Will automation give you better returns than trading manually?

Someone asked this under an AI-trading launch thread recently, politely, in a sea of jokes: "For retail investors, will automation provide us greater returns than manual?" It is the only question in the category that actually matters, and it almost never gets a straight answer — because the straight answer sells worse than the implied one.

Here it is anyway. Automation, by itself, has no returns to give you. A bot is an execution layer: it does whatever the strategy underneath it says, tirelessly and without opinion. Automate a strategy with an edge and you get that edge, minus costs, executed consistently. Automate a strategy with no edge and you get your losses on schedule. The machine adds discipline, not alpha.

So the useful question is not "man versus machine." It is: which of the two things that destroy retail returns is destroying yours — the strategy, or the behavior?

What retail traders actually lose to

The research here is old, large, and brutal, and it mostly is not about bad stock picks.

In the classic Barber and Odean study of 66,465 retail brokerage households (1991–1996), the most active traders earned about 11.4% a year while the market returned 17.9% — the heaviest traders gave up roughly a third of the market's return, mostly to overtrading itself. The same research program documented the disposition effect: retail investors are roughly 50% more likely to sell a winning position than a losing one, which is exactly backwards — it harvests gains early and lets losses compound.

And it compounds at the fund level too: Morningstar's long-running "Mind the Gap" work finds investors in funds earn about 1.1 percentage points a year less than the funds they hold, purely from the timing of their own entries and exits. Same vehicle, worse result — the gap is behavior.

Notice what none of these numbers are about: intelligence, information, or stock selection. They are about what people do — trade too much, sell winners, hold losers, buy after rallies, sell after drops. That is the mistake tax, and it is the one part of this problem a machine genuinely solves.

What automation actually fixes

A rules-following system, run on a schedule, does a short list of unglamorous things a human reliably fails at:

• It sizes the same way every time. No doubling up because the last three trades won. No "conviction" sizing at the exact moment confidence is least trustworthy.
• It takes the exit the plan specified. At the level, on the signal — not at the point of maximum discomfort, which is where humans actually sell.
• It never revenge-trades. After a loss it runs the same procedure as after a win, because it does not remember the loss as an insult.
• It shows up every session. Discipline on the days you are busy, bored, or rattled — which is most of the days that matter.
• It can be told to do nothing. This is the underrated one. A well-built system treats cash as a position and will sit in it. Humans pay for action; a machine does not need the action.

If your losses come from the behavior column — and for most retail traders, the research says a large share do — then automating a disciplined process can improve your results without the strategy beating the market at all. Fewer mistakes is a return stream. It is just not the one the ads sell.

What automation cannot fix

Now the other column, because leaving it out is how this category earns its reputation.

It cannot conjure an edge. If the rules underneath have no advantage, automation industrializes the disadvantage. A bad plan executed perfectly is still a bad plan — you just lose with better paperwork.

It cannot escape the regime problem. Every strategy has markets it was built for and markets it bleeds in. A trend system chops sideways; a dip-buyer walks into the one dip that keeps going. Honest vendors tell you which regime pays their strategy and which punishes it. We publish exactly that about our own research record, and the phrasing matters: through the end of 2025 our backbone ran roughly even with SPY at about a third less drawdown — the outperformance concentrates in leadership regimes. In flat, leaderless tape you should expect roughly market returns with less pain, not a headline.

It cannot repeal arithmetic. Costs, taxes on short-term gains, and overnight gaps apply to machines exactly as they apply to you. A stop is an instruction for the next available price, not the price you wrote down — automated or not.

And it cannot promise that everyone wins. Trading has zero-sum aspects; a sold strategy that guaranteed profits for every buyer would be refuting itself. What a system can honestly offer is a process, its record next to a benchmark, and the losing trades published with the winning ones. Anything more generous than that is a red flag, not a feature.

How to actually decide

Skip the demo videos and ask four questions of any automation — ours included:

1. Is the record graded against a benchmark, losers included? A win-rate without a benchmark is decoration. A record that only contains wins is a filter, not a record.
2. Was the backtest survivorship-free, with costs? Testing today's index members backward quietly deletes every company that died. Here is how to read one before you trust it.
3. Are the rules versioned software, or a model's mood? If the judgment lives in a chat model that changes monthly, there is no stable thing a track record can describe. Rules in code, with a changelog, can be graded across versions.
4. Does the vendor say when it loses? Not whether — when. A strategy without a stated losing regime is a strategy you have not been told the truth about. The passive-income version of this pitch is a myth we have written about before.

If the answers are yes, yes, code, and here is the regime that hurts — you are looking at something worth evaluating with money you can afford to risk. If not, the automation question is moot, because the thing underneath it is not a strategy.

Where Coil sits in this, stated plainly

Coil is the disciplined-process case, sold as software you own: a scanner that scores the S&P 500, the Nasdaq-100 and a macro book every market morning, and a long-only engine that buys scored leaders at scored entries, sizes by rule, exits by rule, and holds cash when nothing qualifies. It runs on your machine, against your own brokerage account, and it ships with live trading switched off. Every closed trade is published with its result — the reds the same as the greens — graded against SPY, because that is the standard we just told you to demand from everyone else.

Will it beat your manual trading? We do not know your manual trading. What we can tell you is what it does, when its style of strategy historically pays, when it does not — and show you the ledger either way.

See how Coil works — $29 once