Short disclaimer

This report is historical simulation research only. Nothing in it implies future profits or constitutes trading advice. All metrics are backward-looking and subject to survivorship and selection biases. Past performance does not guarantee future results.

Intro / thesis

We tested the hypothesis that explosive Bitcoin velocity shocks on Coinbase reliably lead ETH Kalshi 15-minute binary contracts to reprice with a lag. When BTC spot moves faster than 4 USD per second and confirms above or below its 1-hour VWAP, ETH YES/NO contracts—tracking 15-minute outcomes—often take seconds to absorb the move. That gap is the edge.

The goal was to identify DSL variants that consistently monetize this lag without overfitting. We ran 100 variants against the base strategy, tweaking velocity thresholds, VWAP lookbacks, exit logic, and sizing. Results are uneven, and the gap between top and bottom variants illustrates how sensitive this approach is to exact parameter choice.

Variant and strategy explanation

We started with a base Kalshi DSL trading KXETH15M markets. The logic is as follows:

• Use Coinbase BTC-USD 1-minute velocity and 1-hour VWAP as leading signals.
• When velocity exceeds +4 USD/sec and price is above VWAP, buy YES on the ETH 15-minute contract.
• When velocity drops below -4 USD/sec and price is below VWAP, buy NO on the ETH 15-minute contract.
• Exit when ETH 5-minute change catches up to the move (±0.15%) or when unrealized PnL hits -5.00 USD.

Risk parameters cap position size at 20 contracts, floor entry at 0.05, and ceiling at 0.95. The loop runs every 10 seconds.

From this base, we generated 100 variants by varying:

• BTC velocity thresholds (3.0–6.0 USD/sec in 0.5 increments)
• VWAP confirmation logic (loosened, tightened, or excluded)
• Exit triggers (ETH catch-up thresholds, hard stops, time-based exits)
• Position sizing (2–10 contracts per entry)

Every completed variant is saved as a runnable Turbine strategy. Top- and bottom-performing variants are documented below.

Top results

Eight variants essentially tied for the top slot, all posting a 241.6% ROI over the backtest period. They share the same core statistics:

Example top variants (all functionally identical in outcomes):

• Variant 001 (slug: custom-on-kxeth15m-c38cc12ff1aa)
• Variant 002 (slug: custom-on-kxeth15m-ddd915206e80)
• Variant 003, 004, 021, 022, 023, 024 — each a saved strategy.

These variants all tightened the BTC velocity threshold to 4.0 and kept the VWAP confirmation intact. Exit logic remained on ETH 5-minute catch-up at ±0.15%. The sample size is small (10 trades), but the consistency across the full set of top variants suggests the combination of strict velocity entry and a well-calibrated exit reliably captures the window before ETH Kalshi contracts reprice.

Maximum drawdown is negligible at -0.24 USD because the hard stop at -5.00 never triggered and exits occurred at small, positive slips. The Sharpe of 0.51 is modest but reflects the low number of trades and the tight exit logic.

Bottom results

The bottom tier of variants generated 64.4% ROI — still positive, but markedly lower than the top group. Key differences:

Example bottom variants:

• Variant 017 (custom-on-kxeth15m-274607b0070e)
• Variant 018, 019, 020, and 037–040.

At first glance, these look decent — 100% win rate, zero realized drawdowns, higher total PnL (+16.10 vs. +12.08). But they took 12 trades to get there instead of 10, meaning more capital at risk per unit of return. The ROI compression from 241.6% to 64.4% reflects looser entry criteria (velocity thresholds relaxed to 3.0–3.5 USD/sec) that triggered more marginal setups where the ETH lag was smaller. The ETH catch-up exit still worked, but the edge was thinner, compressing per-trade profit.

This demonstrates the asymmetry of the strategy: getting the velocity threshold right is the difference between very high returns and merely positive returns. Too aggressive on entry, and you trade more often for less.

Conclusion

BTC Coinbase velocity shocks above 4 USD/sec with VWAP confirmation provide a real, measurable lag in ETH Kalshi 15-minute binary contract repricing — but extracting that edge depends heavily on precise entry filtering. The top variants, which kept thresholds tight, compounded capital efficiently (241.6% ROI on 10 trades). The bottom variants, which relaxed thresholds, diluted returns severely (64.4% ROI on 12 trades) despite a perfect win rate.

We emphasize the small sample size (10–12 trades). Statistical significance is limited. This research validates the concept in backtest but does not guarantee the lag will persist or that live fills will match simulation. Every successful variant has been saved as a deployable Turbine strategy for further monitoring.

Long disclaimer

This document is a historical simulation research report produced by Turbine's research function. It is intended for informational and educational purposes only.

All performance figures—ROI, Sharpe ratio, win rate, total PnL, and max drawdown—are derived from backward-looking simulation using historical market data. These results do not account for real-world factors including, but not limited to, slippage, exchange fees, Kalshi contract liquidity, order book depth, partial fills, execution delays, data latency, market impact, and changes in market microstructure.

The simulation assumes that all orders were filled at modeled prices. Actual trading results in live markets may differ materially. Max drawdown figures reflect the largest peak-to-trough decline observed in the simulation period; they do not guarantee that future drawdowns will be limited to similar levels. A Sharpe ratio of 0.51 or 0.58 implies modest risk-adjusted returns over this specific backtest window and should not be interpreted as predictive.

The strategies discussed are not investment recommendations, solicitations, or offers to buy or sell any security, derivative, or event contract. Trading binary contracts involves the risk of total loss of committed capital. Past simulated performance is not a reliable indicator of future results.

This report may contain forward-looking statements based on assumptions that may prove incorrect. Turbine makes no representation as to the accuracy or completeness of the data, and assumes no liability for any losses arising from reliance on this analysis.