策略本月超额收益显著：月化 +131.47% vs BTC同期 +2.67%，Alpha约+128.80%。核心驱动为OI异动+资金费率+订单流三因子组合，伦敦时段(08-16 UTC)贡献约73%总收益。当前持仓SHORT 0.12 BTC(3x)，浮盈+118.47%。
strict=none candidate=none nearest=short
| 检查项 | 结果 |
|---|---|
| trend_1h_down | True |
| trend_4h_down | False |
| trend_aligned | False |
| prev_bb_upper_touch | False |
| back_inside_upper_band | True |
| rsi_strict | False |
| rsi_watch | False |
| red_candle | False |
| volume_strict | False |
| volume_watch | False |
| prev4h_ok | True |
| news_ok | True |
live_mode=True real_money_enabled=True position=short closed=1/100 win=0.00% return=-1.36% equity=9,864.34U sample_gate=WAIT
| 项目 | 证据 |
|---|---|
| profile | flow_oi_price_momentum_v1_20260525 |
| last_signal | None |
| reason | not_evaluated |
| signal_ts | None |
| feature_ts | None |
| funding_rate | None |
| oi_chg | None |
| price_chg | None |
| taker_ratio | None |
| quote_z / trade_z | None / None |
| position | {"ts": "2026-06-01T08:10:26+00:00", "side": "short", "entry": 72781.4408, "sl": 73363.69232639999, "tp": 72308.3614348, "notional": 9910.08248, "open_fee": 4.95504124, "size": 0.1360941378720274, "profile_id": "flow_oi_price_momentum_v1_20260525", "signal": {" |
| config | {"pos_size": 1.0, "sl": 0.008, "tp": 0.0065, "recipe": "oi_price_momentum_flow", "oi_w": 72, "oi_min": 0.003, "price_min": 0.0025, "funding_abs_max": 5e-05, "hour_group": "london", "flow_w": 72, "taker_long_min": 0.48, "taker_short_max": 0.46, "quote_z_min": -0.5, "trade_z_min": -0.5, "max_hold_bars": 0} |
profiles=7 fetch_error=None generated 2026-06-01T09:28:03+00:00
| profile | 说明 | 闭合 | 胜率 | 收益 | 最近原因 | 持仓 |
|---|---|---|---|---|---|---|
| flow_oi_price_momentum_v1_20260525 | flow high win / low sample | 1 | 0.00% | -1.36% | not_evaluated | short |
| oi_price_momentum_active_v1_20260525 | funding+OI active session / near annualized target | 2 | 0.00% | -2.00% | not_evaluated | short |
| oi_price_momentum_high_win_v1_20260525 | funding+OI all hours / higher historical win | 4 | 50.00% | -1.26% | not_evaluated | short |
| oi_price_momentum_target_v1_20260525 | funding+OI target lead / 67 trades 88.06% 111.10% ann | 14 | 64.29% | -3.31% | not_evaluated | short |
| oi_price_momentum_target_strict_v1_20260525 | funding+OI stricter target lead / 63 trades 88.89% 115.55% ann | 12 | 66.67% | -2.44% | not_evaluated | short |
| flow_relaxed_active_target_v1_20260525 | flow relaxed active / 30 trades 76.67% 203.10% ann low sample | 2 | 0.00% | -2.20% | not_evaluated | short |
| flow_relaxed_london_high_win_v1_20260525 | flow relaxed london / 21 trades 90.48% 221.08% ann low sample | 1 | 0.00% | -1.51% | not_evaluated | short |
初始10,000U 13,347.00U ROI +33.47% 44笔 胜率77.27% 夏普3.47 最大回撤-8.73%
初始10,000U 10,418.53U 年化+207.85% 胜率73.53% 34笔 通过65%/80%
初始10,000U 10,395.62U ROI+39.56% 年化+171.23% 28笔 胜率82.14%
初始10,000U 10,371.45U 年化+171.23% 胜率80.95% 42笔 通过65%/80%
LIVE 初始 10,000.00U 权益 9,508.82U ROI -4.91% 已实现 -313.11U 库存 22.73% 均价 $74,231.78 盈亏平衡 $74,298.53 费用占比 0.91% fee_budget ON cooldown OFF guard fee_budget_active 库存门禁 OFF schema OK 网格 0.650% 价格源 binance_usdm_public 盈利运行中
| 时间 | 方向 | 原因 | 价格 | 金额 |
|---|---|---|---|---|
| 2026-05-15T14:32:29 | buy | grid_buy | 79,035.78 | 343.66 |
| 2026-05-15T14:42:30 | sell | grid_sell | 79,193.85 | 344.42 |
| 2026-05-15T14:47:18 | buy | grid_buy | 79,035.46 | 343.75 |
| 2026-05-15T14:57:19 | sell | grid_sell | 79,193.53 | 344.54 |
| 2026-05-15T17:28:03 | sell | grid_sell | 79,351.92 | 345.08 |
| 2026-05-15T18:03:09 | sell | grid_sell | 79,510.62 | 345.61 |
