We backtested a raw trading-signal scanner. It lost. Here's all the data.

Name: ClaudeQuantAlgo Raw-Scanner Hypothetical Backtest (Jan–Jul 2026)
Creator: ClaudeQuantAlgo
License: https://creativecommons.org/licenses/by/4.0/

Open dataset · window 2026-01-02 → 2026-07-09 · 40 symbols · updated July 2026 · CC BY 4.0. No real trades were placed. Research and education only — not financial advice.

The one-line finding: traded blind with no human filter, a raw ±5%-move-on-2×-volume momentum scanner produced 161 simulated trades at a undefined% win rate and a 0.82 profit factor — it lost money (net -323 premium-proxy units, undefined% expectancy per trade). This is the baseline most "trading signal" products are built on before any filtering. All numbers below are hypothetical/simulated.

Simulated trades

161

Win rate

undefined%

Profit factor

0.82

Expectancy / trade

undefined%

Net (units)

-323

⬇ Full trade log (CSV) ⬇ 21-variant grid (CSV) ⬇ Dataset (JSON)

Equity curve of a hypothetical raw momentum scanner backtest losing 323 units over 161 simulated trades, Jan to Jul 2026 — Cumulative simulated result of trading every raw signal blind. Hypothetical — no real trades.

Net result by month (simulated)

Month	Trades	Wins	Net units
2026-01	19	7	-5
2026-02	33	16	-38
2026-03	20	6	-126
2026-04	21	13	+108
2026-05	34	17	+73
2026-06	23	9	-309
2026-07	11	7	-26

21 rule variants on the same data — and the one that fooled us

We re-ran the identical frozen data through 21 rule variants. Momentum-following lost in every configuration. The best-looking cell — fading the pops — showed a positive simulated net, but our own statistical audit rejected it: one ticker accounted for 61% of the simulated profit and the result was not statistically significant after testing 21 variants. It is a hypothesis for forward testing, not an edge, and no live card cites it. Full method and audit on the record page.

Table of 21 hypothetical backtest variants, most losing money, with two mean-reversion rows that looked profitable but failed a statistical audit — All 21 variants, losers included. Hypothetical/simulated.

Cell	What changed	Trades	Win %	PF	Net	H1	H2
A baseline	published rules (validation cell — must match backtest.js)	161	46.6%	0.82	-323	-170	-153
B long only	momentum longs only	102	42.2%	0.7	-367	-252	-115
C short only	momentum shorts only	59	54.2%	1.08	+44	+82	-38
D fade both	MEAN-REVERSION: trade against every signal	161	54%	1.22	+330	+193	+137
E fade pops	short the +5% pops only	102	58.8%	1.41	+362	+280	+82
F fade dumps	buy the -5% dumps only	59	45.8%	0.95	-31	-86	+55
G pullback	limit at signal close, wait 2 sessions (trigger-zone style)	143	42%	0.62	-738	-524	-214
H pullback short	pullback entries, shorts only	53	49.1%	0.65	-262	-216	-46
I chg8	bigger shock: \|8%\| signal	103	45.6%	0.82	-197	-111	-86
J chg10	bigger shock: \|10%\| signal	78	41%	0.69	-311	-191	-120
K rvol3	heavier volume: 3x	56	35.7%	0.77	-159	-326	+167
L rvol5	heavier volume: 5x	17	41.2%	0.8	-31	-45	+14
M sess3	shorter hold: 3-session time-stop	161	46%	0.83	-268	-370	+102
N sess10	longer hold: 10-session time-stop	161	50.3%	0.8	-485	-302	-183
O tight exits	tight geometry: TP1 8 / TP2 20 / stop 10	161	47.2%	0.85	-231	-262	+31
P asym exits	asymmetric: TP1 16 / TP2 40 / stop 10	161	43.5%	0.78	-386	-321	-65
Q runner	runner: TP2 60, 10-session hold	161	50.3%	0.81	-463	-302	-161
R short chg8	gen2: shorts only + \|8%\| shock	33	51.5%	0.82	-71	+80	-151
S short sess3	gen2: shorts only + 3-session hold	59	45.8%	0.69	-186	-150	-36
T fadepop rvol3	gen2: short the pops + 3x volume	35	74.3%	1.53	+156	+259	-104
U pullshort tight	gen2: pullback shorts + tight exits	53	43.4%	0.71	-171	-125	-46

Method (one paragraph)

close-to-close >=|5%| on >=2x 30d avg volume -> next-open entry; TP1 +16% underlying (half), TP2 +40% (rest), stop -20%, time-stop 5 sessions; premium proxy 2.5x; stop checked before TP; stops fill at worse of open/level, TPs fill at the level never better. Universe: NVDA, TSLA, AMD, PLTR, SMCI, MU, META, ORCL, MARA, RIOT, COIN, HOOD, SOFI, RIVN, LCID, OPEN, OKLO, RKLB, ACHR, PLUG, AAL, DAL, UAL, BBIO, IONS, QGEN, LASR, AVAV, OUST, HIVE, BBAI, WEN, AMC, GME, IONQ, RGTI, JOBY, CELH, UPST, AFRM. Every ambiguity was resolved against the strategy. Known limitations (disclosed): the universe was chosen with hindsight (survivorship bias); no commissions, spreads, slippage, or implied-volatility effects are modeled; overlapping signals are equal-weighted; six months is a single market regime. Hypothetical performance results have inherent limitations, are prepared with hindsight, do not reflect actual trading, and no representation is made that any account will or is likely to achieve similar profits or losses.

Cite this dataset

Free to quote and link (CC BY 4.0). Suggested attribution:

ClaudeQuantAlgo (2026). "Raw-Scanner Hypothetical Backtest, Jan–Jul 2026." 161 simulated trades, undefined% win rate, 0.82 profit factor. Retrieved from https://www.claudequantalgo.com/data/scanner-backtest/ — hypothetical/simulated results, not investment advice.

Why publish our own losing baseline? Because it's the number every signal seller hides. The filtering that happens after the scan — catalyst checks, adversarial review, liquidity screens — is the actual product, and the only honest proof is a live record that keeps its losers. That record is public and free to watch. Join the floor →

Disclosures. All figures on this page are HYPOTHETICAL and SIMULATED; no real trades were placed. ClaudeQuantAlgo is a research and education community and is not a registered investment adviser or broker-dealer. Nothing here is financial advice or a recommendation to buy or sell any security or derivative. Past performance — real or simulated — does not indicate future results. Trading involves substantial risk of loss.