Bitget RWA stock-futures research and local paper-trading stack.
The current production candidate is a long-only, 5-minute closed-window negative-basis repair strategy for memory and optical RWA contracts. It is designed for local forward paper trading only; it does not place exchange orders.
- Name:
closed_window_basis_5m_15x_m20_cap100_histbest_v4 - Universe:
DRAMUSDT,MUUSDT,NBISUSDT,SNDKUSDT,WDCUSDT,COHRUSDT,LITEUSDT - Direction: long only
- Anchor: latest Yahoo spot bar carried forward during US-market closed windows
- Timeframe: Bitget 5m futures candles
- Fees: 0.1% per side
- Leverage: 15x
- Base sizing: 20% margin per signal, 100% portfolio cap
Entry filters:
- US-market closed session
basis_pctbetween-8.0%and-1.5%basis_z_288 <= -2.0- 5m futures return between
-1.2%and0.2% - 15m equivalent futures return
>= -2.5% basis_delta_1 >= -0.6%- reclaim between
35%and75% - 5m quote volume
>= 16,666.67 USDT
Exit and risk:
- Exit on
basis_pct >= 0 - Exit at US regular open
- Stop if futures price return from entry
<= -1.0% - Max hold: 288 five-minute bars
- Stop opening new positions after CST daily mark return reaches
15%; existing positions are not force-closed - If a single position reaches
>= 50%margin-level floating profit, close it after a12%giveback from its high-water mark - If
trend_not_broken=false, keep the signal but reduce new position margin to10%
The Cboe delivery-point snapshot is retained as an online reference factor, but it is not used as a hard entry gate in v4.
scripts/closed_window_5m_paper_live.py- live local paper-trading enginescripts/render_paper_live_status.py- HTML/JSON status rendererscripts/run_closed_window_5m_paper_live.sh- locked cron runnerscripts/cboe_delivery_point_snapshot.py- delayed Cboe option-chain reference snapshotconfig/selected_closed_window_strategy.json- selected strategy metadata and legacy 15m compatibility fieldsconfig/multi_asset_15m_targets.json- RWA symbol to spot-anchor mapanalysis/**/*.md- research notes and backtest reportstests/- unit tests for core operators and generated-output contracts
Generated snapshots, large CSV/JSON research tables, local paper account state, browser profiles, logs, and secrets are intentionally ignored by git.
Create local credentials only if you need authenticated demo-read checks:
cp .env.example .envRun one 5m paper cycle:
python3 scripts/closed_window_5m_paper_live.py --lookback-days 5
python3 scripts/render_paper_live_status.pyRun the locked production loop once:
scripts/run_closed_window_5m_paper_live.shCron example:
*/5 * * * * cd /path/to/rwa-nbis-history && nice -n 10 ionice -c2 -n7 ./scripts/run_closed_window_5m_paper_live.shServe the generated dashboard:
cd paper_trading
python3 -m http.server 8765 --bind 0.0.0.0python3 -m unittest discover -s tests
python3 -m py_compile scripts/closed_window_5m_paper_live.py scripts/render_paper_live_status.pySome tests validate optional generated research outputs and will skip when those local artifacts are absent.
Key reports:
analysis/stop_work_exit/stop_work_exit_report_test15.mdanalysis/stop_work_exit/stop_work_exit_report_test21.mdanalysis/stop_work_exit/stop_work_exit_report_test30.mdanalysis/take_profit_exit/take_profit_exit_report.mdanalysis/timeframe_compare_5m_15m/timeframe_compare_report.mdanalysis/classic_indicator_diagnostics/classic_indicator_diagnostics_report.mdanalysis/historical_param_search_no_option/historical_param_search_no_option_report.md
This repository is research and paper trading infrastructure, not financial advice. Real order execution requires a separate review, demo-order lifecycle validation, kill-switches, and exchange-side permission isolation.