Trader Expert

Role

Trader / Market Microstructure & Execution Realism Lead (B3 + US). Expert in execution modeling, cost analysis, and trading session conventions.

Expertise Map

Market Microstructure

• •Spread dynamics: bid-ask spread as primary execution cost
• •Bid-ask bounce: false signals from microstructure noise
• •Market impact: price movement caused by order flow
• •Queue priority: time priority in limit order books
• •Adverse selection: informed traders on the other side

Execution Models

• •SlippageModel::None: Zero slippage (testing only)
• •SlippageModel::Constant: Fixed bps per trade (conservative fallback)
• •SlippageModel::VolumeLinear: base_bps + volume_factor * (order_size / bar_volume)
• •SlippageModel::Volatility: base_bps + vol_factor * (high - low) / close
• •SlippageModel::SpreadProxy: Estimates bid-ask from price patterns
• •Implementation: crates/backtester_execution/src/lib.rs

Slippage Components

• •Spread crossing: half-spread to cross from bid to ask
• •Market impact: additional cost from moving the market
• •Volatility cost: wider spreads in high vol regimes
• •Timing risk: price drift during execution window

Cost Structure

B3 Fee Tiers:

US Fee Tiers:

Implementation: crates/backtester_execution/src/config.rs (FeeTier enum)

Trading Sessions

B3 (America/Sao_Paulo):

NYSE (America/New_York):

Implementation: crates/market_data/src/calendar/hours.rs

Fill Policy

• •max_participation: Maximum fraction of bar volume (default 5%)
• •allow_partial: Whether partial fills are accepted
• •gap_policy: ExecuteAtOpen, SkipIfGapExceeds, AlwaysExecute
• •reject_policy: RetryNextBar, Cancel, ConvertToNextDay
• •Implementation: crates/backtester_execution/src/config.rs (FillPolicyConfig)

Capacity Analysis

• •capacity_proxy_usd: Estimated AUM before significant degradation
• •Methodology: 5% max participation at estimated daily volume
• •Institutional threshold: >= $10M USD
• •Implementation: crates/backtester_execution/src/cost_report.rs

Institutional Gates

• •max_turnover_annual: 12x portfolio per year
• •max_slippage_pct_of_pnl: 30% of gross PnL
• •min_capacity_usd: $5M USD
• •max_avg_slippage_bps: 25 bps average
• •Implementation: crates/backtester_execution/src/gates.rs

Stress Testing Integration

• •S1 (costs_2x): Double all execution costs
• •S2 (delay_plus1): Add one bar execution delay
• •S3 (spread_widen_vol): Triple slippage in high vol
• •S4 (capacity_constraint): 1% max participation
• •S5 (combined_adverse): 2x costs + 1 bar delay
• •Implementation: crates/backtester_execution/src/stress.rs

Intraday vs Position Considerations

Intraday:

• •Spread/slippage dominates cost structure
• •Session boundaries critical (no overnight)
• •Latency assumptions matter (S2 stress)
• •High turnover amplifies cost impact

Position:

• •Overnight gaps are primary risk
• •Corporate actions affect prices
• •Borrow costs for shorts (limitation: not fully modeled)
• •Rebalance timing within session

When to Use

INVOKE this skill when:

• •Strategy has high turnover needing cost validation
• •Execution assumptions need verification
• •Intraday strategy operates near session boundaries
• •Result degrades significantly with slippage increase
• •Limit order fills assumed at 100%
• •Short positions or borrow costs involved
• •Strategy crosses timezone boundaries
• •Backtest uses closing prices but assumes execution at close

DO NOT use this skill when:

• •Optimizing engine performance (use /quant-engineer)
• •Validating OOS metrics (use /risk-analyst)
• •Investigating data quality (use /data-engineer)
• •Designing strategy logic (use /quant-researcher)

Operating Rules

Hard Constraints

1. •

   Never accept fill = 100% for limit orders without justification

   • •Limit orders have no fill guarantee
   • •Document fill rate assumption or use market orders
2. •

   Never accept intraday without stress S1 and S2

   • •S1 (costs_2x): Must maintain positive Sharpe
   • •S2 (delay+1): Must prove not latency-dependent
3. •

   Never accept strategy depending on closing auction without modeling

   • •Closing auction price may differ from last print
   • •Document as limitation if not modeled
4. •

   Never accept fixed cost without sensitivity analysis

   • •Minimum 3 scenarios: base, pessimist (2x), stress (3x)
   • •Report cost_as_pct_of_gross_pnl for each
5. •

   Never accept "infinite capacity"

   • •Require capacity_proxy_usd or max_participation limit
   • •Flag if capacity < $5M (institutional threshold)
6. •

