Entry Level Sets

Kalshi KXBTC15M — Entry Level Sets

Short disclaimer: This is historical simulation research on a 30-day window. It is not investment advice, not a live track record, and not a promise of future results. Prediction markets can and do lose money.

Intro / thesis

This batch tested one narrow question on Kalshi's KXBTC15M crypto market: does it matter which of the late-expiry entry levels you allow? The base strategy already defines four late levels — 4 minutes, 2 minutes, 1 minute, and 30 seconds to expiry — each with its own distance-from-strike requirement and price band. The thesis was that some of those levels add nothing but noise, and that trimming the set might change net P&L materially while leaving everything else untouched.

Three variants were defined, all sharing the same exit logic, the same market selection (most_liquid on KXBTC15M), the same 50-contract cap, and the same BTC edge feed from Coinbase (price, 5-minute change, 1-minute velocity, 26-period EMA on the 15-minute). The only thing that moved was the set of entry levels:

• Variant 1 — levels 6 through 9 (4m, 2m, 1m, 30s).
• Variant 2 — levels 7 through 10 (2m, 1m, 30s, 25s).
• Variant 3 — levels 8 through 10 (1m, 30s, 25s).

All three ran over 30 days. Each successful variant is saved as a runnable Turbine strategy, so any of them can be picked up and re-run without rebuilding the rules.

Variant and strategy explanation

Every variant inherits the same base mechanics. The engine evaluates every 10 seconds and applies custom rules top-to-bottom — the first matching rule fires and ends the tick.

The risk-and-exit block is identical across all three. A held YES position is sold if the YES bid falls to 0.35 or lower; a held NO position is sold if the NO bid hits 0.35 or lower. A YES position is also closed whenever spot sits below the strike, and a NO position whenever spot sits above it. There is a take-profit exit for both sides: a YES bid (or NO bid) at 0.995 or higher while between 15 minutes and 2 minutes to expiry. Price bounds are 0.05 to 0.99, max position is 50 contracts.

The entries are where the variants diverge. Each level pairs a time-to-expiry cap with a minimum distance between spot and strike, and requires the EMA to be on the same side of the strike, the 5-minute change to agree in direction, the 1-minute velocity to be non-negative for YES (non-positive for NO), a best ask between 0.90 and 0.99, and a spread no wider than 0.02:

• Level 6 — at most 4m to expiry, spot at least 60.1 above (YES) or 60.1 below (NO) the strike.
• Level 7 — at most 2m, distance at least 40.1.
• Level 8 — at most 1m, distance at least 30.
• Level 9 — at most 30s, distance at least 20.
• Level 10 — at most 25s, distance at least 15.

Level 10 appears only in Variants 2 and 3; level 6 appears only in Variant 1. All three variants share the base DSL as context; each candidate is a complete, self-contained rule set.

Top results

Variant 1 — levels 6 through 9 (4m, 2m, 1m, 30s) took the top spot. It returned +$142.79 total P&L on a 285.58% ROI across 294 trades, with a 96.3% win rate and a max drawdown of -$157.97. The drawdown is larger than the total profit, which is a meaningful observation: the equity path was clearly not smooth, and the peak-to-trough swing exceeded the eventual net gain.

Variant 3 — levels 8 through 10 (1m, 30s, 25s) ranked second: +$23.06 total P&L, 46.12% ROI, 14 trades, 100% win rate, max drawdown -$8.15, Sharpe 0.32. Fourteen trades over 30 days is a very thin sample — the win rate here is not something to lean on, it's just a small-n artifact.

An unresolved discrepancy worth flagging: Variant 3's Sharpe (0.32) is the highest of the three, yet its ROI (46.12%) is well below Variant 1's 285.58%, whose Sharpe is only 0.20. Sharpe and ROI are measuring different things, and both are being computed on short, uneven trade counts — but the direction of the gap between the two variants is not something this batch explains. We're leaving it as an open question rather than attributing it to a specific cause.

Bottom results

Variant 2 — levels 7 through 10 (2m, 1m, 30s, 25s) came in third and only negative: -$19.33 total P&L, -38.66% ROI, 92 trades, 93.2% win rate, max drawdown -$88.15, Sharpe -0.05.

Variant 2 is a useful counterweight to Variant 1. It contains three of the four levels Variant 1 contains (2m, 1m, 30s) plus the 25-second level, and it drops the 4-minute level. It lost money. Variant 1, which adds the 4-minute level and drops the 25-second level, made money. That single swap — 4m in, 25s in; or 4m out, 25s in — is the whole difference between the top and bottom of this batch. With one 30-day window and no sweep, there's no way to say whether that reflects a persistent property of the 4-minute level or just this period. It's the kind of thing you'd want to see replicated across several non-overlapping windows before believing it.

Variant 3 is the third entry here (it appears in both the top and bottom listings since the batch was only three variants). Its 14 trades and -$8.15 drawdown make it look tidy, but 14 trades is not enough to say much beyond "it didn't blow up."

Conclusion

Three variants, one month, one market, one set of base rules. The level-set choice clearly mattered to the P&L — the spread between +$142.79 and -$19.33 is not small — but the batch does not tell us why in any durable sense. Variant 1's win rate (96.3%) is high, but its drawdown exceeds its net profit, so the path was rough. Variant 2 lost money despite winning 93.2% of its trades, which is a reminder that win rate and P&L are not the same thing. Variant 3 made money on 14 trades, which is too few to read as a signal.

No sweep or permutation test ran here. This was an explicit three-way comparison. The winner should not be treated as statistically robust — different periods, different risk budgets, and different market regimes can make these numbers look very different. Each successful variant is saved as a runnable Turbine strategy, so the next step, if there is one, would be to re-run them on additional windows and see whether the ordering holds up.

Long disclaimer: This report describes historical simulation research conducted on Kalshi's KXBTC15M market series over a 30-day lookback window. It is provided for research and informational purposes only and does not constitute investment advice, a recommendation, or an offer to buy or sell any security or contract. The variants described were evaluated on past data under specific rule configurations and cost assumptions; past simulated performance is not indicative of, and does not guarantee, future results. Simulated results are subject to known limitations, including but not limited to: the absence of real fills and real slippage, the assumption that quoted prices were transactable, the use of a single 30-day window that may reflect period-specific market conditions, small trade counts in some variants (notably Variant 3 at 14 trades), and the fact that no parameter sweep, permutation test, walk-forward analysis, or out-of-sample validation was performed. Drawdown figures, Sharpe ratios, ROI, and win rates are point estimates from a single run and should not be interpreted as stable properties of the strategies. Prediction markets involve substantial risk, including the risk of total loss of capital; crypto-related contracts are additionally subject to volatility, liquidity, and venue-specific risks. Nothing in this document should be relied upon as the basis for any trading or investment decision. Anyone considering live deployment should conduct independent due diligence, including their own backtesting on non-overlapping data, careful consideration of transaction costs and market impact, and appropriate position sizing and risk management. The author may hold positions in instruments referenced. This report may be updated or withdrawn without notice.

About this Kalshi strategy backtest

What question does Entry Level Sets investigate?

On KXBTC15M, test whether restricting entries to different subsets of the late expiry levels changes net P&L: levels 6-9 (4m/2m/1m/30s) versus 7-10 (2m/1m/30s/25s) versus 8-10 (1m/30s/25s). All other rules fixed.

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 3 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.