Case Study 1: Trading a Presidential Election — Strategy Comparison

Overview

This case study simulates trading a presidential election prediction market over a six-month period (June through November) using three different strategies: fundamental analysis, event-driven trading, and mean reversion. We track each strategy's performance independently, then compare returns, risk, and risk-adjusted metrics. The goal is not to declare a "winner" but to illustrate how different strategies interact with different market conditions and how their combined use creates a more robust approach.

Market Setup

Market: "Will Candidate A win the presidential election?"

Resolution date: November 5

Starting price: 0.52 (market views the race as a toss-up with slight edge to Candidate A)

Simulation period: June 1 through November 5 (approximately 158 days, 22 weeks)

Starting capital per strategy: $10,000

Transaction costs: 2 cents per contract

The Simulated Price Path

The price path is designed to reflect realistic election dynamics:

Resolution: Candidate A wins (contract settles at 1.00).

Strategy 1: Fundamental Analysis

Approach

The fundamental trader maintains a probability model updated weekly. The model combines: - Weighted polling average (60% weight) - Economic fundamentals index (25% weight) - Historical base rate for incumbent party (15% weight)

Model Estimates vs. Market Price

Trade Log

Note: Contract counts are simplified for illustration. In practice, position sizing would follow Kelly criteria.

Fundamental Strategy Results

• Total P&L: +$216.10 (before transaction costs)
• Transaction costs: 6 trades x 2 sides x avg 155 contracts x $0.02 = ~$37.20
• Net P&L: +$178.90
• Return on capital: 1.79%
• Number of trades: 6
• Win rate: 4/6 = 66.7%
• Maximum drawdown: -$10.00 (during scandal weeks with NO position losses)
• Sharpe ratio: 1.42 (annualized, estimated from weekly returns)

Analysis

The fundamental strategy performed well overall, with its biggest wins coming from the two positions during the scandal period (Weeks 20-21). The model correctly identified that the scandal was being overpriced by the market and held firm in its estimate. The two losing trades (NO positions) occurred when the model's estimate was close to the market price, resulting in small losses. The key strength was conviction during the scandal, where the model estimate provided the confidence to buy when the market was panicking.

Strategy 2: Event-Driven Trading

Approach

The event-driven trader identifies scheduled catalysts and positions around them. Key events in the calendar: - Week 4: Expected endorsement - Weeks 7-8: Convention A - Weeks 10-11: Convention B - Week 17: First debate - Week 20: (Unscheduled) Scandal breaks

Event Calendar and Actions

Event 1: Endorsement (Week 4) - Pre-event estimate: Endorsements historically move markets 1-3 cents - Position: Buy YES at 0.53 (Wednesday before expected Friday announcement) - Outcome: Price moved from 0.53 to 0.54. Sold at 0.54. - Profit: +$0.01 x 200 contracts = +$2.00

Event 2: Convention A (Weeks 7-8) - Pre-event estimate: Convention bounce typically 3-5 cents, fades 50% within 2 weeks - Position: Buy YES at 0.55 on Monday of convention week - Sold 50% at 0.62 (peak of convention bounce) - Sold remaining 50% at 0.60 (beginning of fade) - Profit: 100 x ($0.07) + 100 x ($0.05) = +$12.00

Event 3: Convention B (Weeks 10-11) - Pre-event estimate: B's convention will narrow the gap by 3-4 cents - Position: Buy NO (sell A) at 0.40 before B's convention - Sold NO at 0.45 (A drops to 0.55) - Profit: 150 x ($0.05) = +$7.50

Event 4: First Debate (Week 17) - Pre-event estimate: 60% chance A wins debate; expected market impact +5 cents for A win, -3 cents for A loss - Expected value of YES: 0.60 x (+0.05) + 0.40 x (-0.03) = +$0.018 per contract - Position: Small YES buy at 0.56 (100 contracts) - Outcome: A wins debate, price jumps to 0.63 - Sold at 0.63 - Profit: 100 x ($0.07) = +$7.00

Event 5: Scandal (Week 20) - Not in calendar (unscheduled). Event-driven trader reacts post-event. - Assessment: Scandals in October historically fade. Similar scandals in past elections had 5-8 cent permanent impact, not 15 cents. - Position: Buy YES at 0.48 after initial reaction (200 contracts) - Held to resolution - Profit: 200 x ($0.52) = +$104.00

Event-Driven Strategy Results

• Total P&L: +$132.50 (before transaction costs)
• Transaction costs: 5 events x avg 2 trades x avg 150 contracts x $0.02 = ~$30.00
• Net P&L: +$102.50
• Return on capital: 1.03%
• Number of trades: 9 (across 5 events)
• Win rate: 5/5 = 100% (by event)
• Maximum drawdown: -$6.00 (during scandal when YES position temporarily declined)
• Sharpe ratio: 1.85 (annualized)

Analysis

The event-driven strategy achieved a perfect win rate across events, though the individual trade sizes were generally smaller than the fundamental approach. The trader captured value from each known catalyst and responded quickly to the unscheduled scandal. The high Sharpe ratio reflects the strategy's ability to identify asymmetric opportunities around events. Notably, the strategy was inactive during quiet periods (weeks without catalysts), preserving capital.

Strategy 3: Mean Reversion

Approach

The mean-reversion trader uses a 20-period moving average and 2-standard-deviation Bollinger Band analog. Signals fire when the price moves outside the bands on low-to-moderate volume. The trader holds for a target of reversion to the moving average, with a time limit of 2 weeks.

