jev-trade
Simulated crypto trading loop that turns BTC and ETH tape features into words, asks Jev six judgments per tick and models costs and latency.
Overview
jev-trade is a simulated high-frequency crypto trading loop for BTC and ETH that makes buy and sell calls with Jev. Code computes features (returns, volatility, z-scores, order-book imbalance) from a Kraken or synthetic feed and then discretises every number into words from a closed vocabulary, because Jev is documented as weaker on numbers; for example, the spread is described relative to a typical move rather than in basis points, and a test enforces the vocabulary. Six typed questions go to Jev in one call, and a cost-aware execution simulator turns the decisions into P&L.
The author corrects a brief that called Jev slow, citing TypeSafe's reported 70 to 500 ms, and treats latency as a tunable: a sweep command adds artificial latency to show its cost. The CLI also includes baselines, a single unpacked decision, reading exchange listing announcements, and a central-bank FX experiment including dot-plot and press-conference arms. Without a key the loop runs against an offline stub and says so on every report. It has no required dependencies on Python 3.10+.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Does the tape lean toward buying, selling or neither?(inferred) | — |
| Other | Six judgments per tick about direction, regime and execution conditions | — |
How it works
- Ingest ticks and compute features in code
- Discretise numbers into closed-vocabulary words
- Ask six typed questions in one Jev call
- Simulate execution with costs and latency; compare to baselines
Strengths
- Code owns all numbers; model only judges
- Latency sweep quantifies speed's impact
- Offline stub clearly labelled
- No required dependencies
Caveats
- Simulation only
- Results depend on the discretisation choices
Getting started
python -m jevtrade.cli backtest --baselines (set TYPESAFE_API_KEY and use --provider jev for live Jev).At a glance
- Type
- Project
- Status
- research
- Language
- Python
- Runtime
- CLI
- Has code
- yes
- Has tests
- yes
Use cases and integrations
How curated lists describe it
A simulated crypto trading loop that sends BTC and ETH market features to Jev and models execution costs and latency.
logicrw/awesome-jev-projects
Frequently asked questions
Does Jev see raw prices?
No. Every number is converted into words from a closed vocabulary before reaching Jev.
Can I test the effect of latency?
Yes, the sweep command and --extra-latency-ms simulate slower models.
Does it trade real money?
No, it is a simulated loop with a cost-aware execution simulator.
Related projects
- Jev-Trades★ 51
Next.js crypto dashboard where Jev judges each tick across timeframes and drives a SQLite paper-trading portfolio with automatic TP/SL.
ServiceprototypeTypeScript Local China A-share technical analysis workbench where Jev assesses eight indicator dimensions and returns Buy, Hold or Sell with evidence.
ProjectbetaPython, TypeScriptLocal stock decision service that sends market data and news to Jev or a local model and returns a policy-checked buy/sell/hold.
ProjectbetaTypeScriptPyTorch factor-research stack with 213 factors and cost-aware backtests, plus a preview typed portfolio-decision adapter.
ProjectbetaPythonChinese open-source research and trading workbench whose JEV monitor watches futures contracts and judges entries and exits.
ProjectbetaTypeScriptSelf-hosted trading analysis platform pairing Claude briefings with Jev probabilities for signals like thesis-breaking headlines.
Servicebeta