend (Daily / Overview)

Daily overlay chart (--plot-format png|both):

Panel	Line / Marker	Meaning	Formula / Source	Why it matters
Price + Orders	`Price (1m)` candlestick	1-minute OHLC built from event prices (JST axis)	trade/ticker price aggregated by `group_by_dynamic`	Market context for each order decision
Price + Orders	`Buy Orders` marker	Buy-side submitted/filled order points	marker on order reference/fill price	Visual check of long entries timing
Price + Orders	`Sell Orders` marker	Sell-side submitted/filled order points	marker on order reference/fill price	Visual check of short/exit timing
Price + Orders	`Risk Rejected` marker	Orders blocked by risk rules	risk gate result	Detects if limits are suppressing strategy actions
Position	`Position ETH` line	Net position over the day	simulated execution state	Confirms exposure control and reset/carry behavior
PnL vs Benchmark	`Cum Strategy PnL` line	Cumulative strategy PnL	cumulative sum of step strategy PnL	Raw strategy monetary result
PnL vs Benchmark	`Cum Benchmark PnL` line	Cumulative fixed benchmark PnL	`benchmark_hold_eth * (mid(t)-mid(t-1))` cumulative	Baseline from passive hold (e.g. 10 ETH)
PnL vs Benchmark	`Cum Alpha PnL` line	Strategy excess PnL over benchmark	`Cum Strategy PnL - Cum Benchmark PnL`	Core metric: trading skill vs market drift

Overview chart (monthly summary PNG generated from daily CSV):

Panel	Line / Bar	Meaning	Formula / Source	Why it matters
Daily PnL	`Daily Strategy PnL` bar	Strategy PnL per business day	from each day backtest result	Day-by-day profitability distribution
Daily PnL	`Daily Benchmark PnL (10ETH Hold)` bar	Passive-hold daily PnL	fixed hold benchmark for same day	Market regime effect without strategy skill
Daily PnL	`Daily Alpha PnL` line	Daily excess over benchmark	`Daily Strategy PnL - Daily Benchmark PnL`	Direct daily skill score
Cumulative PnL	`Cum Strategy PnL` line	Running total strategy PnL	cumulative daily strategy PnL	Long-run monetary trajectory
Cumulative PnL	`Cum Benchmark PnL` line	Running total benchmark PnL	cumulative daily benchmark PnL	Drift-only baseline trajectory
Cumulative PnL	`Cum Alpha PnL` line	Running total excess PnL	cumulative daily alpha PnL	Whether edge accumulates over month
Orders/Fills	`Orders` bar	Submitted orders per day	daily backtest report	Activity/turnover intensity
Orders/Fills	`Fills` bar	Filled orders per day	daily backtest report	Execution efficiency with limit-only model

1.3 `atc replay`

Purpose:

Replay canonical events from a path.
Run the same event-driven backtest pipeline.

Command:

uv run atc replay --from data/canonical/events/symbol=ETH_JPY --strategy baseline_flat

Arguments:

--from (required): Path to canonical event root or symbol partition path.
--strategy (required)
--symbol (optional): Explicit symbol override.

Typical output:

Same JSON shape as atc backtest (except report_path is not added here).

1.4 `atc dataset`

Purpose:

Build derived feature datasets from canonical events.
Reuse cached feature results when available.
Support Rust or Python feature-generation backend.

Command:

uv run atc dataset --symbol ETH_JPY --start 2022-01-01 --end 2022-01-31 --name features_jan

Arguments:

--symbol (required)
--start (required)
--end (required)
--name (optional, default features)
--data-root (optional, default data)
--feature-backend (optional, default rust): rust or python.
--workers (optional, default 0): Rust backend worker count (0 = auto).
--no-cache (optional): force rebuild instead of cache reuse.

Typical output (JSON):

dataset_path

Dataset file:

data/derived/datasets/{name}.parquet

1.5 `atc features cache`

Purpose:

Build a reusable feature cache from canonical events.
Cache is stored as Parquet and can be reused by atc dataset.

