Own the Market
Intelligence Infrastructure.

# MA Cross Strategy — using pre-calculated MA columns
import pandas as pd

df = pd.read_parquet("EURUSD/M1/")

# MA columns are already in the dataset — no recalculation needed
# Generate signals from EMA12/26 cross (calculate from MA20/50)
df['ema12'] = df['close'].ewm(span=12).mean()
df['ema26'] = df['close'].ewm(span=26).mean()

# Golden cross (EMA12 crosses above EMA26)
df['golden_cross'] = (
    (df['ema12'] > df['ema26']) &
    (df['ema12'].shift(1) <= df['ema26'].shift(1))
)

# Filter: only crosses where RSI is in 'bullish' zone (RSI 50-70)
golden_filtered = df[
    df['golden_cross'] &
    (df['rsi14_zone'] == 'bullish')
]

# Calculate 24h forward return for each signal
df['return_24h'] = df['close'].pct_change(1440).shift(-1440)
signal_returns = df.loc[golden_filtered.index, 'return_24h']

print(f"Total signals: {len(golden_filtered)}")
print(f"Win rate: {(signal_returns > 0).mean():.1%}")
print(f"Avg return: {signal_returns.mean():.4%}")

# RSI Zone Analysis — zone-based return statistics
import pandas as pd

df = pd.read_parquet("XAUUSD/M1/")  # Gold M1 — 25 years

# RSI zone column is pre-tagged in every bar
# Values: 'overbought' | 'bullish' | 'bearish' | 'oversold'

# Forward return (next 24 bars = ~24 minutes for M1)
df['fwd_return'] = df['close'].pct_change(24).shift(-24)

# Group by RSI zone and calculate stats
zone_stats = df.groupby('rsi14_zone')['fwd_return'].agg([
    'count',
    'mean',
    ('win_rate', lambda x: (x > 0).mean()),
    ('avg_win', lambda x: x[x > 0].mean()),
    ('avg_loss', lambda x: x[x < 0].mean()),
])

zone_stats['mean'] = (zone_stats['mean'] * 100).round(4)
zone_stats['win_rate'] = (zone_stats['win_rate'] * 100).round(2)
print(zone_stats.to_string())

# Correlation Break Events — proprietary dataset
import pandas as pd

# Load the correlation event database (347K+ events)
events = pd.read_csv("correlation_break_events_2000_2026.csv", parse_dates=["event_date"])

# Columns: event_date, symbol, pair, corr_before, corr_after,
#          rsi14, rsi_zone, phase, direction,
#          return_1d, return_3d, return_7d, return_30d

# Example: Events where RSI is in oversold zone
oversold_events = events[events['rsi_zone'] == 'oversold']

# Calculate directional accuracy at 7-day horizon
correct = (
    (oversold_events['direction'] == 'up') &
    (oversold_events['return_7d'] > 0)
) | (
    (oversold_events['direction'] == 'down') &
    (oversold_events['return_7d'] < 0)
)

print(f"Oversold zone events: {len(oversold_events):,}")
print(f"7-day directional accuracy: {correct.mean():.1%}")
print(f"Average 7d return: {oversold_events['return_7d'].mean():.4%}")

# Yield Curve Intelligence — inversion analysis
import pandas as pd

# Load yield curve data (6,556 days, 2000–2026)
yc = pd.read_csv("yield_curve_2000_2026.csv", parse_dates=["date"])

# Columns: date, dgs10, dgs2, dgs3mo, dgs30,
#          spread_10y2y, spread_10y3m, spread_30y10y,
#          curve_regime, inversion_days, event_type
# curve_regime: 'normal' | 'flat' | 'inverted'

# Count inversion episodes
inversions = yc[yc['curve_regime'] == 'inverted']
print(f"Total inversion days: {len(inversions)}")

# Show all 34 inversion episodes
episodes = yc[yc['event_type'] == 'INVERSION_START'][['date', 'spread_10y2y']]
print(episodes.to_string())

# Backtest: market behavior after inversion end
yc['next_regime'] = yc['curve_regime'].shift(-60)  # 60 days forward