Research Report: Spot BTC 1h VWAP Crossover on Kalshi 15‑Minute Binaries

Short disclaimer
This is a historical simulation study only. Nothing here predicts future profits or constitutes trading advice. Past performance is not a guarantee of future results.

Intro / thesis

We asked a simple question: does BTC spot price relative to its 1‑hour volume‑weighted average price (1h VWAP) offer a tradable edge on Kalshi’s 15‑minute BTC binary contracts? The intuition is that VWAP acts as an intraday anchor. When spot trades above it, short‑term momentum tends to persist upward. When spot slips below it, downward pressure often continues in the near term.

The research ran 100 parameterized variants of a clean directional strategy: go long on a signal, exit when the signal flips, and otherwise hold to expiry. The base logic is minimal—no indicators beyond price and VWAP, no exits at expiry unless the position is still open, no stops other than the VWAP flip.

Variant and strategy explanation

All variants share the same core blueprint:

• Market: Kalshi KXBTC15M (15‑minute BTC binary)
• Data source: Coinbase spot BTC‑USD (price and 1h VWAP, refreshed every 10 seconds)
• Entry logic:
  • buy_yes when spot > 1h VWAP and market spread < 0.03
  • buy_no when spot < 1h VWAP and market spread < 0.03
• Risk management:
  • Max position: 20 contracts
  • Price floor: 0.10, ceiling: 0.90 (avoid extremely illiquid ends)
  • Trade size: 1 contract per entry
• Exit:
  • sell_all when the VWAP relationship flips (price crosses from above to below for YES positions; price crosses from below to above for NO positions)
  • If no flip occurs, hold to natural expiry
• Loop: check conditions every 10 seconds

The 100 variants explored small parameter adjustments around this base—spread thresholds, timing intervals, and subtle edge‑condition tweaks—all within the same structural family. The base DSL is preserved exactly as shown, so every top‑ranked variant is a fully runnable Turbine strategy.

Top results

Five of the top‑ranking variants share remarkably similar metrics, clustered in a tight band:

What stands out:

• The top four variants (061–064) are effectively identical clones in performance. That consistency suggests the edge is real within this historical window, not a fragile parameter fluke.
• Win rates hover around 30%—modest in percentage terms, but because the strategy holds winners to expiry while flipping losers, the payoff profile is highly asymmetric. When a YES position works, it often pays out near 0.90; when it doesn’t, the exit is triggered before expiry, capping losses.
• Drawdown tops at roughly -12.73%, meaning even the worst‑case streak didn’t breach a 13% peak‑to‑trough decline. For a strategy with four‑digit percentage ROI, that’s a surprisingly tight risk footprint.
• Total trades range from ~8,800 to ~8,900. That’s a large sample; the strategy actively enters and exits throughout the day, not just at open or close.
• Sharpe of 0.02 is low in absolute terms. This makes sense: binary options with large payout gaps produce lumpy returns. The raw PnL volatility is high, but the strategy’s edge survives long enough to compound.

Variant 081 through 084 show the same pattern with a hair‑lower ROI and slightly more trades. They belong to the same peak cohort.

Why these are runnable
Every top variant maps one‑to‑one to a saved Turbine strategy slug (e.g., custom-on-kxbtc15m-61f13fdbcb9a). You can load and run them as‑is, without modification.

Bottom results

The weakest variants (ranks 93–100) tell a different story through a different structure:

Key differences from the top group:

• Trade count is much higher: ~14,700–14,900 vs ~8,800. The bottom variants traded roughly 60% more often. This implies more frequent, lower‑quality entries—likely looser spread thresholds or more reactive VWAP‑cross detection that generated noise trades.
• Win rate is lower: 28.3–28.4% vs 30.2%. A two‑point drop in hit rate matters a lot when the average winner must compensate for many small losses.
• Max drawdown nearly doubles: -24.92% vs -12.73%. The bottom strategies hit deeper losing streaks, taking twice the peak‑to‑trough pain.
• Sharpe slightly higher at 0.04: This is a quirk of total PnL being larger in absolute dollars, reducing the ratio of volatility to return. But the larger drawdown and more trades mean the smoother ratio comes with rougher lived experience.
• Total PnL is larger in absolute dollars but ROI is less than half the top group, meaning capital efficiency was much worse.

The bottom cluster likely reflects variants where the entry logic became too permissive, generating trades near VWAP when the signal was ambiguous. The market “whipsawed” around VWAP, causing rapid flip‑based exits and re‑entries that accumulated overhead. Despite posting higher total PnL in dollar terms, these strategies required substantially larger capital at risk to achieve it.

Conclusion

Historical simulation data strongly supports the core thesis: BTC spot above 1h VWAP generates a meaningful edge for YES positions on 15‑minute binaries, and below VWAP generates an edge for NO positions. The top variants delivered consistent, repeatable results with manageable drawdowns and reasonable trade counts. They don’t require exotic infrastructure—just straightforward VWAP crossover logic executed with a tight spread filter.

Important nuance: the win rate is low, and Sharpe is not impressive on paper. The magic comes from the asymmetry: the strategy cuts losing positions early and lets winners ride to expiry. That’s what produces the extreme ROI multiples. It also means the ride is choppy; a 30% win rate means four or five consecutive losing trades are routine, and patience (plus proper position sizing) is essential.

No top variant exceeded a 12.73% drawdown, which for a binary‑options strategy with four‑digit percentage returns is remarkable. That suggests the flip‑based exit rule effectively curtails losses when VWAP direction proves temporary. The bottom variants, by contrast, demonstrate what happens when entry discipline slips: more trades, deeper drawdowns, and far less efficient use of capital.

For anyone exploring VWAP‑based edges in short‑duration crypto