BTC RSI Strength Price Sensitivity

BTC RSI Strength on Kalshi KXBTC15M: Entry-Price Sensitivity Sweep

Short disclaimer

This is historical simulation research, not investment advice and not a forecast. It uses backtested data only, and past simulated results do not imply future profits.

Intro / thesis

This report looks at a single fixed screen on Kalshi's KXBTC15M market and asks a narrow question: does it hold up when we only change the prices we're willing to pay?

The strategy trades BTC-linked 15-minute binary markets. It uses Coinbase BTC-USD as an outside signal feed for two RSI readings: a 14-period RSI on one-minute candles and a 14-period RSI on five-minute candles. The entry idea is momentum confirmation. For a YES entry, the one-minute RSI has to cross above 60 while the five-minute RSI sits above 50. For a NO entry, the mirror: one-minute RSI crosses below 40 while the five-minute RSI sits below 50. Both sides only take a fill when the spread is at most 0.03, the relevant best ask is between 0.30 and 0.70, and the market is between 180 and 600 seconds from expiry. Positions are one per market, up to three markets at a time, 10 contracts each, with a $10 notional cap per market.

The exits are deliberately plain. If a YES position is held and one-minute RSI drops below 50, it sells. If a NO position is held and one-minute RSI rises above 50, it sells. A broad stop sells everything if unrealized P&L reaches -$25. These rules are evaluated every 30 seconds, top to bottom, first match wins, and the edge feed reads closed candles only, so values update when a candle closes.

The thesis under test: that this momentum screen may persist through the remaining binary window. This sweep tests whether the positive fixed screen survives entry-price eligibility sensitivity by varying only the global price floor and ceiling. It does not optimize indicator thresholds. Nothing about the RSI levels, the timing windows, or the exit logic changed across the 100 variants.

Variant and strategy explanation

The sweep varies exactly two parameters: the risk price floor and the risk price ceiling. The floor takes ten values (0.05, 0.32, 0.34, 0.36, 0.38, 0.40, 0.42, 0.44, 0.46, 0.48) and the ceiling takes ten values (0.50, 0.52, 0.54, 0.56, 0.58, 0.60, 0.62, 0.64, 0.66, 0.95). That's a 10×10 grid, 100 cells, all 100 completed.

An important note on floor and ceiling here: they are the risk-level bounds on the global price, not the per-side entry ask band. The entry ask band stays fixed at 0.30–0.70 as written in the rules. The floor and ceiling change which candidate fills are eligible before the strategy's own ask conditions are applied, which is why the effective trade counts move substantially from cell to cell (from roughly 62 trades at the tightest cells up to 619 at the widest).

Each successful variant is saved as a runnable Turbine strategy, so any cell in this grid can be pulled up and run directly rather than rebuilt from the report.

Top results

The best cell is floor 0.05 / ceiling 0.66. It produced $315.88 of net P&L, ROI of 3158.8%, a Sharpe of 1.02, max drawdown of -22.57, 497 trades, and a win rate of about 58.7%.

The next tier is close behind and mostly shares the same floor:

• Floor 0.05 / ceiling 0.64: $301.02, Sharpe 0.98, drawdown -22.57, 476 trades.
• Floor 0.05 / ceiling 0.60: $298.67, Sharpe 0.97, drawdown -22.15, 403 trades.
• Floor 0.32 / ceiling 0.66: $292.59, Sharpe 0.97, drawdown -28.03, 478 trades.
• Floor 0.05 / ceiling 0.62: $289.46, Sharpe 0.97, drawdown -22.57, 432 trades.

Two patterns stand out in the top group. First, the lowest floor (0.05) dominates, and raising the floor tends to cost P&L. Second, the highest ceilings and the unrestricted ceiling (0.95) all land in the upper half of the grid, so tight ceilings are not where the returns live.

One honest caveat before reading too much into the top: the floor 0.32 / ceiling 0.66 cell has a larger drawdown (-28.03) than the cells directly above it despite a similar Sharpe, and that appears without a clean explanation in the data. It is an unresolved discrepancy, not something I'll hand-wave.

