# put credit spread with LSTM

Trades SPY put credit spreads based on LSTM-predicted volatility changes.

Canonical: https://prismaticalgos.com/compendium/algo/11375-put-credit-spread-with-lstm
Record ID: 11375
Catalog rank: 1167 (reading order, not performance)
Snapshot: 2026-10-31 15:46 UTC
Family: Statistical, Quantitative & Machine Learning (I)
Source: https://github.com/ikinikar/lstm-put-credit-spread
Evidence label (unverified): backtest_shown

## Scope and limitations

These are machine-extracted research records, not verified trading systems. Catalog rank measures specification detail and claimed evidence, not profitability. The catalog extraction pipeline did not independently backtest these records. Its separate selected testing ledger does not validate the wider corpus. Rules, labels and parameters can be incomplete or wrong; consult the original source and its reuse terms.

No endorsement, ownership claim or source-code reuse permission is implied.

## Extraction warnings

Claim warnings: THIN SAMPLE: 39 trades - below the ~100 needed before a win rate or profit factor carries much information.
Opacity warnings: UNDEFINED ENTRY TERM: the entry conditions are prose naming a concept with no numeric definition, so two readers would implement them differently. Advisory only -- this does not affect the score.
Build warnings: DAILY_LOW_COUNT: daily-or-slower timeframe on a single instrument. Expect well under 100 trades in five years, so a backtest will rarely reach statistical significance on its own.

## Recorded profile

- symbols: SPY
- timeframes: D1
- sessions: any
- market_regime: volatile

## Indicators

- SPY close: {}; applied to: close
- VIX close: {}; applied to: close
- MACD: {}; applied to: close
- Bollinger Bands: {}; applied to: close
- Realized Volatility: {"period": 20}; applied to: close

## Recorded long-entry rules

- model forecasts rising volatility
- model accuracy >= 50%

## Recorded short-entry rules

- Not specified.

## Recorded exit rules

- model forecasts falling volatility
- model accuracy drops below 50%

## Recorded filters

- model accuracy >= 50%
- strike within ±15 of underlying
- expiration between 50 and 100 days

## Risk controls

- Stop loss: -2.5% unrealized P&L triggers full liquidation
- Take profit: Not specified
- Position sizing: min net premium as % of max risk: 0.05