Strategy · High risk · Intermediate

Momentum

Ride trends in prediction-market prices. Simple, directional, and risky — the control variable is discipline, not insight.

Published Apr 10, 2026

The idea

Prices that have risen in the last hour tend to rise in the next hour. The effect is weak but persistent in prediction markets, especially news-driven ones. Momentum captures it: buy what is rising, sell what is falling, exit on reversal.

This is the riskiest polybot strategy. Expect drawdowns, and size accordingly.

How polybot implements it

Three signals, all normalised to [-1, 1]:

• Price momentum. z-scored change over a lookback window (default 1h).
• Volume confirmation. Recent volume vs. rolling average.
• Order-book imbalance. Bid depth vs. ask depth at top-5.

The composite score gates entries. Exit is volatility-scaled stop plus a time-in-trade cap.

score = 0.5 * price_z + 0.3 * volume_z + 0.2 * imbalance_z
if score > self.config.entry and self.inventory[m.id] == 0:
    self.emit(Signal(m.id, side="YES", size=self.config.size))
elif score < -self.config.entry:
    self.emit(Signal(m.id, side="NO", size=self.config.size))

Where it works

• News-heavy windows. Pre-election weekends, Fed days, earnings weeks — volume and price both move.
• Fresh listings. New markets reprice aggressively in their first 24–48 hours. Momentum captures the early trend.

Where it fails

• Mean-reverting markets. Spread-farming regimes produce zero autocorrelation. Your signal is noise and your fees are real.
• Gap risk. A news release during low-volume hours gaps the price; your stop slips. polybot uses synthetic stops (triggering orders at the top of book, not stops at exchange) to minimise but not eliminate this.
• Overfitting. Every backtest is kind to momentum parameters. Shadow live. Demand 60+ days of consistent shadow P&L before promoting.

Configuration

polybot strategy enable momentum
polybot strategy config momentum \
    --lookback-minutes 60 \
    --entry-score 0.6 \
    --exit-score 0.2 \
    --volatility-stop-mult 2.5 \
    --max-hold-hours 6 \
    --max-concurrent 5 \
    --daily-loss-stop-usd 200

Daily loss stop is the important one — momentum drawdowns are steep and fast.

FAQ

How do I avoid over-trading? --max-concurrent caps open positions. --cooldown-minutes prevents re-entering a market you just exited. Both matter.

Is this worth running at all? Honestly — for most operators, momentum belongs in shadow mode as a learning exercise. Start with arbitrage and spread farming; run momentum on 5% of capital after months of shadow data.

Source: src/polybot/strategies/momentum.py.