Command:

uv run atc features cache --symbol ETH_JPY --start 2026-01-01 --end 2026-01-31 --feature-backend rust

Arguments:

--symbol (required)
--start (required)
--end (required)
--data-root (optional, default data)
--feature-backend (optional, default rust): rust or python.
--workers (optional, default 0): Rust backend worker count (0 = auto).
--no-cache (optional): force rebuild.

Typical output (JSON):

cache_path
manifest_path
backend
cache_hit
events
rows
workers

Cache file:

data/derived/features/cache/symbol={symbol}/backend={backend}/features_{start}_{end}_v1.parquet

1.6 `atc ml`

Purpose:

Train machine-learning direction model from event-derived features.

`atc ml train-direction`

uv run atc ml train-direction --symbol ETH_JPY --start 2025-01-01 --end 2025-12-31 --horizon-sec 300

Arguments:

--symbol (required)
--start (required)
--end (required)
--data-root (optional, default data)
--horizon-sec (optional, default 300)
--valid-ratio (optional, default 0.2)
--ridge-alpha (optional, default 1.0)
--model-output (optional): explicit model output path
--workers (optional, default 0): feature-generation workers (0 = auto CPU cores minus one, capped at 5)
--progress-every (optional, default 1): print progress every N chunks
--no-progress (optional): disable intermediate progress lines
--no-update-latest (optional): do not overwrite latest model alias

Outputs:

Model JSON at:
data/derived/models/ml_direction_linear_{symbol}_{start}_{end}.json
Latest alias (default):
data/derived/models/ml_direction_linear_latest.json
Metrics JSON:
{model_path}.metrics.json

`atc ml train-direction-logistic`

uv run atc ml train-direction-logistic --symbol ETH_JPY --start 2025-01-01 --end 2025-12-31 --horizon-sec 300

Arguments:

--symbol (required)
--start (required)
--end (required)
--data-root (optional, default data)
--horizon-sec (optional, default 300)
--valid-ratio (optional, default 0.2)
--learning-rate (optional, default 0.05)
--l2-alpha (optional, default 1e-4)
--epochs (optional, default 40)
--target-mode (optional, default direction): direction or hold10_oracle
--oracle-sell-threshold-bps (optional, default 25.0): for hold10_oracle, sell label threshold in bps
--model-output (optional): explicit model output path
--workers (optional, default 0): feature-generation workers (0 = auto CPU cores minus one, capped at 5)
--progress-every (optional, default 1): print progress every N chunks
--no-progress (optional): disable intermediate progress lines

Outputs:

Model JSON at:
data/derived/models/ml_direction_logistic_{symbol}_{start}_{end}.json
Metrics JSON:
{model_path}.metrics.json

`atc ml build-image-dataset`

uv run atc ml build-image-dataset --symbol ETH_JPY --start 2026-01-01 --end 2026-01-31 --output-name image_alpha

Purpose:

Build a multichannel image dataset for shadow comparison (4 x H x W).
Apply session-aware drops:
- drop anchors in no-new-entry window.
- optionally drop cross-day horizon samples.

Arguments:

--symbol (required)
--start (required)
--end (required)
--data-root (optional, default data)
--output-name (optional, default image_alpha)
--window-minutes (optional, default 30)
--image-size (optional, default 128)
--stride-sec (optional, default 30)
--horizon-sec (optional, default 1800)
--benchmark-hold-eth (optional, default 10.0)
--enforce-session-close (optional, default disabled)
--no-new-entry-minutes (optional, default 10, only when close constraints are enabled)
--drop-cross-day / --allow-cross-day (default drop enabled)

Output JSON:

dataset_path, meta_path
rows_total
dropped_cross_day
dropped_no_new_entry_window
drop_ratio

`atc ml train-image-alpha`

uv run atc ml train-image-alpha --dataset-path data/derived/datasets/image_alpha_ETH_JPY_2026-01-01_2026-01-31.npz

Purpose:

Train a lightweight image alpha regression model for shadow evaluation.

