Deep Research

Momentum v4 Robustness

The v3 dual-regime momentum strategy found a profitable cluster, but the permutation p-value of 0.292 means we cannot rule out luck across the 100-parameter sweep. This v4 sweep narrows the price band and varies polling frequency to test whether the cluster is stable in its neighborhood — prioritizing deflated Sharpe and low sensitivity to small parameter shifts over chasing a single high-PnL cell.

Historical research only. Not investment advice.

Top Strategy Variants

Bottom Strategy Variants

Kalshi BTC Momentum Strategy — Neighborhood Robustness Sweep

Strategy Family: KXBTC15M dual-regime momentum
Sweep: price floor × price ceiling (10×10 grid, 100 variants completed)
Date: Historical simulation only


1. Short disclaimer

This is a historical simulation research report. Past performance does not guarantee future results. Nothing in this report constitutes trading advice. All metrics are derived from backtests and are subject to overfitting, lookahead bias, and fill-assumption error. The strategies described may lose money in live trading.


2. Intro / thesis

The v3 dual-regime momentum strategy on KXBTC15M turned up a cluster of profitable cells, but the permutation p-value of 0.292 across the 100-parameter sweep meant we couldn’t rule out selection noise. Rather than celebrating a single high-PnL cell, this v4 sweep was designed as a focused neighborhood check: we fixed the ceiling at 0.55 (the v3 winner’s value) in most runs and varied the price floor in small steps, while also testing broader ceiling bands to measure how quickly performance degrades when we walk away from the original sweet spot.

The goal was simple: find configurations where PnL holds up across several adjacent parameter values (low neighborhood degradation) and produce a meaningfully positive deflated Sharpe — not just a hot cell that collapses one tick over.

The results are mixed. We found a band of profitable floor values near the v3 winner, but the deflated Sharpe remains modest, the winner sits on thin daily data, and fill assumptions at extremes are stretched. The top-performing cells are still consistent with a lucky draw from the sweep.


3. Variant and strategy explanation

Market traded: Kalshi’s KXBTC15M series — 15-minute binary options on Bitcoin above/below a strike. The strategy trades YES and NO contracts based on short-term BTC momentum signals, with hard risk boundaries on entry price.

What we varied (v4 sweep):

ParameterValues tested
risk.price_floor0.05, 0.09, 0.14, 0.18, 0.23, 0.27, 0.32, 0.36, 0.41, 0.45
risk.price_ceiling0.55, 0.59, 0.64, 0.68, 0.73, 0.77, 0.82, 0.86, 0.91, 0.95

All 100 combinations ran to completion. The base strategy logic was held constant: the same momentum entry rules, same exit rules (2-minute near-expiry sell, 15% take-profit, -25% stop-loss), same position sizing (25 contracts per trade, max 50), and the same 10-second loop interval with Coinbase BTC data refreshed every 5 seconds.

What a “variant” is: Each row in the results is a complete backtest of the base strategy with one (floor, ceiling) pair. The parameter changes only affect which contract prices the strategy is allowed to trade — contracts priced below the floor or above the ceiling are simply ignored. This is a risk filter, not a signal change.

Why these axes: The v3 winner sat at floor 0.05 / ceiling 0.55. By keeping ceiling fixed and walking the floor upward, we can see whether the profitability is concentrated in extremely cheap contracts (where fill assumptions are least reliable) or whether it survives into more moderate price bands. The broader ceiling sweep tests how much the strategy depends on avoiding expensive contracts entirely.


4. Top results

Eight variants produced positive net PnL, all with price_ceiling = 0.55. The top five, ranked by total PnL:

RankFloorTotal PnLROI %SharpeMax DDTradesWin Rate
10.05$33.0166.0%0.72-$30.915237.5%
20.32$29.7459.5%0.67-$19.601471.4%
30.36$29.7459.5%0.67-$19.601471.4%
40.41$29.7459.5%0.67-$19.601471.4%
50.45$28.1356.3%0.97-$1.77875.0%

What the top cluster tells us:

The profitability band is real but narrow. Every positive variant requires the ceiling at 0.55 — raise it even to 0.59 and every cell turns negative. The floor, however, shows some flexibility: values from 0.32 through 0.45 produce nearly identical PnL (~$29–$28) with far fewer trades (8–14 trades vs. 52 for the 0.05-floor winner), much lower drawdown, and higher win rates.

The rank-1 cell (floor 0.05) has the highest raw PnL but also carries three specific warnings from the robustness framework:

  • Thin daily data: Only 7 distinct PnL days. Daily Sharpe estimates on that few observations are unreliable.
  • Extreme fills: 33% of fills are at prices below $0.10 or above $0.90 — exactly where Kalshi’s order book is thinnest and backtest fill assumptions are least trustworthy.
  • Deflated Sharpe of 0.17: After adjusting for the number of trials in the sweep (100 cells), the top Sharpe of 0.72 deflates to 0.17. That means the winner is not distinguishable from the luckiest cell you’d expect to see in 100 random, skill-less trials.

