Rules make the strategy hypothesis explicit

The Python project evaluates ordinary Polars expressions. The Compiled project declares a validated condition graph in YAML. The worked cases use the same hourly assessment interval and two price floors, allowing their Boolean decisions to be checked independently.

Prerequisites

Both projects require the matching wheel and recorded-input bundle. The rule cases use the data extra. The observation clock, feature history, and intended execution policy must be established before interpreting a position as a trade.

Predicate composition defines the entry

The worked hypothesis requires a close above a declared price floor, recent or persistent upward activity, and a rise over three observations. The current bar must also close at or above its open. The comparison varies the floor between 40,000 USDT and 41,000 USDT per BTC.

These predicates evaluate completed hourly observations. Their signals become available after the current close. The execution policy determines the admissible later fill.

Python expresses the predicates as functions

The Python case computes the Boolean conditions on the declared assessment partition. Rolling history starts at that partition's beginning, and early unavailable comparisons become false:

1def evaluate_rules(raw):2    bars = raw.filter(pl.col('datetime').is_between(3        pl.datetime(2024, 1, 18, time_zone='UTC'), pl.datetime(2024, 1, 22, time_zone='UTC'), closed='left'))4    up = pl.col('close') > pl.col('open')5    cross = up & (pl.col('close').shift(1) <= pl.col('open').shift(1))6    persistent = (up.cast(pl.Int8).rolling_sum(2, min_samples=2) == 2).fill_null(False)7    recent = (cross.fill_null(False).cast(pl.Int8).rolling_sum(3, min_samples=1) >= 1)8    rising = (pl.col('close') > pl.col('close').shift(3)).fill_null(False)9    rows = []10    for floor in (40000., 41000.):11        signal = bars.select(((pl.col('close') >= floor) & (persistent | recent) & rising &12                              ~(pl.col('close') < pl.col('open'))).cast(pl.Int8).alias('signal'))['signal']13        rows.append({'price_floor': floor, 'signals': signal.to_list(), 'positive_decisions': int(signal.sum())})14    return {'rows': rows, 'assessment_rows': bars.height, 'history_resets_at_split': True}

The returned signal tape and positive-decision count describe the rule's outputs. A trading comparison additionally needs entry, exit, price, and cost conventions. The project's separate baseline campaign applies its explicit next-open-to-close policy.

YAML names the graph and its parameter domain

The Compiled definition assigns each condition an identity and combines them through named operands. Validation checks references, duplicate identities, dependencies, and cycles before execution:

1schema_version: '1.0'2metadata:3  name: book_rule_contracts4  mode: development5sfd:6  manifest:7    type: rule_based8    data_source:9      method: limen.data.HistoricalData.get_any_file10      params:11        file_path_or_url: inputs/spot_1h_20240101_20241231.parquet12    split_dates:13      train_start: '2024-01-01'14      train_end: '2024-01-15'15      val_start: '2024-01-15'16      val_end: '2024-01-18'17      test_start: '2024-01-18'18      test_end: '2024-01-22'19    strategy:20      conditions:21      - {id: floor, name: Close above the declared floor, type: threshold, column: close, operator: '>=', value: '{price_floor}'}22      - {id: persistent_up, name: Two up bars, type: relative, column: close, operator: '>', other_column: open, persistence_n: 2}23      - {id: recent_cross, name: Recent close-open cross, type: crossover, column: close, other_column: open, direction: above, recency_n: 3}24      - {id: rising, name: Close above three bars ago, type: slope, column: close, direction: rising, lookback: 3}25      - {id: down, name: Current down bar, type: sql_expr, expr: 'close < open'}26      - {id: not_down, name: Exclude down bars, operator: not, operands: [down]}27      - {id: activity, name: Persistent or recent activity, operator: or, operands: [persistent_up, recent_cross]}28      - {id: entry, name: Combine the entry conditions, operator: and, operands: [floor, activity, rising, not_down]}29      entry: entry30    reference_architecture: limen.sfd.reference_architecture.rule_based31    backtest:32      fee_bps: 7.033      slip_bps: 3.034  params:35    price_floor: [40000.0, 41000.0]36uel:37  output_path: compiled-rule-contracts38  n_permutations: 239  search_strategy: {type: grid}40  output_format: csv41  recording:42    predictions: true43    alignment: true44    trade_ledger: true45    fold_details: true

The sql_expr leaf supplies the current down-bar comparison. SQL expressions have their own information requirements. An expression using a future row or a whole-window aggregate changes its availability even when its syntax is valid. Manifest rule contracts define the supported operations and their temporal modifiers.

The signal comparison has a declared population

Both cases assess the half-open UTC interval from 18 January 2024 to 22 January 2024. The independent assertions reconstruct the predicates from recorded prices and compare the complete signal tapes. Their counts therefore measure the same rule conditions and split-local history:

The Compiled run also records its configured backtest. The Python predicate diagnostic reports decisions without attaching a return. The trading chapter develops how a signal becomes a costed position.

Entry and exit can express separate hypotheses

Independent exit rules have their own condition namespace and entry reference. They support the admitted threshold, relative, crossover, slope, and compound predicates. Their contract applies to binary entries from either a rule strategy or an ML model. Exit SQL is excluded by validation.

A separate exit can close a position while its entry signal remains true. Rearming then determines when a new entry becomes eligible. The retained definition must include both policies when research is reconstructed or compared. The independent-exit contract owns the exact supported paths and replay requirements.

The eight packaged common strategies supply explicit baselines: buy-and-hold, SMA, EMA, MACD, RSI, Bollinger, Donchian, and time-series momentum. Their reference specifies parameters, stateful entry/exit rules, costs, and recorded results. The dollar-bar crash-reversal strategy has separate source and availability requirements.

Selected rules preserve their history requirements

A reconstructed rule Sensor retains the chosen predicate configuration and preparation. Rolling conditions depend on the supplied history window. Warm-up, missing features, and selection guards remain visible in inference results. The Trainer reference describes same-window reconstruction and the boundary between research labels and inference inputs.

Continue to define the experiment with the entry hypothesis, parameter domain, and evaluation policy fixed. The subsequent search compares alternatives under those declarations.