jev-trade-cc
Point-in-time O'Neil momentum backtest where Jev picks entries and judges exits, with a jackknife showing no resolvable edge over rules.
Overview
Jev Can Trade Stocks is a backtest of a William O'Neil momentum strategy from January 2022 to September 2026 with $100,000, long only, benchmarked against the S&P 500. Jev picks the entries and decides the exits, while deterministic code owns the data, signals, stops and fills and never delegates risk.
The headline run ended at $207,746 versus $170,234 for the rules alone and $159,500 for S&P 500 buy and hold, but the author explicitly says not to cite it as an edge. A jackknife that deletes 49 tickers and reruns both arms on the same panel shows Jev ahead in only 7 of 12 draws with a mean paired difference of about $4,000 and a $25,000 standard deviation, which the author calls coin-flip territory. The real measured improvement is that prompt and contract fixes moved Jev from reliably worse than the rules (1 of 12, mean -$23,315) to indistinguishable. A Chinese-language write-up distils eight rules for using System One models, starting from an experiment on whether Jev knows what day it is.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Should the strategy enter this candidate stock now?(inferred) | — |
| Yes / no | Should this position be exited?(inferred) | — |
How it works
- Code builds point-in-time data and O'Neil signals
- Jev judges entries among candidates and exits for holdings
- Code applies stops, fills and risk limits
- Compare against rules-only arms and S&P 500; jackknife for robustness
Strengths
- Rigorous paired comparison and jackknife
- Risk management never delegated to the model
- Shows measurable improvement from prompt and contract fixes
Caveats
- No resolvable edge over mechanical rules
- Small portfolio makes outcomes depend on which stocks are caught
- Backtest, not live trading
Getting started
python3 -m http.server 8000 --directory web to view the results page.At a glance
- Type
- Project
- Status
- research
- Language
- Python
- Runtime
- CLI backtest + static web page
- Has code
- yes
Use cases and integrations
How curated lists describe it
Jev Can Trade Stocks — a point-in-time O'Neil momentum backtest where TypeSafe's System One model picks the entries and judges the exits.
hellogumbo/awesome-jev
Frequently asked questions
Does Jev beat the rules?
Not by any margin the experiment can resolve; it led in 7 of 12 jackknife draws.
What improved?
Prompt and contract fixes moved Jev from reliably worse than the rules to indistinguishable from them.
Who controls risk?
Deterministic code owns data, signals, stops and fills.
Related projects
PyTorch factor-research stack with 213 factors and cost-aware backtests, plus a preview typed portfolio-decision adapter.
ProjectbetaPythonTrade Arena proof of concept comparing trading strategies on 13 US stock funds, grown out of a pre-registered Jev study.
ProjectprototypePythonReproducible test of viral Jev HFT claims: direction not calibrated, never under 100 ms, and no edge after trading fees.
ServiceresearchPythonDecision Lab: sixty finance demos asking Jev, OpenAI Decisions and twelve other decision models the same questions, graded.
ProjectresearchJavaScriptBacktest harness that tests Jev as a BUY/SELL/HOLD trader on NQ level-10 order-book data, with replay, viewer and paper-trading modes.
ProjectresearchPythonStudy of whether Jev's reading of financial news predicts stock returns: strong same-day signal, no tradeable alpha.
BenchmarkresearchGo