design-trading-strategies

What this skill does

# Design Trading Strategies

## Quick Start

This skill generates AI-powered trading strategy concepts before you commit to expensive development. It's the cheapest AI tool in Robonet ($0.05-$1.00 vs $1-$4.50 for full strategy creation).

**Load the tool first**:
```
Use MCPSearch to select: mcp__workbench__generate_ideas
```

**Basic usage**:
```
generate_ideas(strategy_count=3)
```

Returns 3 creative strategy concepts with descriptions of market conditions, entry/exit logic, and rationale.

**When to use this skill**:
- Exploring new markets or timeframes
- Stuck for strategy ideas
- Want to validate approach before expensive development
- Need inspiration from current market data
- Brainstorming session before building

**When to skip this skill**:
- You already know exactly what you want to build → Go directly to `build-trading-strategies`
- Iterating on existing strategy → Use `improve-trading-strategies`

## The generate_ideas Tool

**Purpose**: Creates innovative strategy concepts based on current Hyperliquid market data

**Parameters**:
- `strategy_count` (optional, integer, 1-10): Number of ideas to generate (default: 1)

**Returns**: List of strategy concepts including:
- **Strategy name**: Descriptive name for the concept
- **Market conditions**: What market regime it's designed for (trending, ranging, volatile, etc.)
- **Entry logic**: When to enter positions
- **Exit logic**: When to exit positions
- **Risk management**: Stop loss and position sizing approach
- **Indicators**: Which technical indicators to use
- **Rationale**: Why this strategy might work

**Pricing**: Real LLM cost + margin (max $1.00)
- Typical cost: $0.05-$0.50 depending on number of ideas
- Significantly cheaper than `create_strategy` ($1-$4.50)

**Execution Time**: ~20-40 seconds

**Example output**:
```
Strategy 1: "Bollinger Band Mean Reversion"
Market conditions: Ranging market with clear support/resistance
Entry logic: Enter long when price touches lower band and RSI <30
Exit logic: Exit at middle band or when RSI >70
Risk management: 2% position size, stop loss at 1.5× ATR below entry
Indicators: Bollinger Bands (20, 2), RSI (14), ATR (14)
Rationale: Markets tend to revert to the mean in ranging conditions.
           Bollinger Bands identify extremes, RSI confirms oversold/overbought.
```

## Core Concepts

### Strategy Archetypes

**Understanding common patterns helps evaluate AI-generated ideas**:

**1. Trend Following**
- **Market**: Trending markets (strong directional movement)
- **Logic**: Enter in direction of trend, ride momentum
- **Indicators**: Moving averages (EMA, SMA), ADX, MACD
- **Risk**: False breakouts, whipsaws in ranging markets
- **Example**: "Buy when price crosses above 50 EMA and ADX >25"

**2. Mean Reversion**
- **Market**: Ranging markets (price oscillates around mean)
- **Logic**: Buy oversold, sell overbought, profit from reversion
- **Indicators**: RSI, Bollinger Bands, Stochastic
- **Risk**: Trending markets (price may not revert)
- **Example**: "Buy when RSI <30 and at lower Bollinger Band"

**3. Breakout**
- **Market**: Consolidation followed by expansion
- **Logic**: Enter when price breaks support/resistance with volume
- **Indicators**: Bollinger Bands, ATR, Volume, Donchian Channels
- **Risk**: False breakouts, low win rate (requires large profit factor)
- **Example**: "Buy on breakout above 20-day high with volume >1.5× average"

**4. Momentum**
- **Market**: Strong directional moves
- **Logic**: Enter when momentum accelerates, exit when it fades
- **Indicators**: MACD, Stochastic, Rate of Change (ROC), Williams %R
- **Risk**: Late entries, momentum reversals
- **Example**: "Buy when MACD crosses above signal line and price makes new high"

**5. Arbitrage/Market Making**
- **Market**: Any (exploits inefficiencies or spreads)
- **Logic**: Profit from price discrepancies or spread capture
- **Indicators**: Spread analysis, correlation, volume
- **Risk**: Low margins, requires high frequency, execution risk
- **Example**: "Capture spread between related assets or maker/taker fees"