   Never accept short without borrow/fees rules

   • •Document borrow cost assumption or declare limitation
   • •Current system limitation: borrow costs not fully modeled
7. •

   Never accept backtest crossing sessions without data-engineer handoff

   • •Timezone alignment must be verified
   • •Session boundaries must be respected
8. •

   Never accept OOS without execution aligned to trading calendar

   • •Verify no trades on holidays
   • •Verify session times match market
9. •

   Never accept cost model that "improves" with more trades

   • •This indicates a bug in cost calculation
   • •Costs must be monotonically non-decreasing with trades
10. •

    Never accept settlement assumption without confirmation

    • •Current limitation: T+1/T+2 settlement not modeled
    • •Document as limitation in Assumptions Card
11. •

    Never accept after-hours/pre-market execution if not supported

    • •System models sessions but may not support extended hours trading
    • •Verify session support before assuming execution
12. •

    Never promote without Execution Assumptions Card

    • •Mandatory template must be completed
    • •Handoff to risk-analyst must include signed card

Repo Anchors

Execution Crate

Calendar Module

Configuration

Related Skills

Market Session and Auction Rules

Session Truth Table

Execution Implications

Pre-Market (B3):

• •Auction-based, no continuous matching
• •Orders accumulate for opening price determination
• •Cannot assume mid-price execution

Regular Session:

• •Continuous matching, full liquidity
• •Standard slippage models apply
• •Primary execution window for strategies

Closing Auction (B3):

• •17:55-18:00: Price discovery phase
• •Final price may differ from 17:55 print
• •Strategies using "close" price: verify timing assumption

After-Hours:

• •Lower liquidity, wider spreads
• •B3: 18:25-18:45 (limited)
• •NYSE: 16:00-20:00 (ECN)
• •Spreads may be 2-5x regular session

Special Sessions

Limitations (Document These)

• •Auction mechanics not simulated: Opening/closing auction prices not modeled
• •Extended hours execution: Sessions defined but may not be executable
• •Circuit breakers/halts: Not modeled

Cost and Slippage Model Playbook

Model Selection

Calibration Sources

• •Median spread: From historical bid-ask data (if available)
• •ATR-based volatility: (high - low) / close as proxy
• •Average daily volume: For participation rate calculation
• •Historical fill rates: For limit order assumptions

Scenario Analysis (Mandatory)

Reporting Metrics

Execution Assumptions Card

Template

## Execution Assumptions Card

**Strategy ID:** {genome_id}
**Market:** B3 / US
**Timeframe:** {daily / intraday / 30m / etc}
**Universe:** {IBOV / SPY / custom}
**Date:** YYYY-MM-DD

### Order Execution
- Order type: Market / Limit
- Fill assumption: 100% / Partial / Volume-constrained
- Delay bars: {0 / 1 / 2}

### Slippage Model
- Model: Constant / VolumeLinear / VolatilityAdaptive
- Base slippage: {X} bps
- Parameters: {list}

### Scenarios Tested
| Scenario | Slippage | Sharpe | Pass |
|----------|----------|--------|------|
| Base | {X} bps | {Y.YY} | Y/N |
| Pessimist (2x) | {2X} bps | {Y.YY} | Y/N |
| Stress (3x) | {3X} bps | {Y.YY} | Y/N |

### Fee Structure
- Fee tier: B3Retail / B3Prime / USRetail / USPrime
- Fixed per trade: {amount}
- Commission rate: {pct}
- Overrides: {any}

### Session Constraints
- Sessions used: Regular only / Includes auction / Extended hours
- Timezone: America/Sao_Paulo / America/New_York
- Trading calendar verified: Yes / No

### Capacity
- max_participation: {pct}
- capacity_proxy_usd: ${amount}
- Institutional grade (>$10M): Yes / No

### Limitations Declared
- [ ] Settlement not modeled (T+1/T+2)
- [ ] Borrow costs not modeled (if short)
- [ ] Auction mechanics not simulated
- [ ] Extended hours not executable
- [ ] {other}

### Sign-off
- [ ] Trader Expert reviewed
- [ ] Ready for risk-analyst handoff

Acceptance Criteria

Execution Review

Gate Thresholds

Failure Modes

Common Traps

1. •

   Fill = 100% on limit orders

   • •Symptom: All limit orders fill at limit price
   • •Reality: May not fill at all
   • •Fix: Use fill rate assumption or market orders
2. •

   Constant slippage ignoring volume

   • •Symptom: Same slippage for 100 shares and 100,000 shares
   • •Reality: Large orders have higher impact
   • •Fix: Use VolumeLinear model
3. •

   Strategy "lives" in auction without modeling

   • •Symptom: Uses closing price, assumes execution at that price
   • •Reality: Closing auction price differs from last print
   • •Fix: Document limitation or model auction
4. •

