// RESEARCH DIARIES

Blog

Trading hypotheses, honest tests, and lessons from the results

Latest story

October 14, 2026•By StonkHub Research

The Options Profit That Stale Prices Invented

An iron-condor backtest looked profitable until stale option prints were removed. The stronger lesson was about pricing four legs at one executable moment.

Research DiariesOptions researchData quality

October 7, 2026•By StonkHub Research

The Breakout That Trading Costs Ate

An opening-range breakout produced a tiny gross gain. Its trading costs were much larger, and the experiment exposed a second problem: invalid tests.

Research DiariesBacktestingExecution costs

October 1, 2026•By StonkHub Research

Does Reddit Attention Predict Stock Returns?

We tested whether unusual Reddit attention predicted later returns. The daily signal disappointed, and the test exposed questions worth asking next.

Research DiariesBacktestingReddit sentiment

💀 The Graveyard

Rejected research trials and the lessons they leave behind.

Share on X

Research trials recorded: 1,242
Trials rejected: 1,235
Enabled paper / live strategies: 127

Trials are all-time research ledger entries, including repeat tests of the same hypothesis. Rejected trials are a subset of those entries. Enabled strategies are a separate count of composite strategies in paper or live stages, not survivors of these trials or evidence of profitability.

Explore failed experiments

Top causes of death: negative sharpe (603) · failed deflation (575) · failed econ or overfit (57)

R.I.P.
down3 · sma200 ≥h10
2026-10-15 · trial #39
Sharpe 0.07 → PSR 22% (needed 95%)
“Edge vanished under the multiple-testing penalty.”
R.I.P.
down3 · sma200 ≥h5
2026-10-15 · trial #38
Sharpe 0.06 → PSR 9% (needed 95%)
“Edge vanished under the multiple-testing penalty.”
R.I.P.
down3 · sma200 ≥h3
2026-10-14 · trial #37
Sharpe -0.01 → PSR 0% (needed 95%)
“Lost money in-sample. No autopsy needed.”
R.I.P.
down3 · any ≥h10
2026-10-14 · trial #36
Sharpe 0.02 → PSR 1% (needed 95%)
“Edge vanished under the multiple-testing penalty.”
R.I.P.
down3 · any ≥h5
2026-10-13 · trial #35
Sharpe 0.03 → PSR 0% (needed 95%)
“Edge vanished under the multiple-testing penalty.”
R.I.P.
down3 · any ≥h3
2026-10-13 · trial #34
Sharpe 0.01 → PSR 0% (needed 95%)
“Edge vanished under the multiple-testing penalty.”
R.I.P.
IBS Lt0.2 · sma200 ≥h10
2026-10-12 · trial #33
Sharpe 0.03 → PSR 3% (needed 95%)
“Edge vanished under the multiple-testing penalty.”
R.I.P.
IBS Lt0.2 · sma200 ≥h5
2026-10-12 · trial #32
Sharpe 0.03 → PSR 1% (needed 95%)
“Edge vanished under the multiple-testing penalty.”
R.I.P.
IBS Lt0.2 · sma200 ≥h3
2026-10-11 · trial #31
Sharpe 0.01 → PSR 0% (needed 95%)
“Edge vanished under the multiple-testing penalty.”
R.I.P.
IBS Lt0.2 · any ≥h10
2026-10-11 · trial #30
Sharpe 0.08 → PSR 24% (needed 95%)
“Edge vanished under the multiple-testing penalty.”

Signal-family outcomes

Signals triggered in the trailing 30 days with a recorded 24-hour return, for active signal types represented in the research ledger. Win rate is recorded winner flags divided by all eligible signals; missing flags do not count as wins. These are signal evaluations, not executed trades or returns after fees and slippage.

The current evaluator marks bullish moves above +0.5% and bearish moves below −0.5% as wins at 24 hours. Neutral signals use a price excursion of at least 3%. Historical rows retain their recorded flags; a high win rate alone does not establish alpha.

Signal outcomes for the trailing 30 days, ordered by evaluated signal count
Signal family	Evaluated signals	Recorded win rate
Enhanced Confluence	25,791	36.2%
MACD Crossover	20,676	30.6%
composite	6,722	36.8%
Pm Cross Source Bearish	2,712	33.6%
Pm Cross Source Bullish	612	33.3%