Dollar-Cost Averaging With An AI Strategy

Dollar-cost averaging is the strategy for people who don't want a strategy.

Buy the same dollar amount on a fixed schedule. Don't try to time it. Repeat.

When the price is low, your fixed amount buys more. When it's high, it buys less. Over time, your average entry smooths out.

It's not exciting. That's the point. And it's a surprisingly good first AI strategy - because the rules are so clear that you can check every single decision.

We'll cover:

• Why DCA works as a prompt
• The settings
• The prompt
• What it costs to run
• The downsides when trading perpetuals

Why DCA fits an AI strategy

An AI strategy evaluates your market on a schedule and decides whether to buy, sell, close or do nothing. A DCA rule needs exactly two things the AI already has:

• The time and date. The AI receives the current time, date and weekday on every check - in UTC.
• Your past trades. Past trades are part of the default account context, so the AI can see whether it has already bought this week.

That's enough to write a rule like "buy once every Monday".

The settings

Here's a sensible setup for a weekly $100 buy:

• Markets: one market. Each extra market adds cost and complexity.
• Max Trade Quantity: $100. This caps each order, so even a mistaken decision can't buy more than one week's amount.
• Timeframe: 24h. The decision doesn't depend on short-term price action.
• Evaluation frequency: 1 hour - the slowest available. The rule only needs one check inside the buying window.
• Min. Confidence: leave the default 90%. A clear calendar rule is easy to be confident about.
• Discretion: off. You want the AI to follow the rule, not improvise around it.

The prompt

Four things are doing the work:

• The window is a whole day, so a missed check doesn't mean a missed week.
• "No buy since Monday 00:00 UTC" prevents a second buy in the same window.
• "Never sell or close" stops the AI from "helping" on a red day.
• The maximum position size stops the strategy after six months of buys ($100 x 26).

What it costs

At one check an hour on one market, the help centre's cost table gives about $0.0005 per market per hour, and each check also costs the model once more for the strategy itself - so about $0.001 an hour in total.

That's about $0.024 a day, or roughly $0.17 a week to place one $100 buy. Nearly all of it is the 167 checks that correctly do nothing.

If that ratio bothers you, buy fortnightly with a bigger amount, or run a Fast Estimate first to see the cost of the exact setup. The AI credits page shows what each strategy actually used.

The downsides when trading perpetuals

DCA was designed for buying and holding an asset. On a perpetual future, three things are different:

• Leverage creeps in. Each buy adds to the position, but your margin doesn't grow unless you add cash. To keep it a true DCA, keep your account equity at least as large as the position - in other words, 1x.
• Funding runs every hour. A long position pays funding when the rate is positive and receives it when negative. Over months, that adds up either way.
• Liquidation still exists. At 1x the distance is large, but it isn't infinite. Check it as the position grows.

Wrapping up

• DCA = same amount, fixed schedule, no timing.
• The AI has what it needs: UTC time and past trades.
• Cap each buy with Max Trade Quantity and the total with a maximum position size.
• Checking once an hour costs about $0.17 a week for one market.
• On a perpetual, keep it unleveraged and watch funding.

See all strategy settings on the AI Strategies page.