Bottom results

The weakest cell is floor 0.48 / ceiling 0.50: $39.17 net, ROI 304.8%, Sharpe 0.33, drawdown -12.77, only 62 trades, win rate 0.60. The rest of the bottom is a cluster of tight, high-floor cells:

• Floor 0.46 / ceiling 0.50: $57.24, Sharpe 0.44, 76 trades.
• Floor 0.48 / ceiling 0.52: $91.33, Sharpe 0.68, 105 trades, win rate 0.66.
• Floor 0.48 / ceiling 0.95: $137.65, Sharpe 0.58, 453 trades.
• Floor 0.44 / ceiling 0.50: $103.18, Sharpe 0.71, 109 trades.

Note the floor 0.48 / ceiling 0.95 cell: it has 453 trades and a win rate of about 59%, yet only $137.65 of net P&L. That's a wide ceiling and a high floor, and its Sharpe (0.58) sits well below the top cells with similar trade counts. The win rate and trade count look reasonable in isolation, so the low P&L is a discrepancy I can't fully resolve from this data.

The general bottom story is that stacking a high floor against a tight ceiling shrinks the eligible trade set to a small handful and leaves very little to compound. The drawdowns in the bottom group are smaller (-12 to -29), but so is everything else.

On the parameter-wide picture: the marginal averages for the floor slope clearly downward as the floor rises (mean P&L from $266.32 at 0.05 down to $122.99 at 0.48), and for the ceiling they slope upward from $117.35 at 0.50 to $255.54 at 0.66 before easing slightly to $221.62 at 0.95. The neighborhood degradation figure is about 7.6%, which says adjacent cells are not wildly different from the winner.

On robustness testing: the permutation test on the edge feed returned p = 0.010, meaning the real timing beat 99.0% of 100 re-sweeps against time-scrambled versions of the same feed. That's a favorable result, but the design only scrambled the edge feed; the market price series was not permuted, so the price-based conditions in the rules are not tested by this design. The deflated Sharpe probability is 0.9998 against an expected-max-Sharpe benchmark of 0.33 (not directly comparable values). By the rules of these checks, a high deflated Sharpe probability is favorable only for the selection-noise check itself; it cannot establish profitability or make up for missing out-of-sample evidence. With 100 cells and a best-cell selection, the in-sample best is exactly what the grid was designed to find, and there is no out-of-sample test here to confirm it.

Conclusion

The sweep did what a price-eligibility sweep should do: it showed that the strategy's sign survives across the grid, and it showed that the magnitude does not. All 100 cells produced positive net P&L, but the spread from $39 to $316 is roughly an 8× range on identical indicator logic. The best cells are the low-floor, high-ceiling ones; the worst are tight, high-floor cells where the trade set shrinks to a few dozen fills.

The more important finding is the shape of that spread. If only the entry thresholds mattered, price bounds would wash out; here, raising the floor from 0.05 to 0.48 cut average P&L by more than half. That's a real sensitivity, not noise to ignore.

What this does not establish: the permutation test supports the edge-feed timing against scrambled feeds (p = 0.010), but it does not test the price-based conditions. The deflated Sharpe check is favorable but purely a selection-noise check. There is no out-of-sample window in this data. So I'd call this a promising fixed screen with a clear price-sensitivity profile and a plausible persistence story, not a validated strategy.

The saved Turbine variants are the next step, not the end of the work. The recommended starting point, if you're going to look at any cell, is floor 0.05 / ceiling 0.66, but treat that as the best observed in-sample cell rather than the correct one.

Long disclaimer

This report is historical simulation research distributed by Turbine for research and informational purposes only. It is not investment, financial, legal, or tax advice, and it is not a recommendation, solicitation, or offer to buy or sell any security, contract, or derivative, including prediction-market contracts on Kalshi or any other venue.

Everything here is based on backtested and simulated data. Backtests are reconstructions, not records of actual trading. They rely on assumptions about fill availability, order priority, spread, and slippage that may not hold in live markets, and they use closed-candle edge data replayed on minute values, which may not match the timing a live

About this Kalshi strategy backtest

What question does BTC RSI Strength Price Sensitivity investigate?

On Kalshi KXBTC15M, a one-minute RSI14 cross above 60 with five-minute RSI above 50 (or the mirrored cross below 40 with 5m RSI below 50), with neutral RSI-50 reversal exits, may persist through the remaining binary window. This sweep tests whether the positive fixed screen survives entry-price eligibility sensitivity by varying only the global price floor and ceiling; it does not optimize indicator thresholds.

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.