{"paper":{"title":"Sequential Structure in Intraday Futures Data: LSTM vs Gradient Boosting on MNQ","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Four years of single-instrument five-minute OHLCV data prove insufficient for reliable intraday ML forecasting.","cross_cats":["cs.LG","q-fin.CP","q-fin.ST"],"primary_cat":"q-fin.TR","authors_text":"Mathias Mesfin","submitted_at":"2026-05-18T01:03:28Z","abstract_excerpt":"This paper compares gradient boosting and long short-term memory (LSTM) architectures for intraday directional prediction in Micro E-Mini Nasdaq 100 futures (MNQ). Motivated by recent foundation-model research on financial candlestick data, including the Kronos architecture, we test whether five-minute OHLCV bar sequences contain exploitable sequential predictive structure at the scale of a single instrument dataset. Using 944 trading days from 2021-2025, four model configurations are evaluated under strict expanding-window walk-forward validation across three out-of-sample periods. The target"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"The results indicate that four years of single-instrument five-minute OHLCV data are insufficient for reliable sequential ML-based intraday forecasting.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the chosen binary target (close > 10:30 AM open by more than ten points) and the five-minute OHLCV representation are sufficient to reveal any exploitable sequential structure if such structure exists in the market.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Neither LSTM nor gradient boosting models achieve statistically significant out-of-sample accuracy above the 51.8% base rate for intraday MNQ directional prediction using 944 trading days of five-minute OHLCV data under walk-forward validation.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Four years of single-instrument five-minute OHLCV data prove insufficient for reliable intraday ML forecasting.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"aed3acdaf1ac810a9ea7d98d1fab4a911a320578adfecdedb3cb099bb61c3ae6"},"source":{"id":"2605.17724","kind":"arxiv","version":1},"verdict":{"id":"eccea29e-edf0-4c7b-912d-acb6ec144de3","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T22:00:29.571833Z","strongest_claim":"The results indicate that four years of single-instrument five-minute OHLCV data are insufficient for reliable sequential ML-based intraday forecasting.","one_line_summary":"Neither LSTM nor gradient boosting models achieve statistically significant out-of-sample accuracy above the 51.8% base rate for intraday MNQ directional prediction using 944 trading days of five-minute OHLCV data under walk-forward validation.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the chosen binary target (close > 10:30 AM open by more than ten points) and the five-minute OHLCV representation are sufficient to reveal any exploitable sequential structure if such structure exists in the market.","pith_extraction_headline":"Four years of single-instrument five-minute OHLCV data prove insufficient for reliable intraday ML forecasting."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.17724/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_title_agreement","ran_at":"2026-05-19T22:31:19.363782Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T22:10:53.320949Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"shingle_duplication","ran_at":"2026-05-19T21:49:43.464237Z","status":"skipped","version":"0.1.0","findings_count":0},{"name":"citation_quote_validity","ran_at":"2026-05-19T21:49:43.293434Z","status":"skipped","version":"0.1.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T21:33:23.497558Z","status":"skipped","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T21:21:57.398811Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"cited_work_retraction","ran_at":"2026-05-19T21:21:57.110120Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"005f4fc39f17c20f3d4f9595eb202cb8995166723fa14d0d73794faab73b919c"},"references":{"count":8,"sample":[{"doi":"","year":2025,"title":"Introduction The publication of foundation models for financial time series data represents a meaningful development in quantitative research. The Kronos model (Shi et al., 2025), trained on millions ","work_id":"4c1a097d-1d17-48ff-b508-c2e8f751eb85","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2021,"title":"After session boundary filtering and removal of partial days, 947 complete trading days remain for daily feature construction","work_id":"1592a258-ce4c-4275-bc1d-bc79101fb2fa","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2026,"title":"Target A (daily close vs","work_id":"b886a6c5-7175-4cf2-99b9-46571c6b805a","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2026,"title":"Model Architectures and Validation Methodology 4.1 Gradient Boosting Classifier Three gradient boosting configurations are evaluated, each representing a different feature set and target specification","work_id":"b2f96e01-9b46-497a-bbbd-ceb08f09bf1d","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2022,"title":"This is the most conservative test—it does not attempt to use intraday structure and instead relies on multi-day patterns in returns, gaps, and volatility","work_id":"d2ecfea5-e666-4af1-8cce-8c54e7fcaa64","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":8,"snapshot_sha256":"e8b940a13a5717091dc9bc5773caa53286ee49b66a25343ddb81f1e1ab97e291","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"1a48c57f87e438493b7d9deee420640aaa5908841a01a8e67c8b63fa8cbb1fce"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}