Both products are sold with the same word. A vendor with a Pine Script on TradingView and a vendor with a compiled robot for MetaTrader will both say "algo", both will show you entry logic and filters, and both will talk about an edge. What they hand over is a different category of software, and the two categories fail in opposite ways.
An indicator calculates and shouts
An indicator is a calculation drawn on a chart. It reads price and volume history, applies its rules bar by bar, and outputs something you can see: a line, a shaded zone, an arrow, a colour flip. In Pine Script that means a study style script; in MetaTrader it means a custom indicator compiled to .ex4 or .ex5 and dropped into the Indicators folder. Neither has permission to send an order. The platform simply does not expose that function to an indicator, which is a design decision rather than an oversight.
What an indicator can do is raise its hand. TradingView alerts fire the moment a plotted condition becomes true and push the message to email, mobile notification, or a webhook URL. That webhook is how most modern signal services deliver: the alert hits a relay server, the relay posts a formatted message into a Telegram channel, and subscribers see entry, stop loss and target within seconds of the candle event. If you have set up TradingView webhooks before, you already know the whole chain is three moving parts and any of them can be down while the chart still looks perfectly healthy.
The human stays in the loop. Someone reads the alert, checks whether the market is where the alert says it is, sizes the position and clicks. That step costs seconds or minutes, and those seconds are the price of keeping judgement in the system.
An EA holds the order permission
An expert advisor is a program with trading rights inside the terminal. It runs on every incoming tick, evaluates its conditions, and calls the platform's order functions directly. It can open, modify, partially close, trail a stop, add to a position and shut everything down at a session boundary. Nobody has to be awake. Our longer piece on what expert advisors actually do covers the mechanics; the operational point is that an EA only exists while its terminal is open and connected.
That is why EA users end up renting a forex VPS. A laptop that sleeps, updates overnight or drops Wi-Fi for ninety seconds leaves live positions with no manager. The EA does not know it missed those ticks. It resumes on the next one and acts on a chart state it never saw develop.
Automation does not reduce market risk, it changes who is holding the mistake. A wrong rule executed by hand costs one trade. The same wrong rule inside an EA costs every trade it takes until someone stops it. Leveraged trading carries a high risk of loss either way.
Testing them is not the same exercise
An indicator can be checked by eye. Load it on a chart, scroll back, and see where the arrows landed. The failure to watch for is repainting: a script that changes its historical output once later bars arrive, so the backtest looks clean and the live alerts do not match it. Any signal that references the current bar's close before that bar has closed is a candidate.
An EA needs a proper engine. The MT5 strategy tester will run it over historical ticks with modelled spread and commission, which is closer to reality than eyeballing but still an approximation. Real-world spread widening at rollover, slippage on news, and requotes rarely appear in a tester at their true cost. Our notes on backtesting basics go into the assumptions that quietly inflate results.
| Dimension | Algo indicator | Expert advisor |
|---|---|---|
| Can send orders | No | Yes |
| Runs where | Vendor chart or your chart, alerts server-side on TradingView | Inside your terminal, so a VPS in practice |
| Main hidden fault | Repainting and alert lag | Silent failure after a disconnect or symbol change |
| Who sizes the trade | You | The code, from its parameters |
| Time cost per day | Minutes of attention per alert | Setup and monitoring, not per trade |
The failure modes barely overlap
Indicator problems are human problems. Alerts arrive while you are in a meeting. Twenty alerts a day turn into ignored alerts by Thursday. The signal said enter at 2,341.50 and by the time you looked, price was 2,346 and you took it anyway. None of that is the script's fault, which is exactly why it keeps happening.
EA problems are infrastructure problems. The broker adds a suffix to the symbol name after a server migration and the EA can no longer find its instrument. A parameter set that was tuned for a 1.2 pip average spread runs on an account quoting 2.4. The terminal reconnects after an outage and the EA reopens a position it already has. These are quiet failures. There is no error popup on your phone; there is a balance curve that looks wrong a week later.
Choosing by what you want to keep
The useful question is not which is better. It is which part of the process you want to own. If you want to keep the final judgement, the entry timing and the position size in your own hands, an indicator with alerts is the correct tool and an EA is the wrong one. If you have a rule set you can state without any "depends" clauses, and you have somewhere reliable for it to run, an EA removes hesitation and fatigue from the execution.
For a trader holding down a job, my honest position is that indicator plus alerts wins most of the time. An EA is an operations project: hosting, monitoring, version control, a plan for what happens when the broker changes something. Traders who skip that part are not automated, they are unattended. If you are buying rather than building, the same due diligence applies to both, and our checklist on evaluating a signal service transfers almost line for line to robot vendors.
"Ask the vendor one question: what happens when your terminal disconnects for ten minutes with a position open? If they have no answer, you are buying a chart script with a robot's price tag."
— Alex Onta, Executive Director, SINGUARD
Key Takeaways
- Indicators plot and alert; expert advisors hold the platform's order permissions. Everything else follows from that split.
- Repainting is the indicator fault to hunt for; silent failure after a disconnect or symbol change is the EA equivalent.
- An EA without hosting and monitoring is unattended rather than automated, and its mistakes repeat until a human intervenes.
- If your rules still contain a "depends", keep the human in the loop and use alerts rather than full automation.
Frequently Asked Questions
Can a TradingView indicator place trades by itself?
No. Pine Script indicators plot and alert, they do not hold order permissions. Trades only happen if an alert is routed to a broker connection or a third-party bridge, and that bridge, not the indicator, becomes the part that can fail silently.
Do I need a VPS to run an expert advisor?
Practically, yes. An EA only exists while its terminal is open and connected, so a laptop that sleeps or loses Wi-Fi leaves positions unmanaged. A VPS keeps the terminal running and shortens the hop to the broker server.
Which is safer for a beginner, an indicator or an EA?
An indicator with alerts keeps a human between the signal and the order, so mistakes stay small and visible. An EA can repeat the same flawed decision for hours before anyone notices. Neither removes risk: leveraged trading can lose more than the intended amount on any given day.