Kalshi BTC 15‑Minute Market — Mean‑Reversion Fade Report

Short disclaimer: This report is historical simulation research. It does not predict future profits and is not financial advice.

Intro / Thesis

We tested whether fading extreme Kalshi prices for the 15‑minute BTC market (“KXBTC15M”) could produce a reliable edge. The core idea is simple: when Kalshi’s YES price falls below 0.35, buy YES; when it rises above 0.65, buy NO. But not all extremes are created equal. The thesis holds that mean‑reversion is most trustworthy when Coinbase BTC‑USD activity is calm, moving at a moderate clip, or showing a short‑term divergence — regimes where Kalshi overreaction lacks a strong underlying trend. By filtering out violent Coinbase moves (shock regimes) and tightening exits, we hoped to isolate a clean, repeatable fade. Across 100 parameter variants, the research examined which regime filter, entry band, and exit tightness produce the best risk‑adjusted results.

Variant and Strategy Explanation

All variants share the same structural skeleton:

• Market: Kalshi KXBTC15M (15‑minute resolution).
• Edge data: Coinbase BTC‑USD 5m, 15m, 1h, and 4h changes, plus 15m/1h high‑low ranges. Polled every 5 seconds.
• Risk controls: Max position 20 contracts; price floor 0.05, ceiling 0.95; max loss per variant $10.00.
• Loop: Evaluate rules every 10 seconds.
• Entry bands (fade thresholds): YES below 0.35, NO above 0.65.

Entry logic splits into five regime buckets, applied symmetrically for YES and NO fades:

1. Calm regime: Coinbase 5m change between –0.1% and +0.1%, 15m change between –0.25% and +0.25%.
2. Moderate up: Coinbase 5m change +0.1% to +0.35%, 15m change +0.25% to +0.75%.
3. Moderate down: Same ranges but negative.
4. Divergence (5m up, 15m down): 5m > 0, 15m < 0.
5. Divergence (5m down, 15m up): 5m < 0, 15m > 0.

Every entry size is a flat 5 contracts.

Exits are also uniform across the base DSL:

• Take‑profit: price crosses 0.50 (gated on position_size > 0 to avoid cross‑side fire).
• Continuation stop: 5m Coinbase move exceeds ±0.3% against the fade.
• Hard stop: unrealized PnL < –$5.00.
• Expiry exit: time to expiry < 2 minutes.

Variants differ in small tweaks to thresholds, exit levels, or the presence of specific sub‑rules, all derived from this base DSL. The 100 completed variants represent the full sweep of combinations.

Each successful variant listed below is saved as a runnable Turbine strategy under its unique slug.

Top Results

The “best” performers (ranked 1 to 8) are a study in contrast: they produced the least negative total PnL across the tested set, but every single one still lost money.

Key observations:

• Win rates sit just above 50% — effectively a coin flip after fees and spreads.
• Trade counts are enormous (14,000+), meaning the strategy fires constantly. Commission and slippage would only worsen real‑world results.
• Max drawdown is deep: –$1,440 on a $25 account, implying total ruin multiple times over.
• Sharpe ratios are negative or near zero. There is no risk‑adjusted edge.

In short, the top variants are the “least bad” — not profitable. The regime‑filtered fade did not generate positive expectancy.

Bottom Results

The bottom eight variants (ranks 93–100) trade far fewer contracts but lose proportionally more on the capital deployed.

Key observations:

• Win rate dips below 49%. Even with fewer trades, the strategy leaks capital steadily.
• ROI is uniformly catastrophic (–7,100% or worse).
• Max drawdown here is smaller in absolute dollars, but relative to the starting stake it still represents complete loss.

The bottom variants confirm there is no “hidden gem” configuration; the entire family of fades fails to produce a sustainable edge.

Conclusion

Fading Kalshi BTC 15‑minute extremes — even when filtered to calm, moderate, or divergent Coinbase regimes — did not yield a profitable strategy in historical simulation. Win rates hover near 50%, while commissions and spreads turn a breakeven expectancy into a grinding loss. Both the top and bottom variants lose money, differing only in how quickly and how loudly they do so. The regime logic itself may be sound in theory, but it is not sharp enough to overcome the structural costs of trading this market. Mean‑reversion trading in KXBTC15M, as tested, is not a viable standalone approach. Any edge would need to be far more precise, likely employing micro‑structural or order‑flow data beyond what was available in this study.

Long disclaimer: This document is a historical simulation research report prepared for analytical purposes only. It does not constitute an offer, solicitation, or recommendation to buy or sell any financial instrument, including prediction‑market contracts or cryptocurrencies. All performance metrics are derived from simulated backtests, which do not account for actual trading costs, liquidity constraints, execution slippage, exchange fees, or the impact of market‑moving events that occurred outside the test period. Past simulated performance does not guarantee future results. Trading in prediction markets and cryptocurrencies involves substantial risk of loss and may not be suitable for all participants. The author assumes no liability for any trading decisions made based on this report.