My phone buzzed at 3:12 a.m. Position closed, minus 2.1%.
I was awake before I'd read the second line. By 3:20 I had the chart open, and by 3:30 I had a theory about why the next signal was probably going to fail too.
The bot hadn't done anything wrong. The stop was exactly where the rules put it. A normal loss, inside the backtest range, one of dozens I'd already seen. What changed was me. A system built to remove my judgment had just handed it a notification and an invitation.
Here's the part nobody mentions when they talk about automation. A bot removes the decision to enter and the decision to exit. It doesn't remove you. If every fill, every stop and every red number gets pushed to your pocket, you've built a machine that makes you feel every trade anyway.
There's old research on this. In 1997, Richard Thaler, Amos Tversky, Daniel Kahneman and Alan Schwartz ran an experiment where people split money between a stock-like asset and a bond-like one. The only difference between groups was how often they saw their results. Participants who saw returns frequently put roughly 40% into the risky asset. Those who saw results only occasionally put closer to 70%.
Same investment. Same odds. Different amount of looking. The researchers called it myopic loss aversion: the more often you check, the more often you meet a loss, and a loss hurts about twice as much as an equal gain feels good.
Now apply that to a bot with a 35% win rate. By design, most individual trades lose. If you get an alert for each one, you're signing up for a steady stream of small pain, with the wins buried in between. The equity curve goes up. Your experience goes down.
I think this is where most people who "trust their system" actually break. They don't override on the big drawdown. They override on the fourth notification in a row on a Tuesday, when pausing the bot for a day feels harmless. It never is. The trade you skip is statistically not any worse than the trade you take, but you don't get to know which one it was.
Most traders respond by trying to be tougher. Look at the pings and don't react. That's a willpower plan, and willpower is a bad risk control. It runs out right when the bot is losing, which is when you need it most.
So I changed the plumbing instead. Trade-level alerts are off. The bot doesn't tell me when it enters, exits or stops out. I get two messages: one if something is broken (an API error, a stuck order, a disconnected exchange), and one weekly summary with the numbers that matter. That's it.
The first week felt wrong. Quiet felt like neglect. I kept opening the exchange app just to confirm it was still running. By week three the urge faded, and something odd happened. I stopped having opinions about individual trades, because I stopped meeting them one at a time.
Try a simple split. Ask which alerts you'd need to act on, and which ones just tell you what already happened. A failed connection needs you. A stop-loss hit at the planned price doesn't, because the plan already handled it. Anything in the second group is a notification that can only make you worse.
Then check the system on a schedule you set in advance. Weekly works for most strategies. Pick the day, look at the numbers against the backtest range, and close the app.
If you're thinking about removing the emotional element from trading, start with the notification settings before you touch the strategy. We publish our full backtest data at v33systematic.com, and the bots run quietly on purpose.