BTC Horizon Conflict

BTC One-Hour Rebound, Four-Hour Conflict on Kalshi KXBTC15M

Disclaimer (short)

This is historical simulation research on a parameter sweep, not a live trading record and not a promise of future results. Nothing here should be read as investment advice.

Intro / thesis

The idea under test is a momentum-with-conflict setup. The Kalshi series is KXBTC15M, and the underlying read is Coinbase BTC-USD. The thesis asks a fairly narrow question: when the five-minute and one-hour BTC returns agree but the four-hour return points the other way, does following the shorter trend retain a positive after-fee historical result across nearby supported configurations, rather than depending on a single screening pick?

The mechanics are straightforward and fixed. The strategy evaluates every 30 seconds. The first rule liquidates everything if unrealized P&L on the attributed inventory falls to -25 or below. The second rule buys 10 YES contracts when: no position and no orders exist, fewer than 3 portfolio positions are open, time to expiry is between 2 and 5 minutes, the YES best ask sits between 0.35 and 0.80, the spread is 0.03 or tighter, and the three BTC returns line up (5m ≥ 0.0005, 1h ≥ 0.001, 4h ≤ -0.002). The third rule is the mirror image for NO: buy 10 NO contracts under the same position, expiry, price, and spread conditions, but with 5m ≤ -0.0005, 1h ≤ -0.001, and 4h ≥ 0.002. Entry count is capped at one per market, and per-market spend is capped at $50. This is a custom-strategy test with a price-band sweep layered on top.

The sweep tested 100 variants across two axes: the price floor (10 values from 0.05 to 0.45) and the price ceiling (10 values from 0.55 to 0.95). All 100 completed without failure.

Variant and strategy explanation

The base rules above define entries and the single stop. What changed across variants was only the admissible price band — where the strategy is willing to buy YES or NO. The runtime price bounds remain 0.05 to 0.95; the sweep just narrows or shifts that window for entries.

The sweep report shows real spread across configurations. Net P&L ranged from 22.50 at the worst cell to 103.07 at the best. The ceiling axis, in marginal averages, climbs from about 38 at a 0.55 ceiling to roughly 97–98 in the 0.73–0.82 region, then drifts back to about 84 by 0.86 and above. The floor axis is flatter, peaking around 83.7 in marginal average at a 0.27 floor and easing toward 72.8 at 0.45. Neighborhood degradation was 5.7%, meaning the winner's local neighborhood is not wildly worse than the peak — a modestly encouraging sign, though a small one. The deflated Sharpe is 0.896 against the 0.95 threshold, which triggered a low-DSR warning: the top result is not clearly distinguishable from the luckiest of the sweep's skill-less trials. The permutation test scrambled only the edge feed's timing (not market prices) and returned p = 0.010, meaning the strategy's edge-feed timing beat 99.0% of 100 re-sweeps against time-scrambled versions of the same feed. That test does not exercise the price-based rule conditions.

Every successful variant is saved as a runnable Turbine strategy, with its own strategy ID and slug, so any surviving configuration can be loaded and re-run rather than treated as a chart on a page.

Top results

The top eight cluster tightly: net P&L within about 3.5 points, win rates in the high 70s to low 80s, drawdowns around -26 to -30. Ranks 2 and 3 are identical in every metric despite different floor settings, and ranks 5 and 6 likewise. Those ties are noted but unresolved here; nothing in the data explains why those differing floor values produced identical results.

The best cell combines the 0.27 floor with the 0.73 ceiling. It is not a dramatic outlier but it is the peak of a fairly gentle neighborhood, which is what you would want if a real effect were at work — though the DSR warning above argues against reading the peak that way confidently.

Bottom results

The 0.55 ceiling dominates the bottom end. That makes directional sense: a tight ceiling shuts out the higher-priced contracts where the conflict setup probably has more room to pay off, and the win rates collapse toward 50–60%. The two lowest results (0.05/0.55 and 0.36/0.55) are again metric-identical despite different floors, and again that is an unresolved discrepancy rather than something explained by the data provided. Notably, the 0.45/0.55 cell has the smallest drawdown in the entire bottom group (-16.22), which cuts against the idea that the worst cells are worst in every respect — a reminder that net P&L and drawdown are not the same thing.

Ranks 95 and 96 also share identical metrics, and here the trade counts differ by one (65 vs 64) while P&L and win rate match. That inconsistency is unresolved and should not be smoothed over.

Conclusion

Across 100 of 100 completed variants, the configuration space is broadly positive in historical simulation, and the top results cluster rather than sit on a single isolated cell. On the narrow question asked — does the shorter-trend-following read survive nearby configurations — the answer is "mostly yes, with real caveats."

Those caveats matter. The low-DSR warning means the top result is not clearly distinguishable from the luckiest of a hundred skill-less trials; the sweep tested 100 cells, so some selection noise is baked in by construction. The permutation p-value of 0.010 is favorable but only tests edge-feed timing, not the price-based conditions, and the design does not permute the market price series. There is no out-of-sample evidence in what was run. Several tied pairs (ranks 2/3, 5/6, 99/100) and one count/P&L mismatch (95/96) remain unexplained by the data supplied and should be treated as unresolved rather than rationalized.

Practically: the 0.27/0.73 winner and its neighbors in the 0.73–0.82 ceiling range are the configurations worth carrying forward to genuinely out-of-sample testing. Every successful variant is saved as a runnable Turbine strategy, so

About this Kalshi strategy backtest

What question does BTC Horizon Conflict investigate?

When Coinbase BTC five-minute and one-hour returns agree but the four-hour return points the other way, does following the shorter trend on Kalshi KXBTC15M retain a positive after-fee historical result across nearby supported configurations, rather than depend on one selected screening result?

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.