AI covered calls: when NOT to sell another one

A covered call is a trade you can run on a stock you already own: you agree to sell your shares at a set price, and you get paid a small fee, the premium, for the promise. The catch is what happens after one closes. My trade #6 on BMNR closed on 13 May for +$375 in nine days: sold the promise, bought it back for less than half, a clean win. The app pings. I’m holding the shares again. The screen of prices I could sell against is right there.

The instinct is to do it again straight away. The win felt good, and the same trade is sitting there at roughly the same price. Why wouldn’t you?

Because repeating it immediately is the behaviour this small record is testing, and six trades cannot establish whether waiting wins more often. A tightly fenced prompt has been useful to me because it applies the same five questions before I click. It is not predicting the next move. Five checks, six recorded trades, one discipline under test: selling another call is a fresh decision, not the default.

Should you sell another covered call straight after closing one? My answer, from this six-trade BMNR record, is usually not. Before I open the broker’s list of strikes and prices, the same prompt applies five checks: where the stock sits against my basis, whether a price-moving event is near, why the last trade closed, how current IV compares with my chosen window, and whether a clean close is a reason to pause rather than repeat. Any one of them can make my checklist say wait.

Why the urge is the problem

The money creates the pressure. A clean result invites an immediate repeat: nine days, +$375, do it again. In this record, that is the behavioural risk the checklist is designed to interrupt. It cannot tell me whether the next call would have won; it can make me earn a fresh reason for taking it.

The trade I'm trying to avoid is the automatic re-entry: the alert was loud, the chain was open, and I had not earned a fresh setup.

Across the closes covered by this record, I run the same AI prompt before considering another call. It forces me to answer five questions in the same order. That order is the Prompt Stack: fence it to supplied data and let it flag what it cannot verify, give it the facts, ask where the risk is, get a one-line verdict. Left to itself, the urge skips straight to the prices. The prompt makes me stop.

Five conditions where the rules say WAIT

Each of these is a reason the prompt comes back with WAIT, even when the trade looks fine on the screen.

1. The stock is below your basis after a fall

“Basis” is just what the shares cost you, on average. My trade #5 closed in April for a thin profit, barely worth the paperwork: the stock had crept up toward the $23 price I’d agreed to sell at, so I closed it early. The problem was where that left me: I’d been buying more on the way down, which pulled my real average cost a few dollars below the headline figure my broker showed. Selling again at that $23 price would have locked the shares in just as the stock was about to climb back. The rule is to wait. When this was written, I was waiting for the stock to recover toward $22.50–$23.00 before even considering the next trade, a level it has since dropped well below, so that specific number now needs re-basing (see the dated note further down). The reason behind it doesn’t.

2. There’s a known event inside 14 days

If something in the next 14 days could jolt the stock (a company announcement, a move to a bigger stock exchange, results from a similar company), my rule treats that as a reason not to set a sale price immediately beforehand. Waiting may leave premium on the table; it buys flexibility through the event. BMNR moved up to the main New York Stock Exchange on 9 April 2026. Selling into that window did not fit my process.

3. The last trade closed because the stock fell, not because it went well

There’s a difference between a call bought back under a planned management rule and one that became cheap because the stock dropped. In the second case, the agreed sale price may look “safer” because it sits further above spot, but the shares still carry the downside and another call still caps any rebound. After trade #5’s thin close, BMNR drifted lower for weeks. My entry rule therefore requires a fresh price or volatility setup rather than treating the fall itself as permission to sell again.

4. The fee is below this stock’s normal level

The size of the fee is influenced by the move the market prices in. The jargon is “implied volatility”. When this rule was written, my dated context recorded BMNR IV around 75–85%; that is not a timeless normal range. My chosen filter compares current IV with its 90-day median and says wait when it is below. IV Rank and IV Percentile use different one-year calculations, so I do not treat those labels or windows as interchangeable. The filter creates consistency; it does not prove a particular premium is inadequate.

5. You just closed a clean trade: wait for the next fresh setup

Two of my management rules close a trade early: one takes the win once the fee has lost half its value (the 50% target), the other gets out around 21 days before expiry. Those are my process thresholds, not universal points where the odds suddenly flip. When both rules fire together, my process says step back and wait for a fresh setup. Trade #6 closed with 23 days left and both rules met. Under that process, the next move was to watch rather than re-enter because the alert was still flashing.

What you do once you’ve decided to wait

Once the prompt says wait, the job becomes watching rather than a one-off decision. I keep a short list of what would flip the answer: the stock recovering to a price I’d be happy to sell at, the fee spiking back up, the right expiry coming into range, and a strike set a sensible distance above today’s price. I use roughly 30 delta as a selection input. Some traders read delta as a rough probability of expiring in the money, but it moves with price, time and volatility; it is not a literal assignment probability. Assignment can occur on any trading day. The exact inputs move with the stock; under my process all four have to line up before I sell another call.

Update, 12 June 2026: The specific levels in my re-entry table when this was published (a $22.50–$23.00 price trigger, a July expiry, a roughly $26–$27 strike) were built around BMNR trading in an $18–$23 range. It has since fallen to the mid-$16s on ETH weakness, so those numbers are stale and the watch table needs re-basing before the next call is sold. The five conditions below are unchanged; only the levels they point at have moved.

I re-run the prompt weekly with fresh numbers. The answer flips from wait to sell when the conditions earn it, not when an alert pings. The prompt isn’t doing the thinking. It’s holding me to the order I check things in.

Where these five conditions fall short

The five checks stop you selling another call when you shouldn’t. They don’t find you the next trade. When I ran the prompt after closing trade #6, it said wait, and that was the whole of its value. It stopped me selling again; it didn’t hand me a new trade. It also assumes you’re feeding it honest numbers: the fee figure from your broker or a tool like Barchart, an events calendar you’ve actually checked, a cost figure that includes the fees you paid. Bad numbers in, confident wait out. (And it assumes this is still a stock you want to be doing this on at all: that earlier question is Prompt 1 of the entry-side post.)

The deepest limit is provenance. The unconnected chat used for this workflow did not have a broker feed, so it was not allowed to pick a trade, estimate fees, or fill in missing chain data. Connected exceptions now exist: OpenAI documents an Alpaca plugin that can retrieve live option chains. That does not make every chat live-data capable, and it does not remove the need to identify the feed and timestamp. This prompt remains fenced to verified inputs: if a required number is missing, the run stops rather than guessing. The wider boundary is set out in AI for options trading: what it gets wrong.

Garbage in, confident WAIT out.

What would change my mind: a run of three or four “wait” calls that turned out to be obvious misses (a safe sale price, a steady fee, no events ahead, and the stock ran anyway). That would mean the rules are too tight for this stock, and I’d loosen them. Six trades in, the calls have held.

The urge after a win is to sell another call. The prompt is what makes that a decision instead of a reflex.

The private trade ledger remains the source for the six historical results. The public article carries only the bounded figures needed to explain this test; it does not expose position sizes or pretend the retired /trades page still publishes the ledger.