Short answer: Use demo mode to verify that a bot’s orders behave as you expect, then read the trade log critically: sample size, fees, drawdown and whether a few trades carry the whole result. A demo that looks profitable proves the settings run, not that the strategy works live.
What a demo run can and cannot tell you
A simulated account is excellent for catching configuration mistakes: a grid range that is too tight, a safety order that never triggers, a position size that is larger than you intended. It is much weaker at predicting real fills, because simulations are optimistic about price and timing.
So use it as a test of mechanics, not as a forecast of returns.
Five checks on the trade log
First, count the trades: a handful of wins is an anecdote. Second, compare win rate with average win and average loss, since a high win rate can hide a few large losses. Third, add up the fees, because bots that trade often can lose most of their edge to costs. Fourth, find the worst drawdown and ask whether you would have kept the bot running through it. Fifth, remove the best three trades and see whether the result survives.
If you can export the trade history as a CSV, you can run these checks outside the platform and avoid trusting a dashboard summary.
From the demo to live, gradually
When the log holds up, go live with a small allocation, not the full amount, and compare the live log with the demo log after the first batch of trades. Differences in slippage and fees will tell you how optimistic the simulation was.