{"paper_id":"2512.12924","title":"Interpretable Hypothesis-Driven Trading: A Rigorous Walk-Forward Validation Framework for Market Microstructure Signals","authors":["Gagan Deep","Akash Deep","William Lamptey"],"paper_date":"2025-12-15","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2512.12924v1","page_url":"https://thequant.space/flowcharts/interpretable-hypothesis-driven-tradinga-rigorous-walk-forw/","scores":{"empirical_rigor":6,"math_complexity":7,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes a rigorous walk‑forward validation framework for algorithmic trading that combines interpretable hypothesis‑driven signal generation with reinforcement learning and strict out‑of‑sample testing.","quote":"We develop a rigorous walk-forward validation framework for algorithmic trading designed to mitigate overfitting and lookahead bias.","summary":"We develop a rigorous walk-forward validation framework for algorithmic trading designed to mitigate overfitting and lookahead bias. Our methodology combines interpretable hypothesis-driven signal generation with reinforcement learning and strict out-of-sample testing. The framework enforces strict…","strategy":"hypothesis-driven microstructure signal framework"},"assumptions":{"status":"available","items":[{"assumption":"Features, signals, and execution decisions use only data available up to that point in time (strict information set discipline).","type":"data","stated":"explicit","quote":"First, it enforces strict information set discipline where features, signals, and execution decisions use only data available up to that point in time, preventing lookahead bias."},{"assumption":"Walk‑forward validation is performed with rolling windows across 34 independent out‑of‑sample test periods.","type":"model","stated":"explicit","quote":"Second, it employs walk-forward validation with rolling windows, where the system must prove itself repeatedly across 34 independent out-of-sample test periods spanning multiple market regimes."},{"assumption":"All trades must originate from a human‑interpretable hypothesis expressed in natural language.","type":"other","stated":"explicit","quote":"Third, it maintains complete interpretability by requiring every trade to originate from a human-interpretable hypothesis expressed in natural language."},{"assumption":"Transaction costs consist of a $1 fixed commission and 5 basis‑points slippage per share.","type":"execution","stated":"explicit","quote":"where cfixed = $1 commission and cslippage = 0.0005 (5 basis points)."},{"assumption":"Positions are limited to a maximum of 5 concurrent holdings, 20% allocation per position, and 50% per sector.","type":"execution","stated":"explicit","quote":"Position Constraints: Maximum 5 concurrent positions, maximum 20% allocation per position, maximum 50% allocation per sector."},{"assumption":"The sample excludes delisted or acquired stocks, introducing survivorship bias.","type":"data","stated":"explicit","quote":"This process introduces survivorship bias (stocks delisted/acquired during 2015-2024 are excluded), which biases results upward."}]},"datasets":{"status":"available","items":[{"name":"Yahoo Finance daily price data","provider":"Yahoo Finance","kind":"market-prices","access":"public","period":"2015-01-02 to 2024-10-31","universe":"100 US equities","quote":"All data obtained from Yahoo Finance via yfinance API with standard adjustments for splits and dividends.","known_id":"yahoo-finance"}],"known":[{"id":"yahoo-finance","name":"Yahoo Finance","provider":"Yahoo","kind":"market-prices","access":"public","quote":"All data obtained from Yahoo Finance via yfinance API with standard adjustments for splits and dividends.","mentions":3,"used_in":1}],"sample":{"start":"2015","end":"2024","market":"US equities (100 US stocks)","asset_class":"equities","frequency":"daily"},"data_resolution":"daily"},"transaction_costs":{"status":"available","modeled":"yes","headline_net_of_costs":true,"detail":"Fixed $1 commission plus 5 basis‑points slippage per share; orders executed at next‑day open.","quote":"where cfixed = $1 commission and cslippage = 0.0005 (5 basis points). Orders placed at day t close execute at day t + 1 open with slippage.","source":"front_matter","rule":"R3","candidates":{"front_matter":"yes","evidence_card":"yes","evidence_card_quoted":true},"card_detail":"Fixed $1 commission plus 5 basis‑points slippage per share; orders executed at next‑day open.","card_quote":"where cfixed = $1 commission and cslippage = 0.0005 (5 basis points). Orders placed at day t close execute at day t + 1 open with slippage.","meaning":"costs are part of the reported headline results"},"results":{"status":"available","evaluation_type":"trading-pnl","findings":[{"finding":"The framework yields modest annualized returns of 0.55% with a Sharpe ratio of 0.33 and maximum drawdown of -2.76%, while beta to the market is only 0.058.","kind":"performance","quote":"The system yields modest annualized returns (0.55%, Sharpe ratio 0.33) with exceptional downside protection (maximum drawdown -2.76%) and market‑neutral characteristics (beta = 0.058)."},{"finding":"Aggregate results are not statistically significant (p‑value 0.34), indicating no reliable profitability across the 34 folds.","kind":"statistical","quote":"The null hypothesis H0 : µ = 0 cannot be rejected at conventional significance levels: t‑statistic = 0.96, p‑value = 0.34 (two‑sided)."},{"finding":"Performance is regime‑dependent: positive quarterly returns of 0.60% during high‑volatility 2020‑2024 period versus a negative -0.16% in stable 2015‑2019 markets.","kind":"performance","quote":"The key empirical finding reveals that daily OHLCV‑based microstructure signals require elevated information arrival and trading activity to function effectively."},{"finding":"Win rate at the fold level is 41% (14 of 34 folds positive) and trade‑level win rate is 46.5% across 140 trades.","kind":"performance","quote":"Trade-level win rate across all folds is 46.5%, with 140 total trades executed."},{"finding":"The framework is agnostic to hypothesis source and can incorporate rule‑based, genetic programming, neural network, or large‑language‑model generated hypotheses.","kind":"theoretical","quote":"The framework accepts hypotheses from any generator G"}],"reported_performance":{"strategy":"hypothesis-driven microstructure signal framework","sharpe":0.33,"sharpe_benchmark":0.86,"annual_return_pct":0.55,"total_return_pct":null,"max_drawdown_pct":2.76},"out_of_sample":{"method":"walk-forward","headline_out_of_sample":true}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":34,"citations":{"cited_by_in_corpus":2,"by_independent_authors":2,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":1,"unclassified":1},"independent_papers":[],"scope":"Citations between papers in this corpus only; citations from outside it are not counted."}},"implementation":{"status":"available","public":true,"code_url":"https://github.com/akashdeepo/Interpretable-Hypothesis-Driven-Trading","note":null},"replication":{"status":"none_found","card":null,"note":"No replication of this paper has been run by thequant.space yet."},"extraction":{"text_source":"fulltext","model":"openai/gpt-oss-120b","version":1,"extracted_at":"2026-11-14T17:19:05Z"},"provenance":{"claims":"Everything here is what the authors state in the paper. Nothing was re-run or reproduced by the platform.","quotes":"Assumptions, datasets, cost detail and findings were extracted by a language model. Every quote is verbatim paper text. Where an item has `support`, \"full\" means every figure in it appears in the quoted sentences and \"partial\" means a figure is stated elsewhere in the paper; items with a figure found nowhere in the paper were discarded.","datasets_known":"datasets.known lists well-known vendors and public datasets by a stable id, found by a dictionary over the full text where the sentence says the data is used; it is independent of the model and covers only names in the dictionary.","flags":"transaction_costs.modeled, results.evaluation_type and results.out_of_sample.method are platform classifications (structured-decision models); see https://thequant.space/score-guide/#penalty-flags","comparisons":"supports / contradicts / qualifies are a language model's reading of two papers' quoted findings, restricted to closely related papers that share no author. They are pointers to read, not verified replications, and coverage is partial.","implementation":"code_url is a repository the paper itself presents as its own (rule-scored, ambiguous cases judged by a model), checked for liveness when it was recorded."},"docs":"https://thequant.space/agents/"}