优化信号

2026-07-13 01:39:48 +08:00
parent cc79650769
commit a6fefa2c9c
10 changed files with 184 additions and 13 deletions
@@ -34,7 +34,7 @@ HTTP_PROXY=http://127.0.0.1:7890
 # TAVILY_API_KEY=         # 网页搜索 (https://tavily.com)
 # TWITTER_API_KEY=        # Twitter 社交情绪
 # SERPER_API_KEY=         # 网页搜索备用 (https://serper.dev)
-# POLYGON_API_KEY=        # 股票、外汇数据 (https://polygon.io)
+# POLYGON_API_KEY=        # 股票、外汇数据 (https://polygon.io, 已更名为 massive.com)
 # FRED_API_KEY=           # 美国经济指标 (https://fred.stlouisfed.org)
 # ETHERSCAN_API_KEY=      # 链上数据 (https://etherscan.io)
 # CONGRESS_API_KEY=       # 美国立法动态 (https://api.congress.gov)
@@ -0,0 +1,3 @@
+{
+  "$schema": "https://app.kilo.ai/config.json"
+}
@@ -0,0 +1,34 @@
+import sqlite3
+conn=sqlite3.connect('data/signals.db')
+c=conn.cursor()
+
+# Trade size vs IAS relationship
+print("=== Trade Size vs IAS (by size bucket) ===")
+for lo,hi in [(0,2000),(2000,5000),(5000,10000),(10000,20000),(20000,50000),(50000,999999)]:
+    row=c.execute('SELECT COUNT(*), AVG(information_asymmetry_score), AVG(trade_size_usd) FROM signals WHERE trade_size_usd>? AND trade_size_usd<=?',(lo,hi)).fetchone()
+    n=row[0]
+    if n>0:
+        avg_i = round(row[1],3)
+        avg_s = round(row[2],0)
+        print(f"  ${lo:,.0f}-${hi:,.0f}: {n:3d} sigs, avg IAS={avg_i:.3f}, avg size=${avg_s:,.0f}")
+
+print()
+
+# Market question frequency
+print("=== Top 10 Most Frequently Signaled Markets ===")
+for r in c.execute('SELECT market_question, COUNT(*) as cnt, AVG(information_asymmetry_score) as avg_i FROM signals GROUP BY market_question ORDER BY cnt DESC LIMIT 10').fetchall():
+    print(f"  {r[0][:65]:65s} | {r[1]:3d} sigs | avg IAS={r[2]:.3f}")
+
+print()
+
+# Resolution status
+print("=== Unresolved Signals (PENDING) ===")
+pending=c.execute('SELECT COUNT(*) FROM signals WHERE market_resolved IS NULL OR market_resolved=?',(0,)).fetchone()[0]
+print(f"  Pending: {pending}")
+
+print()
+
+# Trader wallet frequency
+print("=== Top 10 Most Frequent Trader Wallets ===")
+for r in c.execute('SELECT trader_wallet, COUNT(*) as cnt, AVG(information_asymmetry_score) as avg_i FROM signals GROUP BY trader_wallet ORDER BY cnt DESC LIMIT 10').fetchall():
+    print(f"  {str(r[0])[:25]:25s} | {r[1]:3d} sigs | avg IAS={r[2]:.3f}")
@@ -0,0 +1,58 @@
+import sqlite3, json
+conn=sqlite3.connect('data/signals.db')
+c=conn.cursor()
+
+print("="*60)
+print("SIGNALS DATABASE ANALYSIS")
+print("="*60)
+
+total=c.execute('SELECT COUNT(*) FROM signals').fetchone()[0]
+resolved=c.execute('SELECT COUNT(*) FROM signals WHERE market_resolved=?',(1,)).fetchone()[0]
+correct=c.execute('SELECT COUNT(*) FROM signals WHERE signal_correct=?',(1,)).fetchone()[0]
+incorrect=c.execute('SELECT COUNT(*) FROM signals WHERE signal_correct=?',(0,)).fetchone()[0]
+
+print(f"\nTotal signals: {total}")
+print(f"Resolved: {resolved} | Correct: {correct} | Incorrect: {incorrect}")
+print(f"Win rate: {correct/resolved:.2%}")
+print(f"Avg ROI: {c.execute('SELECT AVG(theoretical_roi) FROM signals WHERE market_resolved=?',(1,)).fetchone()[0]:.2%}")
+
+print("\n=== IAS Distribution ===")
+for lo,hi in [(0.0,0.3),(0.3,0.4),(0.4,0.5),(0.5,0.6),(0.6,0.7),(0.7,0.8),(0.8,1.0)]:
+    row=c.execute('SELECT COUNT(*), AVG(information_asymmetry_score) FROM signals WHERE information_asymmetry_score>? AND information_asymmetry_score<=?',(lo,hi)).fetchone()
+    n=row[0]; avg=round(row[1],3) if row[1] else 0
+    bar='#'*int(n*0.4)
+    print(f"  {lo}-{hi}: {n:3d} signals, avg={avg:.3f} {bar}")
+
+print("\n=== Win Rate by IAS Tier ===")
+for lo,hi in [(0.3,0.4),(0.4,0.5),(0.5,0.6),(0.6,0.7),(0.7,0.8),(0.8,1.0)]:
+    n=c.execute('SELECT COUNT(*) FROM signals WHERE information_asymmetry_score>? AND information_asymmetry_score<=?',(lo,hi)).fetchone()[0]
+    if n>0:
+        cr=c.execute('SELECT COUNT(*) FROM signals WHERE information_asymmetry_score>? AND information_asymmetry_score<=? AND signal_correct=?',(lo,hi,1)).fetchone()[0]
+        print(f"  IAS {lo}-{hi}: {cr}/{n} correct ({cr/n:.1%})")
+
+print("\n=== Trade Size Distribution ===")
+bins=[(0,1000),(1000,2000),(2000,3000),(3000,5000),(5000,10000),(10000,20000),(20000,50000),(50000,999999)]
+for lo,hi in bins:
+    row=c.execute('SELECT COUNT(*), AVG(trade_size_usd) FROM signals WHERE trade_size_usd>? AND trade_size_usd<=?',(lo,hi)).fetchone()
+    n=row[0]; avg=round(row[1],0) if row[1] else 0