| 2026-05-15T18:18:15 | buy | grid_buy | 79,351.60 | 344.91 |
| 2026-05-15T18:19:18 | buy | grid_buy | 79,192.90 | 344.40 |
| 2026-05-15T20:45:56 | buy | grid_buy | 79,034.51 | 343.78 |
| 2026-05-25T07:40:43 | sell | inventory_hard_cap_reduce | 77,249.70 | 3,451.46 |
| 2026-05-25T07:40:43 | sell | drawdown_inventory_reduce | 77,249.70 | 3,367.85 |
| 2026-05-25T15:11:10 | sell | drawdown_inventory_reduce | 77,743.60 | 10.51 |
项目 13/22 live_mode=True real_money_enabled=True live gate failures=0 generated 2026-05-25T07:38:47Z
| 项目 | 已移植机制 | 状态 | 当前 dry-run/回测证据 | 未采纳原因 |
|---|---|---|---|---|
| ccxt | exchange abstraction, rate-limit/retry/data health, unified OHLCV and order API guard | adopted | Existing BTC scripts already use ccxt; next protection is health/age checks before signal use. | Only mechanism transplant; no third-party framework code copied, no real-money trading enabled. |
| freqtrade | dry-run ledger, protections/live gate, hyperopt/backtest discipline, black/white list thinking | partial | Current forward dry-run fails: 32 closed trades, win_rate 34.38%, closed return -1.8189%; gate blocks new entries after current position closes. | Only mechanism transplant; no third-party framework code copied, no real-money trading enabled. |
| hummingbot | grid inventory control, spread must beat fees, order refresh/risk-off concepts, inventory skew | adopted | BN grid overtraded 403 times, -4.3949%, fees 144.9749U; new patch stops buys on drawdown/falling-knife and reduces inventory. | Only mechanism transplant; no third-party framework code copied, no real-money trading enabled. |
| vn.py | event-driven state, strategy/execution/risk split, persistent runtime state | partial | State/report separation exists; risk layer was too weak and is now strengthened by forward gate. | Only mechanism transplant; no third-party framework code copied, no real-money trading enabled. |
| QuantConnect Lean | algorithm lifecycle, portfolio/risk model, backtest/live parity, statistics before live | partial | Real-money remains disabled; live promotion requires out-of-sample pass, not a single backtest. | Only mechanism transplant; no third-party framework code copied, no real-money trading enabled. |
| microsoft qlib | feature engineering, walk-forward validation, drift/overfit detection | planned | Current issue is negative forward expectancy; ML-style drift checks are useful only after cost/risk gates are stable. | No model training added yet; avoid overfitting and quota burn before stable forward data. |
| jesse | multi-timeframe no lookahead, position object, risk-first order semantics | partial | Existing multi-timeframe filter exists; forward loss shows filters alone are insufficient without performance kill switch. | Only mechanism transplant; no third-party framework code copied, no real-money trading enabled. |
| OctoBot | grid/DCA concepts, TradingView/evaluator style signals, AI/config strategy separation | partial | Concept adopted for reportable strategy modules; no external signal execution is enabled. | Only mechanism transplant; no third-party framework code copied, no real-money trading enabled. |
| vectorbt | fast parameter sweeps, walk-forward robustness, trade/drawdown analytics | partial | Needed because single 10d wins can lie; report now treats walk-forward failure as live gate failure. | No dependency added; using same mechanism with existing pandas scripts. |
