Short Disclaimer

This report is a historical simulation exercise only. Nothing in it constitutes trading advice or a forecast of future returns. All figures are gross of potential real-world costs not captured in the backtest, and past simulated performance does not guarantee future outcomes.

Intro / Thesis

We examined whether a simple intraday Bitcoin momentum signal — specifically, whether the current Coinbase spot price sits above or below its 1‑hour VWAP, combined with a 5‑minute price change threshold — could forecast the direction of Kalshi’s 15‑minute BTC binary option (series ticker KXBTC15M). The goal was to identify a practical rule that buys “yes” or “no” contracts when short‑term momentum aligns with a moderate price regime and controlled spread.

Because the signal is rudimentary, we expected its standalone edge to be thin. What we wanted to learn from a full parameter sweep was whether any combination of entry price boundaries (floor and ceiling) could tilt the risk/reward enough to produce a positive‑expectation system, or whether the results are indistinguishable from noise.

Variant and Strategy Explanation

The base strategy operates on a 30‑second loop. On each tick, it fetches Bitcoin price data from Coinbase (spot price, 1‑hour VWAP, and 5‑minute percent change) and evaluates two entry rules:

• Buy Yes when spot > 1h VWAP and 5‑minute change is positive above a small threshold, provided the contract price is within a configurable floor/ceiling range, the spread is ≤ 3¢, and more than 2 minutes remain until expiry.
• Buy No when spot < 1h VWAP and 5‑minute change is negative below the mirror threshold, with the same price, spread, and time constraints.

Every entry uses a fixed size of 1 contract per signal. The strategy also carries a $25 unrealized PnL stop‑loss and an automatic exit when time‑to‑expiry drops to 2 minutes or less, to avoid holding through settlement noise.

Our sweep varied two parameters across 100 combinations:

• price_floor: 0.05 to 0.45 (10 levels)
• price_ceiling: 0.55 to 0.95 (10 levels)

Every variant is identical in signal logic, risk limits, and exit rules; only the acceptable entry price band changes. This lets us isolate the effect of paying too much (or too little) for directional exposure. Each successful variant is saved as a runnable Turbine strategy, meaning a trader could deploy any of these configurations in live simulation on the platform.

Top Results

The best‑performing variant in the sweep was the combination of floor = $0.45 and ceiling = $0.55 — effectively a very tight price window that only trades contracts priced near the center of the range. Its headline statistics:

• ROI: 122.3%
• Total PnL: +$12.23
• Trades: 4,226
• Win rate: 12.5%
• Max drawdown: –$17.26
• Sharpe (ex‑post): 0.14

The next several ranks share a common ceiling of $0.55 but relax the floor to lower values ($0.27, $0.36, $0.23, etc.). They produce similar total PnL in the +$9.48 to +$10.44 range, with win rates clustered tightly around 12.2‑12.3% and max drawdowns between –$26.57 and –$29.48. Essentially, allowing cheaper entry prices produces marginally more trades and similar gross PnL, but with larger realized drawdowns.

A critical structural detail across all top variants: fees consume approximately 91% of the winner’s gross PnL, according to the simulation’s fee‑drag check. That means the economic edge before trading costs is substantial, but after fees the net PnL is thin and fragile.

Even more important, the deflated Sharpe ratio is 0.50, which falls well below the 0.95 threshold commonly used to suggest that a single top result is statistically distinguishable from the luckiest trial in a random parameter sweep. In plain terms: given 100 variants tested, a result this good (or better) would emerge from random data roughly as often as it appeared here. The top results are consistent with selection noise and overfitting, and we cannot call any variant strong, validated, or promising based on this sweep alone.

Bottom Results

The worst‑performing configurations uniformly pushed the price ceiling too high. Variants with a floor of $0.41 and ceilings from $0.73 to $0.95 produced negative total PnL ranging from –$17.99 to –$19.14, with corresponding ROI of –180% to –191%. Drawdowns in this group reached –$45.77.

What unites these failures is buying contracts near the upper end of the range ($0.73‑$0.95) where the implied probability is already elevated. The momentum signal was not strong enough to overcome the poor risk‑reward of entering when the contract was expensive. Win rates remained statistically indistinguishable from the better variants (still near 11.8‑12.1%), but the magnitude of losses on unsuccessful trades overwhelmed any small edge.

The lesson from the bottom of the sweep is straightforward: the strategy breaks when the price ceiling exceeds roughly $0.64. Above that point, every ceiling tested produced negative mean PnL, and no floor value could rescue the results.

Conclusion

The Coinbase VWAP momentum concept shows a narrow operating window that can produce a modest positive gross edge in backtesting: keep the entry price band tight (ceiling ≤ $0.55‑$0.59, floor in the $0.23‑$0.45 range) and accept that the win rate will hover around 12‑13%. Outside that band, the strategy bleeds money quickly.

However, the results do not survive standard overfitting diagnostics. The deflated Sharpe of 0.50 is in the same neighborhood as the expected maximum Sharpe from a pure noise sweep (0.14), and the margin between the winning variant and its nearest neighbors is thin. Fee drag consumes nearly all the gross edge, leaving net PnL that is fragile to even minor changes in execution or market conditions.

This is not a statement that the signal has no value — only that 100 variants on two parameters are insufficient to identify a reliable configuration. A trader interested in this family would need to test on out‑of‑sample periods, examine different market regimes, and determine whether the modest positive skew persists before deploying any variant with real capital. As it stands, the top variants are best treated as interesting starting points for further research, not as validated strategies.

Long Disclaimer

This report is produced solely for research and educational purposes within a simulated trading environment. All performance figures — ROI, Sharpe ratio, maximum drawdown, total PnL, and win rate — are derived from historical backtests that do not account for real‑world factors such as liquidity constraints, slippage, exchange downtime, API latency, or changes in fee schedules. Backtested results are inherently optimistic and often overstate achievable net returns.

The parameter sweep described here tested 100 variants on the same historical dataset. Statistical diagnostics (including a deflated Sharpe ratio and neighborhood degradation analysis) indicate that the top‑ranked results are consistent with selection noise and overfitting. No variant in this report should be interpreted as validated, robust, or likely to produce similar results in live markets.

The mention of saved runnable Turbine strategies refers to platform functionality for historical simulation replay and configuration management. It does not imply endorsement, recommendation, or a promise of future profitability. Trading binary options on Kalshi or any other venue involves risk of total loss. Past simulated performance is not indicative of future results. Consult a qualified financial professional before engaging in any trading activity.