Signal Log

Detailed Trade Outcomes

Signal 1 (Week 8, Convention Bounce): - Convention drove price to 0.62, exceeding the upper band at 0.60 - Volume was elevated but driven by speculative "convention excitement" - Bought NO at 0.38 (equivalent to selling YES at 0.62) - Price reverted to 0.60 within 1 week; exited at 0.40 (YES at 0.60) - P&L: 130 contracts x $0.02 = +$2.60 - Small win; the reversion was partial and slow

Signal 2 (Week 11): - Z-score of -1.8 did not exceed the 2.0 threshold. No trade.

Signal 3 (Week 17, Post-Debate Spike): - Debate win drove price to 0.63, exceeding the upper band at 0.60 - Bought NO at 0.37 - Price reverted to 0.61 within 1 week; exited at 0.39 - P&L: 140 contracts x $0.02 = +$2.80

Signal 4 (Week 20, Scandal Drop): - Scandal drove price from 0.59 to 0.48, well below the lower band at 0.50 - Volume was extremely high --- this signal should have been filtered as information-driven - Without volume filter: Bought YES at 0.48. Price continued down to 0.45 before recovering. - Exited at 0.50 (2-week time limit, partial reversion) - P&L: 160 contracts x $0.02 = +$3.20 - This trade was risky. The high volume should have been a warning.

Signal 5 (Week 21, Continued Scandal): - Price at 0.45, Z-score -3.2. Extremely oversold. - Volume starting to decline (scandal becoming old news) - Bought YES at 0.45. Price reverted to 0.50 within 1 week. - P&L: 170 contracts x $0.05 = +$8.50

Mean Reversion Strategy Results

• Total P&L: +$17.10 (before transaction costs)
• Transaction costs: 4 trades x 2 sides x avg 150 contracts x $0.02 = ~$24.00
• Net P&L: -$6.90
• Return on capital: -0.07%
• Number of trades: 4
• Win rate: 4/4 = 100%
• Maximum drawdown: -$4.80 (during Signal 4 when price moved against before reverting)
• Sharpe ratio: -0.15 (annualized)

Analysis

The mean reversion strategy won every trade but still lost money due to transaction costs. The edges were small (2-5 cents per contract) and could not overcome the 4 cents of round-trip transaction costs. This is a critical lesson: mean reversion in prediction markets works best when transaction costs are very low or the overreactions are large. The strategy was correct directionally but the profit per trade was insufficient. Had transaction costs been 0.5 cents per contract instead of 2 cents, the net P&L would have been +$11.10.

Comparative Analysis

Performance Summary

Key Insights

1. The fundamental strategy captured the most total profit because it was willing to hold positions for extended periods, allowing large directional moves to compound. Its biggest wins came from the scandal period, where the model provided the conviction to buy when the market was fearful.

2. The event-driven strategy had the best risk-adjusted return (highest Sharpe ratio) because it only deployed capital around identifiable catalysts with favorable expected value. Its capital was mostly in cash, reducing exposure to random fluctuations.

3. The mean reversion strategy was unprofitable despite winning every trade. Transaction costs consumed the small edges. This strategy needs either lower costs, larger overreactions, or more frequent signals to be viable.

4. The scandal was the defining event. All three strategies responded to the Week 20-21 scandal, but differently: - Fundamental: Bought aggressively because the model said the market was wrong (+$203 from scandal trades) - Event-driven: Bought based on historical analysis of scandal impacts (+$104) - Mean reversion: Bought because the Z-score was extreme (+$11.70, but net of costs only +$5.70)

5. Strategy correlation was moderate. The fundamental and event-driven strategies were positively correlated (both bought during the scandal) but diverged during calm periods (fundamental was invested; event-driven was in cash). Mean reversion had lower correlation with both.

If Strategies Were Combined

A trader using all three strategies with equal capital allocation ($3,333 each) would have achieved:

• Combined Net P&L: +$91.50 ($59.63 + $34.17 + $-2.30)
• Combined Return: +0.92%
• Combined Sharpe Ratio: approximately 1.45 (improved diversification)
• Maximum Drawdown: approximately -$6.50 (reduced vs. fundamental alone)

The combined approach would have smoothed returns and reduced drawdowns compared to any single strategy, at the cost of somewhat lower total return than the best individual strategy.

Lessons Learned

1. Strategy selection depends on the fee environment. In markets with high transaction costs, strategies with small edges (mean reversion) are not viable. Focus on strategies with larger per-trade edges (fundamental, event-driven).

2. Conviction matters more than frequency. The fundamental strategy made fewer trades but made them with higher conviction and larger size, leading to the highest total return.

3. Event-driven strategies excel at capital efficiency. By deploying capital only around catalysts, the event-driven trader achieved the highest risk-adjusted return.

4. Mean reversion is a supporting strategy, not a standalone approach (at least in markets with significant transaction costs). It works best as a timing tool within a broader framework.

5. The biggest profits come from the biggest dislocations. All three strategies earned the majority of their profits during the scandal period, when the market overreacted most severely. The traders who had both the model (fundamental) and the framework (event-driven) to act decisively during the dislocation earned the most.

6. Combining strategies reduces risk. Even though the combined approach earned less than the best individual strategy, it had lower drawdowns and more consistent returns.

Code Reference

See code/case-study-code.py for the complete simulation, including price path generation, strategy implementations, and comparative analysis.