Strategy Design

This section walks you through a tactical workflow for designing a strategy that is testable, debuggable, and deployable in PredictBack.

Step-by-step workflow

1. Pick a strategy objective

   • Decide what you’re trying to capture:

     • Trend continuation

     • Mean reversion

     • Volatility expansion / contraction

     • Passive liquidity capture

   • Define a single measurable outcome (e.g., positive expectancy with bounded drawdowns).

2. Select your market set (regime coverage)

   • Choose 1–N related markets (same category) and include at least:

     • A trending market

     • A flat / choppy market

     • A jumpy / event-driven market

   • Goal: avoid designing only for one regime.

3. Define the signal inputs (indicators)

   • Choose the minimum indicators needed to express the hypothesis.

   • Keep it simple first; add filters only after you see failure modes.

4. Write explicit entry logic

   • Use either:

     • Crossover conditions for “state changes” (trend flips)

     • Compare conditions for thresholds (RSI > 65)

   • Ensure the entry rule is unambiguous.

5. Write explicit exit logic

   • Always define exits before optimizing entries.

   • Exit types:

     • Invalidation exit (signal flips)

     • Time-based exit (hold N bars)

     • Risk exit (stop-loss)

     • Profit exit (take-profit)

6. Set sizing + constraints

   • Configure:

     • Order size

     • Initial balance

     • Optional stop-loss / take-profit

   • Keep sizing fixed for first pass; add dynamic sizing later.

7. Run the backtest and inspect mechanics

   • Check:

     • Trade count (too many trades = likely noise)

     • Drawdowns (path-dependence)

     • Equity curve smoothness

     • Order history (are trades firing where you expect?)

8. Diagnose and iterate

   • Identify why it fails:

     • Overtrading in chop?

     • Late entries in trends?

     • Large losses during jumps?

   • Then adjust one variable at a time:

     • Indicator period

     • Threshold

     • Add a confirmation filter (AND)

     • Add alternative exits (OR)

---

Key components in PredictBack (quick mapping)

• Price Data

  • The market(s) your strategy runs on.

• Indicators

  • Signals computed from price (SMA, EMA, RSI, MACD, Bollinger).

• Compare

  • Threshold logic using operators (>, <, ≥, ≤, =).

• Crossover

  • Detects when A crosses above/below B.

• AND / OR

  • Combine multiple conditions into a single decision.

• Buy / Sell

  • Trade actions triggered when upstream logic evaluates to true.

Operator definitions

• > greater than

• < less than

• ≥ greater than or equal to

• ≤ less than or equal to

• = equal to

How to Read Backtest Results (What to Check First)

After running a backtest, inspect in this order:

1. Trade Count

   • Too many → noise / overfitting

   • Too few → regime dependence

2. Max Drawdown

   • Indicates path-dependency and tail risk

3. Cumulative PnL Curve

   • Smooth → structural edge

   • Stair-step or cliff-like → event sensitivity

4. Order History

   • Verify trades trigger where logic says they should

   • Check for clustering or over-reaction

5. Win Rate vs PnL

   • High win rate + low PnL → execution drag

   • Low win rate + positive PnL → convex payoff