Trading prompts that end in a backtest

I trade crypto perpetual futures on daily bars. Give me five strategy ideas that each have a clear reason to work, such as trend following or funding carry. For each one, give the reason in one sentence, the market where it tends to fail, and the data it needs.

A good answer: Each idea comes with a reason and a failure case. Ideas that need data you don't have are flagged, and no returns are quoted.

Here is an idea I have: [describe it]. Ask me questions, one at a time, until it is specific enough to test: which coins, which timeframe, when to enter, when to exit and how much to hold.

A good answer: The assistant asks before it answers. By the end you have every piece a backtest needs, written in your words.

Turn this idea into rules a computer can follow: [idea]. Put the entry, the exit, the position size and the coins on separate lines. Use only indicators with standard definitions, and state the value of every setting.

A good answer: Every line can be checked by a machine. No line says things like when momentum looks strong.

Using Keel, find the components for these rules and build them as a strategy: [rules]. Show me which blocks you used, and tell me anything you could not express instead of approximating it.

A good answer: A saved strategy, a list of the blocks in it, and an honest note on anything that did not fit. You can open it in the Keel app and read it as blocks.

Backtest this strategy on Keel over its default window. Tell me the return, the worst drawdown and the number of trades, explain what drove the result, and give me the link to open it in the Keel app.

A good answer: Numbers from a real run, a plain reading of why, and a link you can open to check every trade.

Run this strategy on Keel with [setting] at three different values and compare the backtests side by side. Tell me whether the result holds across them or depends on one value.

A good answer: Three runs on the same window and a straight answer. A result that only works at one value is a warning, and a good answer says so.

Is this result real or curve-fit? Run the strategy on Keel over a period it was not built on, and over a rising and a falling stretch, then give me one verdict and the runs behind it.

A good answer: A verdict backed by runs on windows the strategy never saw, with the weakest one shown, not hidden.

Here is my backtest result: [paste it]. List the ways it could be misleading, such as look-ahead, too few trades, one lucky coin or costs left out. For each, tell me how to check it.

A good answer: A checklist you can work through, specific to the result you pasted, without reassurance.

My account is [size] and I can accept losing [amount] in a bad stretch. This strategy's worst backtest drawdown was [drawdown]. How much of the account should it use, and what leverage does that mean? Show the arithmetic, and assume the next drawdown can be worse.

A good answer: The working shown step by step, a margin for a deeper drawdown than the backtest, and a size you can say out loud.

Summarise the strategies running on my Keel account: positions, recent fills and how each one's drawdown compares with its backtest. Flag anything that looks stale or out of line, and change nothing.

A good answer: A read-only summary that flags gaps between live and backtest. Starting or stopping a strategy stays with you, in the Keel app.