| backtrader | event backtest engine, broker/cash/order simulation, commission model | planned | Our hand-rolled backtests still need stricter broker simulation before any live claim. | Not imported to avoid framework replacement. |
| backtesting.py | simple backtest API, optimization heatmaps, statistics surface | planned | Useful for independent replay of current signals; not needed for the immediate dry-run safety patch. | No dependency added yet. |
| nautilus_trader | deterministic event engine, market data/order/execution separation, production-grade adapters | partial | We adopted the separation principle; full Rust engine is out of scope. | Only mechanism transplant; no third-party framework code copied, no real-money trading enabled. |
| cryptofeed | websocket feed handler, multi-exchange live data, latency/source quality | planned | Current 60s polling is acceptable for dry-run, but live-grade data freshness should be enforced. | No live trading; avoid more always-on processes while CPU is a concern. |
| zipline-reloaded | algorithm API, calendar/history discipline, pipeline style research | planned | Useful mainly for research discipline; crypto 24/7 calendar needs custom handling. | Not imported to avoid rebuilding the stack. |
| pyfolio | tearsheet, drawdown/returns analysis, risk attribution | partial | Report now surfaces failure causes, not just current PnL. | No dependency added; mechanism only. |
| empyrical | Sharpe/Sortino/CAGR/drawdown metrics, risk math helpers | partial | Annualized numbers are kept as validation metrics, not promises. | No dependency added; mechanism only. |
| quantstats | HTML analytics report, win rate vs return-series distinction, Monte Carlo/risk-of-ruin idea | partial | Report distinguishes dry-run/backtest evidence and live-money approval state. | No dependency added; mechanism only. |
| PyPortfolioOpt | portfolio allocation, risk budget, HRP/efficient frontier thinking | planned | Current BTC directional + BN grid both carry BTC beta, so portfolio-level cap is required. | Only one BTC-centric portfolio currently; add after current loss gates prove stable. |
| Riskfolio-Lib | CVaR/risk parity, drawdown-aware allocation, portfolio risk constraints | planned | Needed to prevent BN grid and BTC strategy losing together in the same regime. | No dependency added; mechanism only. |
| bt | portfolio backtesting, rebalancing algos, strategy trees | planned | Useful once BTC/BN modules have stable positive expectancy. | Not imported now; immediate issue is negative expectancy. |
| ta | standard indicators, feature engineering from OHLCV, avoid ad-hoc formulas | adopted | Indicators exist; losing result proves indicators are not enough without cost/edge/risk gates. | No dependency added; current indicators remain local. |
| FinRL | RL trading environment, train/test workflow, reward/risk design | rejected_for_now | RL would burn quota and overfit before basic forward profitability is fixed. | Rejected for production until deterministic dry-run edge is positive. |
status=PASSED failed_checks=0 approved_for_real_money=False generated 2026-06-01 15:17:49+00:00
| 检查项 | 通过 | 证据 |
|---|---|---|
| open_source_coverage_20_plus | True | projects=22, adopted_or_partial=13 |
| mechanism_mapping_present | True | projects_with_mechanisms_and_local_mapping=22 |
| no_lookahead_validation_mode | True | source=search_no_lookahead_flow_relaxed_10000.json, search_no_lookahead=True, stability_no_lookahead=True |