The rank-5 cell (floor 0.45) is arguably the more interesting configuration: $28.13 PnL on just 8 trades, -$1.77 max drawdown, 75% win rate, and a raw Sharpe of 0.97. But with only 8 trades, confidence intervals on all those metrics are wide, and the deflated Sharpe concern applies to the whole sweep, not just the rank-1 cell.

Each successful variant is saved as a runnable Turbine strategy under its unique slug (e.g., momentum-v4-robustness-61d3ce259ddb for the rank-1 cell). None are live or recommended; they are research artifacts.


5. Bottom results

The worst performers cluster at price_ceiling = 0.95 — effectively no ceiling at all. Once the strategy is allowed to buy contracts up to $0.95, losses are deep and consistent across all floor values:

RankFloorCeilingTotal PnLROI %SharpeMax DDTradesWin Rate
940.450.95-$57.26-114.5%-0.46-$108.8615065.3%
950.230.95-$61.36-122.7%-0.66-$107.3415466.2%
970.180.95-$67.15-134.3%-0.75-$113.1315665.4%
980.050.95-$68.32-136.6%-0.75-$114.3015864.6%

What the bottom cluster tells us:

The ceiling parameter dominates. Opening the strategy to expensive contracts (above ~$0.55) destroys performance regardless of floor. Even with high win rates (64–67%), the size of the losses on losing trades overwhelms the winners — a classic negative expectancy profile dressed in a respectable win rate. Trade frequency explodes (150+ trades vs. 8–52 in the winning cells), suggesting the strategy is firing on marginal signals when it doesn’t have the price filter to protect it.

The marginal means confirm this: average PnL across all floor values is +$24.01 at ceiling 0.55, drops to -$7.73 at ceiling 0.59, and craters to -$61.26 at ceiling 0.95. The strategy’s edge — if there is one — lives entirely in the sub-$0.55 contract space.


6. Conclusion

This v4 sweep achieved its objective: it mapped the neighborhood around the v3 winner and confirmed that the profitable region is real, narrow, and tied to a strict price ceiling of 0.55. Within that band, floor values of 0.32–0.45 produce similar PnL with better risk characteristics and fewer extreme-price fills than the raw PnL leader at floor 0.05.

However, three cautions override any temptation to call these results validated:

  1. The deflated Sharpe is 0.17. After accounting for the 100 cells swept, the top result is statistically indistinguishable from selection noise. This is the single most important number in the report. The raw Sharpe of 0.97 on the floor-0.45 cell looks impressive, but with only 8 trades across 7 PnL days, it simply doesn’t survive multiplicity adjustment.

  2. The extreme-price fill warning on the top cell (33% of fills at <$0.10 or >$0.90) means real-world execution would almost certainly be worse than the backtest shows. Kalshi’s liquidity at those price extremes is sparse; assuming fills at back

About this Kalshi strategy backtest

What question does Momentum v4 Robustness investigate?

The v3 dual-regime momentum strategy found a profitable cluster, but the permutation p-value of 0.292 means we cannot rule out luck across the 100-parameter sweep. This v4 sweep narrows the price band and varies polling frequency to test whether the cluster is stable in its neighborhood — prioritizing deflated Sharpe and low sensitivity to small parameter shifts over chasing a single high-PnL cell.

Read the analysis alongside the individual variant results. Check which entry, exit, price-band or risk settings changed, which settings stayed fixed, and whether the report documents enough historical coverage to support its comparison. A different parameter value defines a different simulation; it does not establish a new source of live trading returns.

How many strategy variants are included?

This report lists 100 completed and 0 failed strategy variants. Failed comparisons do not count as successful tests with zero profit. Review trade counts and drawdown alongside net P&L, win rate and Sharpe, and read any coverage or execution caveats in the analysis. A result based on very few trades needs particular care, even when its return looks high.

Does a profitable Kalshi backtest predict live performance?

These are historical simulations. Live results can differ because of liquidity, fees, slippage, latency, market resolution, outages and data quality. Selecting the best result after comparing many variants can also overstate how reliably it will perform on new data. Compare nearby settings, inspect losing variants and test on a separate period before considering a live run. The report is research, not a promise of profit or investment advice.

How can I inspect or test a strategy from this report?

Use an available Open strategy link to inspect that variant. Review its rules, market, date window, fill assumptions, fees and trade history before changing parameters or rerunning it. Confirm which data and execution assumptions the selected market actually supports; a replay does not establish that every historical order could have filled live. For the workflow, read the Kalshi and prediction market backtesting guide. The Kalshi backtesting software page explains the product's supported testing workflow.

This report is generated from historical simulations. Backtests can be wrong or incomplete, and live trading can differ materially because of liquidity, fees, slippage, latency, market resolution, outages, and data quality. Do your own review before running any strategy.

Momentum v4 Robustness · a41f2a87ba1d | Turbine