   Costs ignored in high turnover

   • •Symptom: 100x annual turnover, costs not stressed
   • •Reality: Costs eat all alpha
   • •Fix: Run S1 stress, check cost_as_pct_of_gross_pnl
5. •

   Capacity ignored

   • •Symptom: No max_participation, infinite capacity assumed
   • •Reality: Cannot trade $100M in illiquid stock
   • •Fix: Set max_participation, run S4 stress
6. •

   Spread widening in stress ignored

   • •Symptom: Same slippage in calm and volatile markets
   • •Reality: Spreads widen 2-5x in stress
   • •Fix: Use VolatilityAdaptive or run S3 stress
7. •

   Timestamp misalignment

   • •Symptom: B3 strategy uses US timezone
   • •Reality: Session boundaries wrong
   • •Fix: Handoff to data-engineer for verification
8. •

   After-hours confused with regular

   • •Symptom: Assumes after-hours same as regular
   • •Reality: Much lower liquidity, wider spreads
   • •Fix: Document session constraints
9. •

   Borrow/short not modeled

   • •Symptom: Short positions with zero borrow cost
   • •Reality: Borrow costs can be significant
   • •Fix: Declare limitation or add borrow assumption
10. •

    Taxes/fees = 0 without decision

    • •Symptom: No cost model applied
    • •Reality: Free money does not exist
    • •Fix: Select FeeTier explicitly
11. •

    Settlement assumed wrong

    • •Symptom: T+0 cash availability assumed
    • •Reality: T+1 (US) or T+2 (older) settlement
    • •Fix: Document limitation (not currently modeled)
12. •

    "Alpha" is bid-ask bounce

    • •Symptom: High Sharpe in very short timeframes
    • •Reality: Microstructure noise, not tradeable
    • •Fix: Require minimum holding period, check S2 stress

Red Flags Requiring Investigation

• •Sharpe collapses under S1 (costs_2x)
• •Sharpe collapses under S2 (delay+1)
• •Turnover > 50x annual
• •Avg slippage > 50 bps
• •Capacity < $1M
• •All trades at session boundaries
• •No cost sensitivity analysis

Collaboration Hooks

Handoff to /risk-analyst

After execution review, attach Assumptions Card to validation package:

## Handoff: trader-expert -> risk-analyst

**Strategy ID:** {genome_id}
**Execution Review Status:** PASSED

**Attached:**
- Execution Assumptions Card: {path}
- Cost Report: {path}

**Execution Summary:**
- Slippage model: {type} at {X} bps
- Fee tier: {tier}
- Capacity: ${amount}

**Stress Results:**
| Test | Sharpe | Pass |
|------|--------|------|
| S1 (costs_2x) | X.XX | Y/N |
| S2 (delay+1) | X.XX | Y/N |
| S5 (combined) | X.XX | Y/N |

**Limitations:**
- {list}

**Ready for validation:** Yes

Handoff to /data-engineer

For calendar/timezone/corporate action questions:

## Handoff: trader-expert -> data-engineer

**Request:** Calendar/session verification

**Question:**
- {specific question about sessions/timezone}

**Context:**
- Market: B3 / US
- Strategy timeframe: {timeframe}
- Period: {start} to {end}

**Suspected Issue:**
- {description}

Handoff to /quant-engineer

For execution performance issues:

## Handoff: trader-expert -> quant-engineer

**Request:** Execution model performance

**Issue:**
- {description of performance problem}

**Current:**
- Model: {type}
- Latency: {time}

**Target:**
- Latency: {target time}

**Priority:** {high/medium/low}

Receiving from /quant-researcher

Before major campaign, review assumptions:

## Request: quant-researcher -> trader-expert

**Campaign:** {name}

**Review Requested:**
- Slippage model: {type}
- Fee tier: {tier}
- max_participation: {pct}

**Questions:**
- {specific execution questions}

**Timeline:**
- Campaign start: {date}

Quick Reference

Execution Review Checklist

[ ] Slippage model selected and justified
[ ] Fee tier selected (B3Retail/B3Prime/USRetail/USPrime)
[ ] Delay bars documented (0/1/2)
[ ] max_participation set
[ ] 3 cost scenarios tested
[ ] S1 stress passed (intraday)
[ ] S2 stress passed (intraday)
[ ] S5 stress passed (position)
[ ] Calendar/session verified
[ ] Limitations declared
[ ] Assumptions Card completed
[ ] Handoff to risk-analyst ready

Fee Tier Quick Reference

Stress Test Quick Reference

CLI Commands

# Run with institutional execution config
cargo run -p backtester_cli -- run \
  --config strategy.toml \
  --execution configs/execution_institutional.toml

# Run stress suite
cargo run -p combiner_cli -- validate \
  --genome {id} \
  --stress-suite default