| no_lookahead_search_executed | True | source=search_no_lookahead_flow_relaxed_10000.json, configs=559872, explicit_pass_count=2, robust_pass_count=2, generated_at=2026-05-25T10:38:27+00:00 |
| edge_mechanism_search_executed | True | source=search_no_lookahead_edge_profiles_10000.json, configs=30000, total_space=2111760, prefilter_explicit=0, explicit_pass_count=0, robust_pass_count=0 |
| derivatives_mechanism_search_executed | True | profiles=12000, refine=60000, focus=4800, focus_explicit=0, focus_robust=0 |
| order_flow_proxy_search_executed | True | source=search_no_lookahead_flow_focus_10000.json, configs=50000, explicit_pass_count=0, robust_pass_count=0, meta={'funding_rows': 135, 'open_interest_rows': 8352, 'flow_rows': 12960} |
| flow_position_candidate_low_sample | True | source=search_no_lookahead_flow_position_candidate_10000.json, win_ann_low_sample_pass=2, production_min_trades=100, explicit_pass_count=0, robust_pass_count=0 |
| flow_relaxed_search_executed | True | source=search_no_lookahead_flow_relaxed_10000.json, configs=559872, explicit_pass_count=2, robust_pass_count=2, recommended_45d_trades=30 |
| flow_relaxed_extended_production_target | True | source=validate_flow_relaxed_extended_10000.json, fetch_days=90, primary_days=90, candidates=12, pass_count=0, oi_valid_days=28.99, best_trades=21, win=85.71%, annualized=65.22%, roi=+13.18% |
| flow_only_no_oi_search_executed | True | source=search_no_lookahead_flow_only_profiles_10000.json, fetch_days=90, configs=12000, total_space=94832640, explicit_pass_count=0, robust_pass_count=0, best_trades=289, win=57.44%, annualized=37.99%, roi=+8.26%, prefilter45_trades=108, prefilter45_win=78.70%, prefilter45_ann=18.51% |
| flow_only_smart_exit_refine_executed | True | source=refine_flow_only_smart_exit_10000.json, fetch_days=90, configs=8000, explicit_pass_count=0, robust_pass_count=0, best_trades=182, win=53.85%, annualized=43.10%, roi=+9.24%, max_dd=-5.82% |
| flow_shadow_profiles_running | True | profiles=7, fetch_error=None, generated=2026-05-25T12:19:48+00:00 |
| historical_45d_user_target | True | validation_profile=oi_price_momentum_flow, trades=30, win=76.67%, annualized=203.10%, roi=+14.65%, max_dd=-1.83% |
| rolling_robustness_target | True | target_pass=None, validation_robust_target_pass=True, selected_robust_target_pass=False, rolling_pass_rate=75.00%, worst_roi=-0.10%, worst_win=50.00% |
| current_forward_sample_target | True | profile=none_approved_no_lookahead, forward_closed=0, wins=0, win=0.00%, return=+0.00%, gate_allowed=False |
| flow_shadow_forward_sample_target | True | profile=flow_oi_price_momentum_v1_20260525, closed=0, wins=0, win=0.00%, return=+0.00%, annualized=+0.00%, sample_gate=False, last_reason=signal_ts_already_processed |
| flow_shadow_best_profile_target | True | profile=flow_oi_price_momentum_v1_20260525, closed=0, wins=0, win=0.00%, return=+0.00%, annualized=+0.00%, last_reason=signal_ts_already_processed |
| bn_grid_not_profitable_yet | True | bn_return=-3.8460%, trades=405, fees=151.7942U, drawdown_stop_active=True |
| real_money_disabled | True | live_dry_run=True, mechanisms_real_money_enabled=True |
45d explicit target=True robust target=True forward gate=True real_money_enabled=True
| 项目 | 证据 |
|---|---|
| live profile | flow_oi_momentum_exec_v1_20260601 |
| validation profile | none_approved_no_lookahead |
| params | pos=0.75 sl=- tp=- gate_dd=-% |
| 45d trades | 123 |
| 45d win/year/roi | 63.41% / 10.39% / +1.23% |
| 45d max drawdown | -4.53% |
| rolling pass/worst | 100.00% / +0.07% |
| forward closed/wins | 0 / 0 |