+    if n>0:
+        print(f"  ${lo:,.0f}-${hi:,.0f}: {n:3d} signals, avg ${avg:,.0f}")
+
+print("\n=== Signal Time Range ===")
+earliest=c.execute('SELECT MIN(detected_at) FROM signals').fetchone()[0]
+latest=c.execute('SELECT MAX(detected_at) FROM signals').fetchone()[0]
+print(f"  First: {earliest}")
+print(f"  Last:  {latest}")
+
+print("\n=== Top 10 Signals by IAS ===")
+for r in c.execute('SELECT trade_side, trade_outcome, trade_size_usd, trade_price, market_question, information_asymmetry_score, signal_correct, theoretical_roi FROM signals ORDER BY information_asymmetry_score DESC LIMIT 10').fetchall():
+    c_str='CORRECT' if r[6] else ('WRONG' if r[6] is not None else 'PENDING')
+    roi=f'{r[7]:.2%}' if r[7] else 'N/A'
+    print(f"  IAS={r[5]:.2f} | {r[0]} {r[1]} @ {r[3]:.4f} ${r[2]:,.0f} | {r[4][:55]}... [{c_str}] ROI={roi}")
+
+print("\n=== Sample LLM Analysis Reasoning (highest IAS) ===")
+top=c.execute('SELECT reasoning, insider_evidence FROM signals ORDER BY information_asymmetry_score DESC LIMIT 3').fetchall()
+for i, (reasoning, evidence) in enumerate(top):
+    print(f"\n--- Signal #{i+1} ---")
+    print(f"Reasoning (first 500 chars): {str(reasoning)[:500]}")
+    print(f"Insider evidence: {str(evidence)[:500] if evidence else 'None'}")
@@ -0,0 +1,78 @@
+import json, collections, math
+from datetime import datetime
+
+with open('data/processed_transactions.json') as f:
+    txns = json.load(f)
+
+print(f"Total transactions in file: {len(txns)}")
+print()
+
+# Sample
+print("=== Sample (first 3) ===")
+for t in txns[:3]:
+    print(json.dumps(t, indent=2))
+    print()
+
+# Fields available
+print("=== Available Fields ===")
+sample = txns[0]
+print(list(sample.keys()))
+print()
+
+# Direction distribution
+print("=== Direction ===")
+dirs = collections.Counter()
+for t in txns:
+    dirs[t.get('side','?')]+=1
+print(dirs)
+print()
+
+# Side breakdown
+print("=== BUY vs SELL ===")
+for s in ['BUY','SELL']:
+    n=sum(1 for t in txns if t.get('side')==s)
+    print(f"  {s}: {n}")
+print()
+
+# Size distribution
+print("=== Trade Size (USD) Distribution ===")
+bins=[(0,100),(100,500),(500,1000),(1000,2000),(2000,5000),(5000,10000),(10000,20000),(20000,50000),(50000,100000),(100000,999999)]
+for lo,hi in bins:
+    n=sum(1 for t in txns if lo<t.get('usdc_size',0)<=hi)
+    if n>0:
+        avg=sum(t.get('usdc_size',0) for t in txns if lo<t.get('usdc_size',0)<=hi)/n
+        print(f"  ${lo:,.0f}-${hi:,.0f}: {n} txns, avg ${avg:,.0f}")
+print()
+
+# Price distribution
+print("=== Price Distribution (BUY only) ===")
+buy_txns=[t for t in txns if t.get('side')=='BUY']
+price_bins=[(0,0.1),(0.1,0.2),(0.2,0.3),(0.3,0.4),(0.4,0.5),(0.5,0.6),(0.6,0.7),(0.7,0.8),(0.8,0.9),(0.9,1.0)]
+for lo,hi in price_bins:
+    n=sum(1 for t in buy_txns if lo<=t.get('price',0)<hi)
+    if n>0:
+        print(f"  {lo}-{hi}: {n}")
+print()
+
+# Largest trades
+print("=== Top 20 Largest Trades (BUY) ===")
+largest=sorted(buy_txns, key=lambda t: t.get('usdc_size',0), reverse=True)[:20]
+for t in largest:
+    print(f"  ${t['usdc_size']:,.0f} | {t['side']} {t.get('outcome','?')} @ {t['price']:.4f} | {t.get('title','?')[:60]}")
+print()
+
+# Check what happens at each filter stage
+print("=== Filter Stage Analysis (BUY only, 0.0-0.95 price) ===")
+print(f"BUY txns with price in 0.0-0.95: {sum(1 for t in buy_txns if 0<=t.get('price',0)<=0.95)}")
+print(f"BUY txns size >= 1000: {sum(1 for t in buy_txns if t.get('usdc_size',0)>=1000)}")
+print(f"BUY txns size >= 3000: {sum(1 for t in buy_txns if t.get('usdc_size',0)>=3000)}")
+print(f"BUY txns size >= 5000: {sum(1 for t in buy_txns if t.get('usdc_size',0)>=5000)}")
+print()
+
+# Check if there are BUY txns >= 5000 with price in range
+big_buy=[t for t in buy_txns if t.get('usdc_size',0)>=5000 and 0<=t.get('price',0)<=0.95]
+print(f"BUY txns >= $5K and price in range: {len(big_buy)}")
+if big_buy:
+    print("  Sample of qualifying transactions:")
+    for t in big_buy[:5]:
+        print(f"    ${t['usdc_size']:,.0f} @ {t['price']:.4f} | vol={t.get('volume',0):,.0f}")
@@ -56,6 +56,7 @@ class Settings(BaseSettings):
    min_trade_size_usd: float = Field(default=1000.0, alias="MIN_TRADE_SIZE_USD")
    min_price: float = Field(default=0.10, alias="MIN_PRICE")
    max_price: float = Field(default=0.90, alias="MAX_PRICE")
+    premium_threshold: float = Field(default=0.01, alias="PREMIUM_THRESHOLD")

