A strategy that looks good in your head isn't the same as one that's proven to work. Before any strategy touches real money, it needs to survive two separate checks: backtesting against historical data, and demo trading against live-but-fake-money market conditions. They catch different kinds of problems, and skipping either one is how a fatal flaw gets discovered with real capital on the line instead of before it.
What backtesting actually checks
Backtesting runs your strategy against historical price data to see how it would have performed. It's not a guarantee of future results, but it's a fast, free way to catch the obvious problems: a stop loss that's way too tight for the instrument's typical volatility, a condition that almost never triggers, or one that triggers so often it churns through fees. MT5 and MT4 both have a built-in Strategy Tester for this; cTrader has its own backtesting under Automate. Point either one at your exported file and a date range, and it replays every trade your rules would have made.
What backtesting can't tell you
Historical data is cleaner than live markets. A backtest doesn't experience slippage the way a real order does, doesn't see requotes, and can't tell you whether your broker's execution during a volatile news release holds up as smoothly as the history file assumes. It also can't catch a bug that only shows up once the platform is genuinely running your EA or cBot tick by tick, rather than replaying stored data in seconds. A strategy can backtest beautifully and still fail this second, different kind of test.
Demo trading: the check backtesting can't replace
A demo account trades fake money against the platform's real, live price feed - so unlike a backtest, it exposes your strategy to actual execution conditions: real spreads, real order fills, real connectivity hiccups. It's slower than backtesting, since you're waiting for the market to actually move rather than replaying history in seconds, but it's the only step that catches problems that only appear when a strategy is genuinely running unattended against a live feed. A strategy that survives a few weeks on demo, still doing roughly what the backtest predicted, has cleared a bar backtesting alone can't set.
Where this fits in the workflow
Every file AlgoPuzzle exports comes with a README that says the same thing for a reason: always test on a demo account first, and run it through your platform's own backtester before that. Neither step happens inside AlgoPuzzle itself - the builder's job ends once your rules are defined and a working EA or cBot is generated. Proving it actually behaves the way you expect is a job for MT5's or MT4's Strategy Tester, cTrader's backtesting, and then time on a demo account. Going straight to a live account skips the two checks most likely to catch a real problem while it's still free to find.