I debated risk rules with my AI, and got a better rule

1. The warning

I was about to open a separate account and start smaller, keeping the same rules.

Billie is an AI I built for my own trading. It isn't a general chatbot — it is wired into my observation logs and my trading database, and it remembers the rules I've written over the past several months.

I told Billie the plan. It answered:

"🚨 This rule doesn't fit the account. Your daily max risk becomes 50% of the account. That's a rule that lets you put half the account at risk in one day. Two losing days and it's empty."

On the numbers, that's correct. Halve the account and the same limit becomes twice the weight.

I disagreed.

2. My pushback

Two points.

① Defined risk is collateral, not expected loss

With a credit spread, the maximum loss is fixed the moment you enter, and that amount is locked up as margin.

So it isn't "I am exposing 50% of the account to risk." It is "I set that much aside before I started." The worst case is calculated in advance and quarantined. The account as a whole is not swinging.

② In options you cannot earn without carrying the risk

This is the more important one.

A credit spread structurally loses more than it wins. Take a 5-point-wide spread collecting 1.90:

	per contract
max profit	$190
max loss	$310
R:R	1 : 0.61

Halve the risk and you halve the return, exactly. "Let's just risk a little" is, in premium selling, the same sentence as "let's just make a little."

So I asked: "What do you think?"

3. Billie's answer — conceded

The first line came back:

"You're right. My '50%' framing was wrong. I conflated two things."

Then it explained what.

I confused collateral with expected loss. The max risk on a credit spread is the amount fixed at entry and locked as margin — not a forecast that says "I expect to lose this much."

It accepted the R:R point too, and put numbers on what I had only said in words.

breakeven win rate = 310 ÷ (310 + 190) = 62%

Win rate	expected value per contract
85%	+$115
70%	+$40
62%	0 (breakeven)
50%	−$60

And then one sentence:

"This strategy doesn't make money because the R:R is favorable. It makes money because the win rate is high."

That sentence made me understand my own strategy differently. I had been targeting 85–90% win rates, and this calculation explains why it has to be that high. Below 62% the structure loses money.

4. But the real problem was somewhere else

If Billie had just conceded and stopped, this would only be a story about an AI being wrong.

It continued:

"So the real question isn't the size of the risk. It's how many days of ammunition it represents."

And it quoted a principle I had written myself, months earlier:

"The account ceiling is a last line of defense, not a daily rule. Spend the whole account as a daily limit and the box empties in one day, with nothing left to recover with."

Applied to the current plan:

Account	daily limit	days of ammunition
original plan	unchanged	3 days
smaller account	unchanged	2 days

Two worst-case days and the account is empty.

Then one more thing:

"On 0DTE credit spreads, max loss is not a theoretical tail."

I have lived this. On one day, several spreads I had entered settled at maximum loss across the board. Held to expiry, in-the-money means max loss, full stop. There is no partial outcome the way there is with a long option.

Max loss is not a rare extreme. It is the default on a losing day.

5. The fix — days of ammunition, not dollars

Billie's proposal wasn't to lower the limit. It was to change the unit.

account ÷ daily max risk ≥ N days

Write the rule in dollars and it breaks every time the account changes. Write it as a ratio and it follows the account up or down.

I set N = 3.

There is a second benefit. As the account grows, the max risk grows with it automatically. I'm starting small, but the size I can carry scales as the account scales. No renegotiating the rule.

⚠️ The cost is just as clear. The larger N is, the fewer contracts, and the lower the daily ceiling on profit. The objection I raised at the start — "then you can't earn that much" — comes right back here.

There was no free lunch. I just get to see the trade-off in numbers now, and choose it deliberately.

6. What I learned from the exchange

Billie gets things wrong. And the wrong spot is useful

Billie's first warning was wrong. It conflated collateral with expected loss.

But what if I hadn't pushed back? I would have tightened the rule unnecessarily, without understanding why I was tightening it.

And if I had only pushed back and stopped there? I would never have seen the two-days-of-ammunition problem.

A wrong objection brought the right question.

Push back with structure, not feeling

If I had said "50% doesn't feel right to me," the conversation would have ended there.

What I offered was structure. Defined risk is collateral. The R:R is 1:0.6, so cutting risk cuts return. Those are facts about how options work, not opinions.

Push back with structure and you get calculations back. That calculation was the 62% breakeven — a number I did not know.

It remembered a note I had forgotten

The part that struck me most: Billie pulled up a principle I wrote two months ago. I had forgotten the sentence.

"The account ceiling is a last line of defense, not a daily rule."

A conclusion I reached in the past argued against the present me. Billie only made the connection.

That is what working with AI actually looks like to me. Not an oracle handing over answers — something that remembers my own record better than I do, and joins the pieces I dropped.

Summary

Defined risk is collateral. Judging a credit spread by account percentage alone misreads it
Credit spreads pay because the win rate is high, not because the R:R is good. On a 5-wide spread at 1.90, breakeven is a 62% win rate
Max loss is not an extreme — it is the default on a losing day. Especially with 0DTE
Write risk rules as ratios, not dollars. account ÷ daily limit ≥ N days

This is a record of how I set a risk rule. It is not a recommendation to use any strategy or trade any security. The breakeven math uses illustrative figures and is ordinary options arithmetic. Your decisions and their outcomes are your own.