Research Report: Kalshi ETH Mean Reversion Exploration

Short Disclaimer

This report summarizes historical simulation research. Past simulated performance does not guarantee future results. All figures represent hypothetical outcomes derived from backtesting.

Intro / Thesis

We ran 100 systematic variants on Kalshi’s 15‑minute Ethereum binary option markets (KXETH15M) under a mean‑reversion framework. The central idea was that these short‑duration markets occasionally overreact early in their 15‑minute life, and you can exploit that by buying when the price dips into a defined entry band and exiting once the price moves back inside that same band or reaches a risk floor. This is not a directional ETH call — it is purely a trade on intra‑market price behavior.

The simulation loop checked every 60 seconds. A position was entered only when the market price traded inside an entry band, held until it either moved above that band’s upper bound or hit a hard price floor, and was subject to a 30‑second cooldown after every close to avoid whipsaw re‑entries.

Below I walk through what worked, what failed, and the patterns that emerged across all 100 variations.

Variant and Strategy Explanation

Every variant inherited the same core structure from the base DSL but altered a small set of parameters:

• Market: KXETH15M — 15‑minute binary options resolving to an ETH price condition.
• Loop interval: 60 seconds (the strategy evaluated entry/exit once per minute).
• Position limits: Max 10 contracts per entry, hard price floor between 0.05 and 0.20, price ceiling stayed at 0.85.
• Entry band: Lower bound ranged from 0.20 to 0.40; upper bound from 0.45 to 0.70 depending on the variant.
• Cooldown: Fixed at 30 seconds across all variants.
• Exit logic: The position was closed when the market price moved above the entry band’s upper edge. If the price instead fell to the hard floor, the position was exited at a loss.

The key tension is between “buying early at a discount” (lower entry bands, lower floors) and “avoiding catching a falling knife” (higher floors, tighter entry ranges). Each completed variant is saved as a runnable Turbine strategy.

Top Results

The highest‑performing variants converged on a consistent theme: buy low, cut losses quickly, and don’t be greedy on the exit.

What stands out:

• The top eight strategies all share a 0.55 upper band. Exiting at 0.55 consistently outperformed higher exits — momentum fades quickly in these 15‑minute windows.
• The sweet spot for the lower entry band is 0.20–0.35. Buying below 0.20 would have been too rare to matter, while buying above 0.35 crept into “late to the party” territory.
• The price floor parameter (0.05 vs 0.10 vs 0.15) had minimal impact on the top performers. Max drawdowns were identical across floors for several bands, suggesting that once you’re in the right entry band, the floor mainly serves as catastrophic protection rather than a frequent exit trigger.
• Win rates are modest — none cracked 50%. The edge comes from a handful of large winners outweighing smaller, controlled losses. Sharpe ratios near 1.0 confirm the strategy is generating returns proportional to the risk taken.
• Tying the top two entries, the variant with a 0.15 floor matched the 0.10 floor point‑for‑point. Practically speaking, the 0.15 floor offers slightly less overnight gap risk and may be a safer default for live deployment (though that is not a forward‑looking recommendation).

Bottom Results

Failures were equally instructive. The worst variants almost without exception had entry bands shifted too high and combined them with floors that were set too conservatively high (0.20).

What broke:

• High entry bands (>0.35 low bound): You were buying after the easy reversion move had already happened. The market wasn’t “cheap” anymore; it was fairly priced, and you’d get trapped when it faded further.
• High exit bands (>0.60): Waiting for a 0.65 or 0.70 exit meant holding through noise and eventual mean reversion failure. The 15‑minute clock works against you here.
• 0.20 price floor: Setting a hard floor at 0.20 sounds prudent, but in practice it forced exits on positions that might have recovered within the remaining window. The market routinely traded below 0.20 intra‑window, only to rally. A 0.20 floor plus a high entry band was a recipe for buying high and selling low.
• Zero‑win variant: The bottom‑ranked strategy with band 0.25‑0.65 and floor 0.20 notched a goose egg — zero wins across 20 trades. It only took 20 entries, meaning the conditions were rare but lethal when they triggered.

Conclusion

The KXETH15M market displays a reproducible mean‑reversion signature when you catch the price between roughly 0.20 and 0.35 and exit by the midpoint at 0.55. The most successful variants accepted sub‑50% win rates and relied on cutting losses quickly — either via a tight floor or by not entering too early. The losing variants illustrate the mirror error: buying too late, holding too long, and placing the safety net where it becomes a guillotine.

The top strategies are structurally simple and now exist as saved, runnable Turbine strategies for further review. This is a backward‑looking analysis only; it describes what worked in the simulated data, not what will work tomorrow. But the pattern is clear enough to warrant attention from anyone following volatile, short‑horizon binary markets.