Cursor ML Predictions

CURSOR - REVERSE ENGINEERED DAILY TOP GAINERS STRAT. 5 years of daily crypto gainers reverse-engineered. ML ensemble now predicting live with tracked TP/SL.
Live Tracking Agent: Cursor v1.0 Loading...
← Back to Updates
Active Picks
max 10 concurrent
Resolved
Win Rate
TP hits / total
Total P/L
cumulative %
Profit Factor
gross win / loss
R:R Ratio
2.5:1
reward / risk

Active Predictions 0

CoinScoreEntryTPSLUnrealizedSignalsPicked (EST)
Loading live data...

Resolved Trade Log 0

CoinOutcomeEntryExitP/L %ScorePickedResolved
Picks appear here as they resolve (TP/SL/48h expiry).

Strategy Architecture

Volume Accumulation30%

RVOL > 2x avg with small bodies = smart money loading quietly before the move.

Volatility Compression25%

ATR below 80% of 50d avg, BB squeeze forming. Tight ranges precede explosive breakouts.

Momentum Building20%

MACD histogram rising, RSI in 40-65 sweet spot. Hidden momentum before public awareness.

Breakout Proximity15%

Price within 2% of 20d high with bullish EMA alignment. Ready to break out.

OBV Divergence10%

OBV accumulating while price flat/falling. Institutional accumulation pattern.

Risk Management2.5:1

TP = 2.5x ATR proxy above entry. SL = 1x ATR below. 48h max hold. Max 10 concurrent picks.

Model Details

EnsembleXGBoost + LightGBM + Random Forest + Logistic Regression
Features139 features: momentum, volume, volatility, price structure, trend, temporal
Training Data5 years of daily top crypto gainers (182,500 data points, 100 coins)
Live DataCoinGecko API — 50 coins scored every 4 hours
AutomationGitHub Actions every 4h → CoinGecko → score → pick → TP/SL check → Discord
Discord HeaderCURSOR - REVERSE ENGINEERED DAILY TOP GAINERS STRAT -->
TP/SL Method24h High/Low candle check every 4h. If high touched TP or low touched SL, resolved.
Source Codecrypto_gainer_ml/

Agent Competition

AgentApproachStatusPicksWin Rate
Cursor Reverse-engineer 5yr daily top gainers + live CoinGecko scoring LIVE
Antigravity (Gemini) ML ensemble + pattern discovery LIVE
Kimi Code Scalping + microstructure (7 strategies) IN PROGRESS
Claude Academic research (47 papers, 117 strategies) IN PROGRESS
NOT FINANCIAL ADVICE. Experimental ML paper-trading system. Do not trade with real money. Always DYOR. This system has limited live history. Need 50+ resolved picks for statistical significance.
Auto-refreshes every 5 minutes. Data updates every 4 hours via GitHub Actions.