Research Report: M5 Remote-Plane Validation — Kalshi KXBTC15M

Short Disclaimer

Historical simulation only. No guarantee of future performance. This report examines a specific strategy family and found no reliable edge.

Intro / Thesis

This validation tested whether multi-horizon BTC trend confirmation could predict repricing in Kalshi's KXBTC15M market (15-minute BTC price contracts). The idea: when both 5-minute and 1-hour BTC returns are positive, buy YES; when both are negative, buy NO. Only enter at prices between $0.25 and $0.75, with spreads at or below $0.03, and more than 2 minutes remaining before expiry.

The hypothesis was that trend agreement across short and medium timeframes would capture momentum-driven repricing before contract settlement. After running 100 variants, the answer is clear: it does not.

Variant and Strategy Explanation

The base strategy ran on a 30-second loop, monitoring Coinbase BTC-USD data with 10-second refreshes. Entry conditions required:

• 5m and 1h BTC returns both positive → buy YES
• 5m and 1h BTC returns both negative → buy NO
• Position sizing: 1 contract per entry
• Risk controls: max position of 100 contracts, exit at -$25 unrealized PnL, flatten with 2 minutes remaining

The parameter sweep varied two axes:

This produced a 5×5 grid of 25 cells. All 25 cells completed successfully. Each successful variant is saved as a runnable Turbine strategy.

Important: the parameter sweep was effectively inert. Varying either entry_low or entry_high produced identical trades in every cell. The strategy never actually entered positions at prices outside of whatever narrow band the market happened to trade in during the test window. The flatness across the sweep does not represent robustness — it represents parameters that were never exercised.

Top Results

Across all 100 completed variants, the results are uniformly negative. Here are the top 8 by rank:

The "best" variant lost $12.99 with a Sharpe of -0.10. Win rate was slightly above 50% (51.3%), but that was not enough to overcome fees and spread costs. The spread cap made no material difference across the top cohort.

These results are consistent with selection noise. The deflated Sharpe ratio is 0.29, well below the 0.95 threshold needed to distinguish the winner from the luckiest result among skill-less trials. The permutation test p-value is 0.003, which seems interesting at first glance, but this only tells us that the edge feed's timing was not randomized — not that the strategy has a positive edge. The actual returns are negative in every single variant tested.

Bottom Results

The worst-performing variant:

The bottom variant lost $13.19 with a 46.6% win rate and a much larger drawdown of -$44.18. Notably, the spread cap of 0.01 actually reduced trade count (3,845 vs 4,158) while still losing more money. The tighter spread requirement didn't filter for better trades — it just filtered for fewer trades.

The range between best and worst variant is narrow: -$12.99 to -$13.19. This confirms the parameter sweep was largely decorative. Whether you entered at $0.25–$0.55 or $0.25–$0.65, whether you required a $0.01 or $0.04 spread, the outcome was the same: small, consistent losses.

Conclusion

The multi-horizon BTC trend confirmation thesis fails validation on Kalshi KXBTC15M. Every one of the 100 variants tested lost money after fees. The best variant lost 13% ROI with a negative Sharpe and a drawdown more than double its total loss.

The strategy's core problem is structural: a 51.3% win rate on a binary contract priced around $0.50 is not sufficient to overcome the bid-ask spread and fees when each win and loss is roughly symmetric. The edge simply isn't there.

The robustness statistics support this conclusion. The deflated Sharpe of 0.29 is far too low to claim the winner is anything other than the luckiest draw from a set of equally ineffective configurations. The parameter axes were inert — the strategy traded identically regardless of entry band settings, meaning we never actually tested different trading behaviors, just the same behavior labeled differently.

No variant should be considered strong, validated, or promising. The thesis is rejected.

Long Disclaimer

This report is a historical simulation research document. It describes the results of running a specific strategy configuration against historical market data on the Kalshi KXBTC15M market. Nothing in this report should be interpreted as investment advice, a recommendation to trade any financial instrument, or a prediction of future performance.

Historical simulation has significant limitations:

1. Survivorship and data quality: The data used may not perfectly reflect actual market conditions, including order book depth, execution latency, or price feed discrepancies.
2. Slippage and fees: The simulation includes certain fee assumptions but may not capture the full cost of execution, especially during volatile periods.
3. Overfitting risk: With 100 variants tested against the same dataset, the probability of finding spurious "good" results by chance is high. The robustness statistics in this report indicate that the results are consistent with selection noise.
4. Regime dependence: BTC and crypto market behavior changes over time. A strategy that loses money in one period could theoretically perform differently in another — but there is no evidence in this data to suggest this one would.
5. The permutation test cited in this report shuffles only the edge data (BTC price changes), not the market price series. It tests whether the timing of the BTC feed relative to trades was meaningful, not whether the overall strategy has predictive power.

The negative results reported here are internally consistent across all 100 variants and both the robustness sweep and the permutation test. The conclusion that this strategy family has no edge on this market in this window is robust to the limitations of the methodology.

Always do your own research. Past performance, whether simulated or real, does not guarantee future results.