Spread Convergence

Mean-reversion trading between two correlated prediction markets.

Pro Feature. Requires a Pro or Ultra subscription. Get started at api.mathematicalcompany.com

Spread Convergence

Track the price spread between two correlated markets. Enter when the spread deviates beyond a z-score threshold, exit when it reverts to the mean.

How It Works

1. Compute spread = price_a - price_b each cycle
2. Maintain a rolling window of spread values
3. Compute z-score of current spread vs historical mean/std
4. Enter when |z| > entry_zscore, exit when |z| < exit_zscore

Spread history: mean=0.02, std=0.01
Current spread: 0.05 → z-score = 3.0
Signal: long_b_short_a (spread too high, expect reversion)

Pipeline: spread_convergence

scanner = hz.spread_convergence(
    market_a="btc-100k",
    market_b="eth-5k",
    feed_a="btc_feed",
    feed_b="eth_feed",
    lookback=100,
    entry_zscore=2.0,
    exit_zscore=0.5,
    max_spread=0.20,
    size=10.0,
    auto_execute=False,
    cooldown=30.0,
)

hz.run(pipeline=[scanner], ...)

Parameter	Type	Default	Description
market_a	str	required	First market ID
market_b	str	required	Second market ID
feed_a	str	required	Feed for market A
feed_b	str	required	Feed for market B
lookback	int	100	Rolling window size
entry_zscore	float	2.0	Z-score to enter
exit_zscore	float	0.5	Z-score to exit
max_spread	float	0.20	Max absolute spread to consider valid
size	float	10.0	Trade size per leg
auto_execute	bool	False	Auto-execute signals
cooldown	float	30.0	Seconds between trades

SpreadSignal

Stored in ctx.params["last_spread_signal"].

Field	Type	Description
market_a	str	First market
market_b	str	Second market
spread	float	Current spread value
spread_mean	float	Rolling mean
spread_std	float	Rolling standard deviation
zscore	float	Current z-score
signal	str	"long_a_short_b", "long_b_short_a", "exit", or "hold"

Signal Logic

Condition	Position	Signal
z > entry_zscore	flat	long_b_short_a
z < -entry_zscore	flat	long_a_short_b
`	z	< exit_zscore`	positioned	exit
otherwise	any	hold

Example

import horizon as hz

# Spread trade between two correlated political markets
spread = hz.spread_convergence(
    market_a="gop-senate",
    market_b="gop-house",
    feed_a="gop_senate_feed",
    feed_b="gop_house_feed",
    lookback=200,
    entry_zscore=2.5,
    exit_zscore=0.3,
    auto_execute=True,
)

hz.run(
    name="spread_trader",
    exchanges=[hz.Polymarket(private_key="0x...")],
    markets=["gop-senate", "gop-house"],
    feeds={
        "gop_senate_feed": hz.PolymarketBook("gop-senate"),
        "gop_house_feed": hz.PolymarketBook("gop-house"),
    },
    pipeline=[spread],
    interval=1.0,
)

For more rigorous pairs trading with cointegration testing and half-life filtering, see Statistical Arb.