Arguments:

--dataset-path (required)
--model-output (optional): default sibling ml_image_alpha_linear_latest.json
--valid-ratio (optional, default 0.2)
--ridge-alpha (optional, default 1.0)

Output JSON:

model_path
rows_total, rows_train, rows_valid
train_mse, valid_mse
metrics_path

`atc ml shadow-compare`

uv run atc ml shadow-compare \
  --dataset-path data/derived/datasets/image_alpha_ETH_JPY_2026-01-01_2026-01-31.npz \
  --output-csv data/derived/reports/shadow_compare_ETH_JPY_2026-01.csv

Purpose:

Compare Tabular and Image expected-alpha scores on the same event sequence in shadow mode.

Arguments:

--dataset-path (required)
--output-csv (required)
--image-model-path (optional): default sibling ml_image_alpha_linear_latest.json
--tabular-model-path (optional): default logistic latest if present, otherwise linear latest
--benchmark-hold-eth (optional, default 10.0)

Output JSON:

rows
tabular_target_corr
image_target_corr
image_better_rows
tabular_better_or_equal_rows
output_csv_path

1.7 `atc report`

Purpose:

Inspect generated report JSON.
Build multi-strategy leaderboard for a fixed date window.
Run walk-forward optimization for baseline_regime_switch_micro.

`atc report show`

uv run atc report show --path data/derived/reports/backtest_ETH_JPY_2026-01-01_2026-01-01_baseline_momentum_micro.json

Arguments:

--path (required): JSON report path.

`atc report leaderboard`

uv run atc report leaderboard --symbol ETH_JPY --start 2026-01-01 --end 2026-01-07

Arguments:

--symbol (required)
--start (required)
--end (required)
--data-root (optional, default data)
--strategy (optional, repeatable): Evaluate only the specified strategy names.
--order-latency-sec (optional, default 0.3)
--min-order-interval-sec (optional, default 300)

Output:

JSON ranking sorted by alpha_sharpe descending.
Parquet snapshot at:
data/derived/reports/leaderboard_{symbol}_{start}_{end}.parquet

`atc report walk-forward`

uv run atc report walk-forward --symbol ETH_JPY --start 2026-01-01 --end 2026-01-31 --train-days 14 --valid-days 7 --step-days 7

Arguments:

--symbol (required)
--start (required)
--end (required)
--data-root (optional, default data)
--train-days (optional, default 14)
--valid-days (optional, default 7)
--step-days (optional, default 7)
--session (optional, repeatable): TOKYO, EUROPE, US, or ALL.
--max-candidates (optional, default 64)
--workers (optional, default 0): parallel workers for candidate evaluation (0 = auto CPU cores minus one).
--order-latency-sec (optional, default 0.3)
--min-order-interval-sec (optional, default 300)
--progress-every (optional, default 10): print progress every N train evaluations.
--no-progress (optional): disable intermediate progress lines.

Output:

JSON summary with train/valid split results and best params per split.
JSON file at:
data/derived/reports/walk_forward_regime_switch_{symbol}_{start}_{end}.json

`atc report daily-eval`

uv run atc report daily-eval --symbol ETH_JPY --start 2026-01-01 --end 2026-01-31 --strategy baseline_regime_switch_micro

Purpose:

Run day-by-day backtests with per-day reset (start_position=0, start_cash=0).
Export daily metrics CSV, overall summary CSV, and overview PNG.
Enforce fixed benchmark comparison (10 ETH hold default).

Arguments:

--symbol (required)
--start (required)
--end (required)
--strategy (required)
--data-root (optional, default data)
--benchmark-hold-eth (optional, default 10.0)
--target-abs-qty-eth (optional, default 10.0)
--max-abs-position-eth (optional, default 10.0)
--order-latency-sec (optional, default 0.3)
--min-order-interval-sec (optional, default 300)
--enforce-session-close (optional, default disabled)
--no-new-entry-minutes (optional, default 10, only when close constraints are enabled)
--force-close-start-jst (optional, default 05:55:00, only when close constraints are enabled)
--close-deadline-jst (optional, default 05:59:50, only when close constraints are enabled)
--allow-market-fallback / --no-market-fallback (default enabled, only when close constraints are enabled)
--plot-interval (optional, default 1m)
--plot-format (optional, default png): png, html, or both for each daily overlay.
--plot / --no-plot (default plot enabled): skip per-day overlay generation for faster iteration.

