You have a profile that has behaved itself on a crypto pair for a few months and the obvious next thought is to point the same logic at a stock. The engine takes both. The profile settings are the same fields. Nothing in the interface objects.
What objects is the calendar. A crypto pair trades every hour of every day including the ones you are asleep for, and a stock trades for about six and a half hours on weekdays and then stops, and that single difference silently rescales half the parameters on your profile. Some of them travel anyway. Some of them mean something completely different on the other side and will keep producing plausible looking results while measuring the wrong thing.
The cadence on your existing profiles is already telling you this
Look at how a working set of profiles tends to be organised. On the Autopilot page I am looking at, the equity and index followers run on a daily cadence, and the crypto and metals pairs run on four hour cadences. Eleven profiles, ten of them enabled.
That split is not decoration. A four hour bar on a 24 hour market gives you six evenly spaced observations a day, every day, forever. A four hour bar on a stock gives you one full bar and a stub before the close, five days a week, with a hole every night and a bigger hole every weekend. The same setting produces a regularly sampled series in one market and a ragged one in the other, and any parameter you expressed in bars now means two different things.
So the first honest answer to the question in the title is that a profile which works on a crypto pair at four hours is not being ported to equities when you point it at a stock. It is being changed, whether or not you edited anything.
Setting up a comparison the calendar cannot rig
The Backtest a Profile panel takes a profile, a number of days and a start equity, and replays historical signals against that profile with no real orders, honoring its filters, sizing, leverage caps and risk guards. Two runs, one profile pointed at a crypto series and one at an equity series, is the experiment. Getting the setup right is most of the work.
Use the same day count and the same start equity, and run both on the same afternoon, since the day count runs backward from when you press the button. Then apply the correction that everyone skips: do not compare trades per calendar day. A crypto market offers roughly 168 trading hours a week and a US equity market offers about 32.5. The same logic will produce something like five times as many opportunities per calendar week on the crypto side for no reason connected to the strategy.
Normalise on trades per hundred hours of market time, or if that is fiddly, compare trades per bar rather than per day. Then normalise the results too. A profile that made twice as much on crypto over ninety days while taking five times as many trades earned less per trade, and per trade is where the friction is applied, so per trade is what determines whether it survives contact with a broker.
One more setup note. Pick a window that contains a weekend crash. Crypto's distinguishing feature for a retail trader is that the market moves while you cannot do anything about it and equities mostly do not, and a window with no weekend event in it will report that this makes no difference.
The three parameters that never travel
These change meaning across the boundary and should be re-derived rather than copied.
Anything counted in time. A twenty four hour cooldown blocks re-entry for one full day of trading on a crypto pair and for nearly four sessions on a stock, because the clock keeps running while the market is shut. If you want equivalent behaviour, express the cooldown in bars or sessions and convert, and accept that the conversion is approximate.
Anything expressed as a percentage of price. A stop three percent below entry is a normal, slightly tight stop on a mid cap crypto pair and an extremely wide one on a large cap stock, where a typical daily range is a fraction of that. Ported unchanged, the stop simply never triggers on the equity side, your losing trades run to the exit signal instead, and the loss distribution you get back has nothing to do with the one you tested. Re-derive stops from the instrument's own average range rather than from a number that felt right elsewhere.
Anything about liquidity or minimum size. A volume filter calibrated for a token screens nothing in a large cap equity universe, and one calibrated for equities screens out most of crypto. Worse, the cost model differs in kind: crypto venues generally charge proportionally, while many equity brokers charge a fixed amount per order, so a small position that is cheap on one side has a fee floor on the other. Apply your actual equity commission on a round trip to your intended position size and see what percentage it represents. If your average trade in the crypto run was worth well under one percent, the equity version may not be viable at your size regardless of how the signal performs.
The two that usually do travel
The portable parameters are the ones expressed as fractions of your own account rather than as properties of the market.
Risk per trade travels. One percent of the account between entry and stop is one percent in either market, and it is a statement about you, not about the instrument. The dollar amount is unchanged and only the position size that delivers it differs, which is exactly what you want a sizing rule to do.
Concurrency travels approximately. The number of simultaneous positions you can psychologically and financially carry does not depend on which market they are in. What changes is how quickly you reach the ceiling, since the crypto side generates entries around the clock and will hit a concurrency limit faster in calendar terms. Expect the same limit to bind more often on crypto, and treat that as correct behaviour rather than something to tune away.
The entry logic itself may or may not travel, and the backtest is the only way to find out. Be suspicious of a strong result in both, though, particularly over a window where crypto and equities were both trending. A profile that works on everything during a broad rally is frequently a long bias with extra steps, and the way to check is to run it over a period where the two markets diverged.
What breaks first when you actually port it
The thing that breaks first is not the strategy, it is the supervision, and the module is blunt about the stakes on the same page as the backtest. With paper mode off, Autopilot places real orders on your connected exchanges, execution prices may differ from signal prices because of market conditions and latency, and you are solely responsible for the trades. Its instruction is to start in paper mode, verify the behaviour, then switch to live with small sizes. Do that on the ported profile specifically, even though the profile is one you already trust, because what you are testing is not the logic. It is the logic in a market with different hours.
Going the other direction, from equities to crypto, is where retail traders get hurt most reliably. An equity profile is implicitly supervised: it can only act while the market is open, and the market is open during hours when you are broadly awake. Port that same profile to a crypto pair and it acquires the ability to open positions at four in the morning on a Sunday, into the thinnest order book of the week. Nothing about the settings changed. The entire supervision model did.
So before you enable a ported profile, decide two things and write them down. What the maximum position size is during the hours you are not watching, which should generally be smaller than your daytime size rather than the same. And where the pause and emergency kill switch controls are, since they sit at the top of the Autopilot page and are the only response available when a profile is doing something you did not anticipate in a market that will not close to give you time to think.