← ML Lab  ·  Dashboard

OmmAlpha · V8 ML History Builder

Builds the historical setup/outcome population required before XGBoost training. Processing is resumable and intentionally conservative for shared hosting.

Historical price coverage

2023-09-21 Earliest candle
2026-09-29 Latest candle
498 Stocks with price rows
353,672 Total OHLCV rows

Create ML historical replay

A completed ML History V8 run already exists: #2, 2,721 run signals.
2,721 Correct OmmAlpha-v8 signals
0 Legacy / mislabeled signals
Strategy audit passed. All signals attached to this completed run are stamped OmmAlpha-v8.
Recommended first run: use the complete available date range and DEVELOPING + DEVELOPED + PRE_BREAKOUT. OmmAlpha still requires the historical watch-candidate hard gates to pass, so this does not mean every daily candle becomes a training signal.

Run #2 · COMPLETED

ML V8 History · NIFTY500 · 2023-09-21 ? 2026-09-25
2023-09-21 → 2026-09-25

501 / 501 Stocks processed
2,721 Historical signals found
0 Stocks pending
18 Stock errors
100.0% complete
Open ML Lab Export ML Dataset Backtest runs

Recent stock errors

SymbolError
URBANCO Insufficient history: 254 candles.
TMCV Insufficient history: 217 candles.
TENNIND Insufficient history: 212 candles.
TATACAP Insufficient history: 237 candles.
PWL Insufficient history: 213 candles.
PIRAMALFIN Insufficient history: 220 candles.
PINELABS Insufficient history: 215 candles.
MEESHO Insufficient history: 197 candles.
LGEINDIA Insufficient history: 236 candles.
LENSKART Insufficient history: 219 candles.
JAINREC Insufficient history: 244 candles.
ICICIAMC Insufficient history: 190 candles.

What this builder does

  1. Uses the existing NIFTY 500 price history already stored in MySQL.
  2. Replays each stock one historical session at a time.
  3. Uses only candles available on that historical date.
  4. Runs OmmAlpha indicators, VCP detection and watch-candidate gates.
  5. Stores qualifying historical signals in setup_signals.
  6. Evaluates later candles separately into setup_outcomes.
  7. ML Lab then converts those rows into leakage-audited XGBoost training samples.

Default batch size is one stock per browser request to reduce shared-hosting timeouts. The page automatically refreshes until the run finishes, and you can pause/resume at any time.