Generated outputs:

Daily CSV:
- strategy_total_pnl_jpy
- benchmark_total_pnl_jpy
- alpha_total_pnl_jpy
- orders, fills, fill_rate, turnover
- forced_close_orders, forced_close_fills
- market_fallback_count
- deadline_breach_flag
Overall CSV:
- totals, means, daily win-rate, max drawdown
Overview PNG:
- daily Strategy/Benchmark/Alpha
- cumulative Strategy/Benchmark/Alpha
- orders/fills

2) `just` wrapper commands

just provides convenient wrappers around the atc CLI.

2.1 `just sync`

just sync

Runs uv sync --extra dev.

2.2 `just test`

just test

Runs uv run pytest -q.

2.3 `just lint`

just lint

Runs uv run ruff check src tests.

2.4 `just typecheck`

just typecheck

Runs uv run mypy src.

2.5 `just ingest symbol start end`

just ingest ETH_JPY 2026-01-01 2026-01-31

Equivalent to:

uv run atc ingest historical --symbol ETH_JPY --start 2026-01-01 --end 2026-01-31

2.6 `just backtest symbol start end strategy`

just backtest ETH_JPY 2022-01-01 2022-12-31 baseline_momentum_micro

Equivalent to:

uv run atc backtest --symbol ETH_JPY --start 2022-01-01 --end 2022-12-31 --strategy baseline_momentum_micro

2.7 `just features-cache symbol start end [backend]`

just features-cache ETH_JPY 2026-01-01 2026-01-31 rust

Equivalent to:

uv run atc features cache --symbol ETH_JPY --start 2026-01-01 --end 2026-01-31 --feature-backend rust

2.6b `just backtest-default symbol strategy`

just backtest-default ETH_JPY baseline_momentum_micro

Equivalent to:

uv run atc backtest --symbol ETH_JPY --strategy baseline_momentum_micro

2.7 `just replay path strategy`

just replay data/canonical/events/symbol=ETH_JPY baseline_flat

Equivalent to:

uv run atc replay --from data/canonical/events/symbol=ETH_JPY --strategy baseline_flat

2.7b `just backtest-plot symbol strategy`

just backtest-plot ETH_JPY baseline_momentum_micro

Equivalent to:

uv run atc backtest --symbol ETH_JPY --strategy baseline_momentum_micro --plot --plot-interval 1m

2.8 `just compare-baselines symbol start end`

just compare-baselines ETH_JPY 2022-01-01 2022-12-31

複数 baseline 戦略を同条件で順番に backtest します。
実行対象:
baseline_flat
baseline_momentum_micro
baseline_mean_reversion_micro
baseline_inventory_mm
baseline_queue_imbalance_breakout
baseline_sweep_follow
baseline_regime_switch_micro
baseline_flow_imbalance_pulse

2.9 `just leaderboard symbol start end`

just leaderboard ETH_JPY 2026-01-01 2026-01-07

Equivalent to:

uv run atc report leaderboard --symbol ETH_JPY --start 2026-01-01 --end 2026-01-07

Output:

Ranked strategy table in JSON (sorted by alpha_sharpe descending).
Parquet snapshot at:
data/derived/reports/leaderboard_{symbol}_{start}_{end}.parquet

2.10 `just walk-forward symbol start end`

just walk-forward ETH_JPY 2026-01-01 2026-01-31

Equivalent to:

uv run atc report walk-forward --symbol ETH_JPY --start 2026-01-01 --end 2026-01-31

2.11 `just ml-train-direction symbol start end`

just ml-train-direction ETH_JPY 2025-01-01 2025-12-31

Equivalent to:

uv run atc ml train-direction --symbol ETH_JPY --start 2025-01-01 --end 2025-12-31