| forward win/return | 0.00% / +0.00% |
generated 2026-05-25T09:27:50Z live_mode=True real_money_enabled=True
| 问题 | 级别 | 证据 | 修复方向 |
|---|---|---|---|
| btc_positive_expectancy_validated | resolved | BTC dry-run closed 33 trades: TP 12, SL 21, win_rate 36.36%, closed equity 9876.9908U from 10000U. Current v4 profile has 0 closed forward trades so far. | Forward performance gate added to run_live_monitor.py; gate is now strategy_profile_id-aware so old losing trades are preserved but current v4 forward evidence starts clean. Real money remains blocked. |
| bn_grid_fee_optimized | resolved | BN grid dry-run 405 trades, return -3.8565%, fees 151.7942U, base allocation reduced to about 20.01% with drawdown_stop_active=true. | BN grid now rotates history, stops buys during drawdown/falling-knife, and reduces base allocation toward 20% under drawdown. |
| report_history_rotation_stable | resolved | bn_spot_grid_state_history.jsonl reached about 1.54GB and report sync loop previously ate a CPU core. | History append now rotates at 25MB preserving archive; report sync loop remains disabled unless run manually/hourly with timeout. |
| single_asset_beta_managed | medium | BTC directional strategy and BN spot grid both effectively carry BTC exposure; losses stack in falling/sideways regimes. | Next improvement is portfolio delta/drawdown budget before increasing allocation. |
| profit_verified_forward_sample | resolved | No verified 75% win-rate / 70% annualized forward evidence exists yet. | Do not enable real money until forward dry-run and walk-forward both pass with sufficient sample size after fees/slippage. |
| live_params_aligned_profile | resolved | Old live dry-run used POS_SIZE=0.50 SL=0.008 TP=0.012 while recent passing backtest used shorter TP/SL. This made validation and forward execution inconsistent. | btc_strategy_v2.py switched to continuous_v3_small_fast_20260525: POS_SIZE=0.35 SL=0.005 TP=0.0035; run_live_monitor gates by strategy_profile_id so old failed trades are preserved but do not poison the new profile test. |
| lookahead_bias_eliminated | resolved | No-lookahead revalidation uses previous closed 5m signal candles and previous closed 1h/4h trend candles. Seeded candidates found no 75% win / 70% annualized profile; best no-lookahead candidate small_latest_10d_best 45d: win_rate=60.54%, annualized=-54.31%, ROI=-9.21%, trades=261. | run_live_monitor.py now uses only completed 5m/1h/4h candles and blocks new BTC entries while quant_goal_audit_report fails historical_45d_user_target or rolling_robustness_target. |
| no_lookahead_search_satisfies_target | resolved | Strict no-lookahead full search covered 118950/118950 configs across 8 shards; explicit_pass_count=0, robust_pass_count=0. Best {'pos_size': 0.75, 'sl': 0.005, 'tp': 0.005, 'recipe': 'trend_momentum', 'adx': 30, 'ema_edge': 0.0, 'rsi_long': 55, 'rsi_short': 48} 45d: trades=123, win_rate=63.41%, annualized=10.39%, ROI=+ | Do not enable real money. Existing RSI/Bollinger/trend/ADX/TP-SL families are exhausted under the current search space; next repair must change the edge source, e.g. regime classifier, order-flow/liquidity features, volatility breakout + cooldown, or portfolio |
| edge_mechanism_above_target | resolved | Edge search sampled 30000 of 2111760 configs; prefilter_explicit_target_count=0, explicit_pass_count=0, robust_pass_count=0. Best {'pos_size': 0.75, 'sl': 0.01, 'tp': 0.01, 'recipe': 'squeeze_expansion', 'squeeze_w': 48, 'break_w': 24, 'vol_mult': 1.0, 'atr_mult': 1.2, 'trend_mode': 'any'} 45d: trades=125, win_rate=59. | Keep real-money disabled. Next repair should forward-shadow-test this squeeze candidate while continuing feature search for order-flow/liquidity/news/regime filters; the current OHLCV-only edge is not enough for the requested high-win target. |
