One EA on many charts, or one chart scanning many symbols?

[Fairman]

One EA on many charts, or one chart scanning many symbols?

If you want an EA to trade several pairs, there are two common setups. Attach the EA to a chart for each pair, so each instance trades its own symbol. Or attach it to one chart and have it scan all the pairs from there.

The one-chart-per-pair setup is simpler to write. Each instance only cares about its own symbol, uses its own chart's data, and runs when that symbol ticks. It's easy to see what's happening on each pair. The downside is managing many charts, and keeping settings consistent across them. It also makes any logic across pairs, like limiting total exposure, harder, because instances don't know about each other without extra work.

The single-chart scanner is more complex. The EA has to request data for each symbol, which is slower and can cause issues if data isn't loaded. It runs on the ticks of the chart it's attached to, so a quiet pair on the main chart means slow reactions for others, unless you use a timer. But it's easier to manage total risk across all pairs in one place.

I'd go with one chart per pair for most strategies, with a shared file or global variables for anything that needs coordination. The scanner approach makes more sense when cross-pair logic is the main thing.

[Fairman]

Backtesting is another factor. MT5 can test multi-symbol EAs, but the tester has to load data for every symbol, and results can depend on how it synchronises ticks between them. Single symbol EAs are faster to test and easier to verify. If your strategy doesn't need cross pair logic, testing each pair separately and combining the results is simpler and gives you more confidence that each pair works on its own. MT4's tester handles multi-symbol EAs poorly, which pushes people towards the one chart per pair setup there. Simpler testing is a good reason on its own.

[Fairman]

If you go with one chart per pair, use different magic numbers per instance or filter by symbol everywhere, so instances don't touch each other's trades. A total risk check across instances can be done by counting all open positions in the account, regardless of symbol, before opening a new one. That covers the most important cross-pair rule without needing a scanner. It's simple, and it prevents the situation where five instances all open a trade at once on correlated pairs. One extra check before each new order handles the most common risk without much code.

[FXS Prop Desk]

One chart per pair is easier to debug, since each instance has its own log and settings. The one-chart scanner is more efficient but harder to troubleshoot, because a bug on one symbol can affect all of them.

A scanner also has to handle ticks from other symbols, which do not trigger OnTick on its chart. Using a timer with OnTimer is a common solution, but it means the EA checks prices at fixed intervals rather than on every tick.

[Fairman]

Resource use is another difference. One EA per chart means the terminal loads one copy of the EA code for each chart, and each copy processes its own ticks. With many charts, that adds up, especially on a small VPS. A single EA scanning many symbols uses one copy but has to request data for each symbol, which has its own cost. I'd test both on the VPS you'll actually use, watching CPU and memory. The better design on paper isn't always the one that runs smoothly on a cheap server.

[Shadow Trader]

A platform with thirty charts, each with three indicators, is doing a lot of work on every tick. Each tick triggers recalculation, redraw, and sometimes script execution for each of them.

Under normal conditions, a modern computer handles it. During a fast market, when ticks arrive in bursts of dozens per second, the processing queue grows. The terminal's user interface slows, and clicks register a moment late.

The lag can be 200 to 500 ms, which in scalping is the difference between a clean exit and a missed one.

Several culprits are common. Custom indicators that recalculate their entire history on every tick are the worst. Charts on symbols that aren't actually being traded, but kept "just in case". Large chart history settings, where the platform keeps hundreds of thousands of bars in memory for no reason.

A cleanup helps a lot: close unused charts, cut the history to what's needed, and remove indicators that don't earn their place. Some traders run a separate lean terminal for trading and a heavier one for analysis.

Checking the CPU usage during a news spike is a quick way to find out whether the platform is the bottleneck.

How many charts and indicators does the main trading terminal carry, and has trimming it made a difference?

[Fairman]

Is error handling simpler with one approach than the other? With one chart per pair, a crash or error in one instance usually only affects that pair. With one EA scanning everything, a bug in one symbol's logic could stop the whole EA, affecting all pairs at once. That's a good argument for one chart per pair if reliability matters most. On the other hand, a single EA makes it easier to manage total risk across pairs, since it knows about all positions in one place. Neither design is clearly better in every case.