### Realistic Performance Expectations

**Set realistic expectations to evaluate generated ideas**:

**Sharpe Ratio** (risk-adjusted return):
- **>2.0**: Exceptional (rare for algorithmic strategies)
- **1.0-2.0**: Good (achievable with solid strategy)
- **0.5-1.0**: Acceptable (worth testing)
- **<0.5**: Poor (likely not profitable after costs)

**Max Drawdown** (largest peak-to-trough decline):
- **<10%**: Conservative (lower returns, safer)
- **10-20%**: Moderate (balanced risk/reward)
- **20-40%**: Aggressive (higher returns, higher risk)
- **>40%**: Very risky (difficult to recover from)

**Win Rate** (percentage of profitable trades):
- **45-65%**: Realistic for most strategies
- **>70%**: Suspicious (likely overfitting or unrealistic assumptions)
- **<40%**: Needs improvement or higher profit factor

**Typical characteristics of good strategies**:
- Clear entry and exit rules (not ambiguous)
- Realistic indicator parameters (not over-optimized)
- Appropriate for market regime (trend-following in trending markets)
- Risk management included (stop loss, position sizing)
- Based on sound trading principles (not curve-fitted)

### Market Regime Awareness

**Different strategies work in different market conditions**:

**Trending Markets** (strong directional movement):
- **Best**: Trend following, momentum strategies
- **Worst**: Mean reversion (buys dips that keep dipping)
- **Current state**: Check BTC/ETH price action (making new highs/lows?)

**Ranging Markets** (sideways price action):
- **Best**: Mean reversion, oscillator-based strategies
- **Worst**: Trend following (whipsawed by false signals)
- **Current state**: Price oscillating between clear support/resistance?

**Volatile Markets** (large price swings):
- **Best**: Breakout strategies, volatility-based position sizing
- **Worst**: Tight stop losses (get stopped out frequently)
- **Current state**: ATR elevated? Wide intraday ranges?

**Low Volatility Markets** (compressed ranges):
- **Best**: Range trading, spread capture
- **Worst**: Momentum strategies (insufficient movement)
- **Current state**: ATR compressed? Narrow ranges?

**How generate_ideas uses market data**:
- Analyzes current Hyperliquid market data
- Considers recent volatility, trends, and patterns
- Generates ideas appropriate for current conditions
- You should still validate ideas against your own market view

## Best Practices

### Cost Optimization

**Why use generate_ideas before create_strategy**:

```
Without generate_ideas:
1. create_strategy("vague idea") → $2.50
2. Code doesn't match expectations → Wasted
3. create_strategy("try again") → $2.50
4. Still not quite right → Wasted
Total cost: $5.00, no strategy yet

With generate_ideas:
1. generate_ideas(strategy_count=3) → $0.30
2. Review 3 concepts, pick best
3. create_strategy("refined concept from idea #2") → $2.50
4. Code matches expectations → Success
Total cost: $2.80, working strategy
```

**Savings: $2.20 (44% reduction) + better outcome**

### Effective Idea Generation

**Request multiple ideas to compare approaches**:
```
generate_ideas(strategy_count=3-5)
```

**Why 3-5 ideas**:
- See multiple approaches to same problem
- Compare trade-offs (risk vs return, complexity vs simplicity)
- Identify common themes (e.g., all ideas use RSI → probably important)
- Cost is still minimal ($0.15-$0.50 total)

**Avoid**:
- `strategy_count=1`: Limited perspective, may miss better approaches
- `strategy_count=10`: Diminishing returns, most ideas will be variations

### Critical Evaluation of AI Ideas

**Not all AI-generated ideas are good. Evaluate critically**:

**Red flags** (be skeptical of these):
- **Unrealistic parameters**: "Buy when RSI exactly equals 27.6" (over-optimized)
- **Too many indicators**: Uses 8+ indicators (overfitting risk)
- **Vague logic**: "Enter when conditions are favorable" (not actionable)
- **No risk management**: Doesn't mention