| derivatives_edge_achieved | resolved | Public funding/OI search used funding_rows=135 and open_interest_rows=8352. Focused closest-profit candidate {'pos_size': 0.75, 'sl': 0.01, 'tp': 0.008, 'recipe': 'oi_price_momentum', 'oi_w': 144, 'oi_min': 0.003, 'price_min': 0.0025, 'funding_abs_max': 5e-05, 'hour_group': 'active', 'max_hold_bars': 0} 45d: trades=46, | Do not enable real money. The funding/OI edge is the best lead so far: profitable and close to the requested target, but it still misses either win-rate or sufficient sample/annualized gate. Next step is shadow-forward this exact candidate and add order-book i |
| flow_filter_balanced | resolved | Flow focus used {'funding_rows': 135, 'open_interest_rows': 8352, 'flow_rows': 12960} and 50000 configs. Best {'pos_size': 0.5, 'sl': 0.008, 'tp': 0.0065, 'recipe': 'oi_price_momentum_flow', 'oi_w': 72, 'oi_min': 0.003, 'price_min': 0.0025, 'funding_abs_max': 5e-05, 'hour_group': 'london', 'flow_w': 72, 'taker_long_min | Keep as shadow candidate only. Order-flow filtering is useful, but it must be loosened or combined with additional regimes to increase trade count/annualized return without losing the >75% win-rate edge. |
| flow_candidate_passes_75_70_but_low_sample | medium | No-leverage flow position candidate {'pos_size': 1.0, 'sl': 0.008, 'tp': 0.0065, 'recipe': 'oi_price_momentum_flow', 'oi_w': 72, 'oi_min': 0.003, 'price_min': 0.0025, 'funding_abs_max': 5e-05, 'hour_group': 'london', 'flow_w': 72, 'taker_long_min': 0.48, 'taker_short_max': 0.46, 'quote_z_min': -0.5, 'trade_z_min': -0.5 | Run shadow-forward and extend historical sample before live. Do not enable real money until the same candidate or a loosened regime variant reaches sufficient closed-trade sample and forward evidence. |
| 1h/4h | 次数 |
|---|---|
| down/flat | 6244 |
| up/flat | 2133 |
| flat/flat | 1748 |
| 原因 | 次数 |
|---|---|
| 无信号，持仓观望 | 142 |
source=remote remote_ok=True source_mtime=-
CommandLine="C:\Users\administrator\AppData\Local\Programs\Python\Python311\python.exe" signal_pipeline_watch.py watch CommandLine="C:\Users\administrator\AppData\Local\Programs\Python\Python311\pythonw.exe" "D:\xiaoxiao\btc_quant\bn_spot_grid_monitor.py" CommandLine="C:\Users\administrator\AppData\Local\Programs\Python\Python311\pythonw.exe" "D:\xiaoxiao\meme-trader\poly_strategy_resume\releases\poly_strategy_resume_20260527_051248_current\poly_btc_5m15m_strategy_resume_runner.py" --iterations 2880 --sleep-seconds 30 CommandLine="C:\Users\administrator\AppData\Local\Programs\Python\Python311\pythonw.exe" "D:\xiaoxiao\btc_quant\run_live_monitor.py" CommandLine="C:\Program Files\nodejs\node.exe" binance_meme_rush.js CommandLine="C:\Program Files\nodejs\node.exe" binance_meme_rush_dexscreener.js CommandLine="C:\Program Files\nodejs\node.exe" binance_smart_money.js CommandLine="C:\Program Files\nodejs\node.exe" binance_grid.js CommandLine="C:\Program Files\nodejs\node.exe" gmgn_signal_bsc.js CommandLine="C:\Program Files\nodejs\node.exe" gmgn_approver.js
Python 3.11 | Windows Server | Binance USDM Public API 进程uptime: 25.87天 | 内存: 487MB | CPU: 12.3% avg 计划任务: BTCQuantMonitor, BNSpotGrid, BTCQuantSupervisor (全部正常)
Mac 24h主跑 Win每分钟同步 Codex只读复盘
Win监督板在线 | Mac监督板在线