jgrusewski 4817bd4e7c test: FinBERT news-sentiment (real LLM, real FNSPID data) -> dead for retail ...

User wanted to download a small open LLM and test news. Did it: FinBERT scoring 10408 FNSPID
headlines (7 tickers, 1332 obs). corr(sentiment, same-day ret) +0.082 (weak, contemporaneous);
next-day +0.026 (~0); next-day trade net 10bp OOS Sharpe +0.05 (dead). News real but reaction
INSTANT (HFT-captured); no tradeable next-day drift for retail; net of cost = nothing. Matches
Lopez-Lira/Tang + FinGPT research. Model was never the constraint -- market structure is. Closes
the news/LLM thread; deployable answer remains the adaptive multistrat book (news = risk-managed).
(note: pin transformers<5 for torch 2.6.)

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

2026-06-07 21:13:54 +02:00

..

cross_exchange_funding.py

feat(surfer): cross-exchange funding frontier (a) — fails (carry trap re-found)

2026-06-06 19:55:55 +02:00

cross_venue_funding.py

feat(crypto): cross-venue harness -> hysteresis (deployable) + deep full-year history

2026-06-07 18:58:24 +02:00

crypto_funding_harvest.py

feat(crypto): delta-neutral funding harvest — the one edge that breaks 0.7

2026-06-07 01:16:46 +02:00

crypto_funding_liveness.py

feat(crypto): funding carry liveness — ALIVE (regime not decay), regime filter rescues 2025-26

2026-06-07 01:20:04 +02:00

crypto_funding_oos.py

spec(crypto): funding harvest — clean OOS validated + deployable spec

2026-06-07 01:25:23 +02:00

crypto_funding_paper.py

fix(crypto): MANDATORY hedgeability filter — funding edge was partly phantom

2026-06-07 11:03:41 +02:00

crypto_sweep.py

fix(surfer): universe-robustness exposes carry as cherry-picked; momentum is robust

2026-06-06 15:16:56 +02:00

energy_battery_gate.py

feat(energy): repurpose the engine — battery arbitrage is real, growing, forecasting-shaped

2026-06-06 22:58:54 +02:00

energy_capture_gate2.py

result(crypto): cross-venue funding arb PASSES clean OOS (the breakthrough)

2026-06-07 19:08:08 +02:00

energy_capture_gate.py

result(crypto): cross-venue funding arb PASSES clean OOS (the breakthrough)

2026-06-07 19:08:08 +02:00

energy_intraday_gate.py

feat(harvest): premium-harvesting tested — simple 60/40 beats the sophisticated system

2026-06-07 00:42:32 +02:00

equity_cohort_challenge.py

feat(surfer): the challenge — liquid high-vol cohort fits (low-vol/lottery-aversion)

2026-06-06 20:40:58 +02:00

equity_factor_gate.py

feat(surfer): small/mid-cap equity factor gate — no factor survives small-cap cost

2026-06-06 20:24:43 +02:00

equity_lowvol_detail.py

feat(surfer): +0.74 autopsy — real anomaly, un-harvestable for our constraints

2026-06-06 21:21:39 +02:00

equity_lowvol_gauntlet.py

feat(surfer): equity low-vol gauntlet — real structure, marginal after realistic cost

2026-06-06 20:55:17 +02:00

equity_vrp2.py

feat(surfer): equity-index VRP frontier — real+affordable, measurement needs engineering

2026-06-06 22:13:34 +02:00

equity_vrp3.py

feat(surfer): equity-index VRP frontier — real+affordable, measurement needs engineering

2026-06-06 22:13:34 +02:00

equity_vrp.py

feat(surfer): equity-index VRP frontier — real+affordable, measurement needs engineering

2026-06-06 22:13:34 +02:00

fetch_caiso.py

feat(harvest): premium-harvesting tested — simple 60/40 beats the sophisticated system

2026-06-07 00:42:32 +02:00

fetch_crypto_pit.py

feat(surfer): point-in-time survivorship test PASSED — momentum is real

2026-06-06 16:04:19 +02:00

fetch_crypto.py

result(crypto): cross-venue funding arb PASSES clean OOS (the breakthrough)

2026-06-07 19:08:08 +02:00

fetch_daily.py

…

fetch_earnings.py

research: AI4Finance debunked + PEAD dies OOS (efficient-market wall confirmed again)

2026-06-07 17:15:22 +02:00

fetch_equities.py

feat(surfer): small/mid-cap equity factor gate — no factor survives small-cap cost

2026-06-06 20:24:43 +02:00

fetch_es_1m.py

fix(surfer): full-16y MFT settle — regime-adaptive surfer significantly negative

2026-06-06 12:15:37 +02:00

fetch_more_futures.py

feat(surfer): multi-asset futures test — uncorrelated but no edge (dilutes)

2026-06-06 17:08:40 +02:00

fetch_xsp.py

feat(surfer): equity-index VRP frontier — real+affordable, measurement needs engineering

2026-06-06 22:13:34 +02:00

fetch_xvenue_hist2.py

feat(crypto): cross-venue harness -> hysteresis (deployable) + deep full-year history

2026-06-07 18:58:24 +02:00

fetch_xvenue_hist.py

research(crypto): cross-venue funding arb survives net-of-cost with hysteresis

2026-06-07 18:33:28 +02:00

floor_experiment.py

…

fng_gate.py

feat(surfer): Fear&Greed gate — sentiment regime-dependent, not robust

2026-06-06 19:35:37 +02:00

mft_quality_test.py

fix(surfer): full-16y MFT settle — regime-adaptive surfer significantly negative

2026-06-06 12:15:37 +02:00

micro_gate.py

feat(surfer): minute-horizon gate — deep ML (Mamba2/CfC/TLOB) has no fuel

2026-06-06 19:09:04 +02:00

momentum_robustness.py

feat(surfer): momentum robustness battery — crypto XS momentum is a robust real edge