2.12 `just ml-train-direction-logistic symbol start end`

just ml-train-direction-logistic ETH_JPY 2025-01-01 2025-12-31

Equivalent to:

uv run atc ml train-direction-logistic --symbol ETH_JPY --start 2025-01-01 --end 2025-12-31

3) Built-in help

To confirm current options from source of truth:

uv run atc --help
uv run atc ingest historical --help
uv run atc backtest --help
uv run atc replay --help
uv run atc dataset --help
uv run atc ml --help
uv run atc ml train-direction --help
uv run atc ml train-direction-logistic --help
uv run atc ml build-image-dataset --help
uv run atc ml train-image-alpha --help
uv run atc ml shadow-compare --help
uv run atc report --help
uv run atc report leaderboard --help
uv run atc report walk-forward --help
uv run atc report daily-eval --help
just --list

4) Website and report article workflow

4.1 Install and run docs website

cd website
pnpm install
pnpm start

Build static site:

cd website
pnpm build

4.2 Convert backtest report JSON to website article

Command:

uv run python scripts/report_to_article.py \
  --report data/derived/reports/backtest_ETH_JPY_2025-10-01_2026-02-11_ml_direction_logistic_hold_overlay_regime_guard.json

Useful options:

--output-dir (default: website/blog)
--date (default: today, YYYY-MM-DD)
--decision (default: 保留)
--note (optional free text)
--overwrite (replace existing markdown)

Generated file:

website/blog/{YYYY-MM-DD}-{symbol}-{strategy}.md

Prerequisites​

1) atc CLI commands​

1.1 atc ingest historical​

1.2 atc backtest​

1.2b Chart Legend (Daily / Overview)​

1.3 atc replay​

1.4 atc dataset​

1.5 atc features cache​

1.6 atc ml​

atc ml train-direction​

atc ml train-direction-logistic​

atc ml build-image-dataset​

atc ml train-image-alpha​

atc ml shadow-compare​

1.7 atc report​

atc report show​

atc report leaderboard​

atc report walk-forward​

atc report daily-eval​

2) just wrapper commands​

2.1 just sync​

2.2 just test​

2.3 just lint​

2.4 just typecheck​

2.5 just ingest symbol start end​

2.6 just backtest symbol start end strategy​

2.7 just features-cache symbol start end [backend]​

2.6b just backtest-default symbol strategy​

2.7 just replay path strategy​

2.7b just backtest-plot symbol strategy​

2.8 just compare-baselines symbol start end​

2.9 just leaderboard symbol start end​

2.10 just walk-forward symbol start end​

2.11 just ml-train-direction symbol start end​

2.12 just ml-train-direction-logistic symbol start end​

3) Built-in help​

4) Website and report article workflow​

4.1 Install and run docs website​

4.2 Convert backtest report JSON to website article​

Prerequisites

1) `atc` CLI commands

1.1 `atc ingest historical`

1.2 `atc backtest`

1.2b Chart Legend (Daily / Overview)

1.3 `atc replay`

1.4 `atc dataset`

1.5 `atc features cache`

1.6 `atc ml`

`atc ml train-direction`

`atc ml train-direction-logistic`

`atc ml build-image-dataset`

`atc ml train-image-alpha`

`atc ml shadow-compare`

1.7 `atc report`

`atc report show`

`atc report leaderboard`

`atc report walk-forward`

`atc report daily-eval`

2) `just` wrapper commands

2.1 `just sync`

2.2 `just test`

2.3 `just lint`

2.4 `just typecheck`

2.5 `just ingest symbol start end`

2.6 `just backtest symbol start end strategy`

2.7 `just features-cache symbol start end [backend]`

2.6b `just backtest-default symbol strategy`

2.7 `just replay path strategy`

2.7b `just backtest-plot symbol strategy`

2.8 `just compare-baselines symbol start end`

2.9 `just leaderboard symbol start end`

2.10 `just walk-forward symbol start end`

2.11 `just ml-train-direction symbol start end`

2.12 `just ml-train-direction-logistic symbol start end`

3) Built-in help

4) Website and report article workflow

4.1 Install and run docs website

4.2 Convert backtest report JSON to website article