    # Monitoring Settings
    trending_markets_limit: int = Field(default=50, alias="TRENDING_MARKETS_LIMIT")
@@ -254,16 +254,16 @@ class TradeMonitor:
                pass

        # --- 4. Size ---
-        if activity.usdc_size < 3_000:
+        if activity.usdc_size < self.settings.min_trade_size_usd:
            return False

        # --- 5. Dynamic size ---
-        base_size = 5_000.0
+        base_size = max(self.settings.min_trade_size_usd, 5_000.0)
        baseline_volume = 1_000_000.0

-        if market and market.volume > 0:
-            threshold = base_size * math.sqrt(market.volume / baseline_volume)
-            threshold = max(3_000.0, min(threshold, 50_000.0))
+        if market and market.volume_24hr > 0:
+            threshold = base_size * math.sqrt(market.volume_24hr / baseline_volume)
+            threshold = max(self.settings.min_trade_size_usd, min(threshold, 50_000.0))
        else:
            threshold = base_size

@@ -271,12 +271,9 @@ class TradeMonitor:
            return False

        # --- 5.5 Normalized size (like normalized_premium) ---
-        # usdc_size / √(volume) makes signals comparable across market sizes.
-        # A $5K trade in a $50K market is far more significant than $20K in a $10M market.
-        if market and market.volume > 0:
-            normalized = activity.usdc_size / math.sqrt(market.volume)
-            # Minimum normalized threshold: filters out trades that are trivial
-            # relative to market size (calibrated: $5K in a $1M market → 5.0)
+        # Use 24h volume so old markets don't suppress small trades.
+        if market and market.volume_24hr > 0:
+            normalized = activity.usdc_size / math.sqrt(market.volume_24hr)
            if normalized < 1.5:
                return False

@@ -289,7 +286,7 @@ class TradeMonitor:
            else:
                market_mid = 1.0 - market.outcome_prices[0]

-            if activity.price < market_mid + 0.01:
+            if activity.price < market_mid + self.settings.premium_threshold:
                return False

        return True