2026-06-06 15:18:59 +02:00

multiasset_futures.py

feat(surfer): multi-asset futures test — uncorrelated but no edge (dilutes)

2026-06-06 17:08:40 +02:00

multisignal_combine.py

feat(surfer): ML multi-signal combination — definitive (market not tools)

2026-06-06 22:44:43 +02:00

multisignal_walkforward.py

feat(surfer): ML multi-signal combination — definitive (market not tools)

2026-06-06 22:44:43 +02:00

multistrat_book_v2.py

result: hedge-fund reframe -> moat is cheap leverage, not the strategy

2026-06-07 19:48:46 +02:00

multistrat_book_v3.py

result: hedge-fund reframe -> moat is cheap leverage, not the strategy

2026-06-07 19:48:46 +02:00

multistrat_book.py

result: hedge-fund reframe -> moat is cheap leverage, not the strategy

2026-06-07 19:48:46 +02:00

multistrat_etf_backtest.py

backtest: deployable ETF book Sharpe +1.20, maxDD -5.4%, every year positive (beats 60/40)

2026-06-07 20:28:37 +02:00

multistrat_etf_expanded.py

test: diversification maxed in base-6; more streams dilute (they're beta-repackaging)

2026-06-07 20:50:27 +02:00

multistrat_paper.py

feat: adaptive multi-strat paper-forward harness (spec Phase 1)

2026-06-07 20:02:51 +02:00

multistrat_vol_test.py

test: ML vol-nowcaster (Mamba2/CfC/TLOB proxy) doesn't help risk layer

2026-06-07 20:14:25 +02:00

multistrat.py

result: hedge-fund reframe -> moat is cheap leverage, not the strategy

2026-06-07 19:48:46 +02:00

news_sentiment_test.py

test: FinBERT news-sentiment (real LLM, real FNSPID data) -> dead for retail

2026-06-07 21:13:54 +02:00

pead_gate.py

research: AI4Finance debunked + PEAD dies OOS (efficient-market wall confirmed again)

2026-06-07 17:15:22 +02:00

pead_real.py

research: AI4Finance debunked + PEAD dies OOS (efficient-market wall confirmed again)

2026-06-07 17:15:22 +02:00

phase1_signals.py

feat(surfer): Phase 1 RL-gate — no rich signal set; RL not justified (+ resid-mom upgrade)

2026-06-06 18:36:21 +02:00

pit_realism.py

feat(surfer): execution-realism stress — honest momentum magnitude ~0.4-0.6

2026-06-06 16:07:31 +02:00

pit_sweep.py

feat(surfer): wire tail-managed VRP into PoC as second sleeve (two-sleeve book)

2026-06-06 17:29:40 +02:00

portfolio_harvest2.py

feat(harvest): premium-harvesting tested — simple 60/40 beats the sophisticated system

2026-06-07 00:42:32 +02:00

portfolio_harvest3.py

feat(harvest): Sharpe-ceiling question answered — triangulated ~0.7-0.9 for retail

2026-06-07 01:00:27 +02:00

portfolio_harvest.py

feat(harvest): premium-harvesting tested — simple 60/40 beats the sophisticated system

2026-06-07 00:42:32 +02:00

reflexivity_test.py

feat(surfer): Reflexivity Harvester design — killed by marginal-alpha gate

2026-06-06 19:43:42 +02:00

research_ab.py

feat(surfer): proper (a)robust-construction + (b)diversifier research

2026-06-06 16:12:55 +02:00

research_book.py

feat(surfer): capstone — low-vol fails battery; momentum is small-capacity

2026-06-06 16:17:33 +02:00

research_capacity.py

feat(surfer): turnover optimization resolves capacity — deployable momentum spec

2026-06-06 16:20:12 +02:00

research_diversifier.py

feat(surfer): diversifier hunt — no robust partner (TS-trend real but stale)

2026-06-06 16:24:55 +02:00

signal_sweep.py

feat(surfer): deflated signal-sweep engine + Batch 1 (futures factor zoo = null)

2026-06-06 14:54:06 +02:00

sixtyforty_paper.py

feat: 60/40 paper-forward harness (the ~0.7-Sharpe non-crypto baseline)

2026-06-07 15:55:44 +02:00

surfer_poc.py

feat(surfer): residual (beta-stripped) momentum upgrade — net +0.40->+0.57 @$20M

2026-06-06 18:41:40 +02:00

vrp_diversifier.py

feat(surfer): crypto VRP diversifier check — first real diversifier, but tail+decay

2026-06-06 17:13:06 +02:00

vrp_tail.py

feat(surfer): tail-managed VRP — real second sleeve (with honest haircut)

2026-06-06 17:17:09 +02:00

xcorr_leadlag.py

test: cross-market connectedness = diversification, not alpha

2026-06-07 20:46:38 +02:00

xvenue_sim2.py

research(crypto): cross-venue funding arb survives net-of-cost with hysteresis

2026-06-07 18:33:28 +02:00

xvenue_sim3.py

research(crypto): cross-venue funding arb survives net-of-cost with hysteresis

2026-06-07 18:33:28 +02:00

xvenue_sim_oos2.py

correction(crypto): cross-venue arb is MARGINAL not breakthrough (honest fixed-pair OOS)

2026-06-07 19:10:13 +02:00

xvenue_sim_oos.py

test(crypto): clean OOS for cross-venue arb (IS-pick config -> OOS-blind apply)

2026-06-07 19:03:30 +02:00

xvenue_sim.py

research(crypto): cross-venue funding arb survives net-of-cost with hysteresis

2026-06-07 18:33:28 +02:00