What is TradingView backtesting good at?
TradingView is unmatched at iteration speed. You can go from a chart observation to a rule set to a strategy report in a few minutes, on any instrument the platform carries, without installing anything.
The charting is the best in the retail market, the Pine Script language is deliberately small and learnable in an afternoon, and the community library means most standard ideas already exist in some form you can read and modify rather than write from scratch.
For exploration — does this idea look like anything at all? — that speed genuinely matters. Rejecting fifteen ideas in an evening is a legitimate use of a tool with approximations, because approximations rarely rescue a strategy that is obviously broken.
Where does TradingView's strategy tester flatter results?
The tester flatters results in three specific places, and all three become more significant the shorter your holding period.
Intrabar behaviour is approximated. On a standard backtest the engine sees OHLC for each bar and applies an assumption about the order in which those prices occurred. If your stop and target both sit inside one bar, that assumption decides the trade. Higher-resolution intrabar checking is available but is not the free default, and most published results do not use it.
Costs are opt-in. Commission and slippage default to zero unless you set them in the strategy properties. A zero-cost report on a strategy trading several times a day is not a slightly optimistic estimate — it is a different strategy that does not pay to trade.
Repainting. Pine has well-known ways to accidentally use information that would not have existed yet: `security()` calls without proper offsetting, indicators that reference the current unclosed bar, or strategies that evaluate on a bar which later changes. The tester reports these results with the same confidence as clean ones.
How do you sanity-check a TradingView strategy report?
Check five things before you believe a Pine strategy report, in this order, because each one kills more results than the next.
Number of trades. Below a few hundred, treat the metrics as provisional regardless of how good they look.
Costs. Open the properties and confirm that commission and slippage are set to something realistic for your broker and instrument. If they are zero, the report is fiction.
Percent of equity versus fixed size. Sizing settings change the equity curve far more than most rule tweaks, and comparing two strategies under different sizing settings tells you nothing about the strategies.
A single dominant trade. Look at the trade list sorted by profit. If removing the best two or three trades destroys the result, you have a lottery ticket rather than an edge.
Out-of-sample. Set the date range to an earlier period, fix the parameters, then run the later period once. If performance evaporates, the parameters were fitted — which is the normal outcome and the reason for checking.
Is TradingView enough on its own?
TradingView is enough for exploration and insufficient for a final verdict, which is a reasonable place for a tool to sit as long as you know which half you are in.
The gap is not the language or the charting — Pine is capable and the charts are excellent. The gap is that nothing in the workflow insists on the discipline that separates a real result from a fitted one: no enforced split, no trial counting, no default costs, no statement of what the engine could not resolve inside a bar.
A sensible pattern is to use TradingView to find and shape ideas, then re-test the survivors somewhere that resolves intrabar behaviour with finer data and applies the out-of-sample split without being asked. The ideas are the scarce part; the verification is the part worth automating.
Can you trust published Pine Script strategies?
Treat published strategies as ideas to test rather than results to believe, because the incentives around publishing select for impressive backtests rather than robust ones.
A script with a spectacular equity curve in its description was, in almost every case, tuned until the curve looked like that on the instrument and period shown. That is not necessarily dishonest — it is what happens when the feedback loop rewards screenshots.
The useful move is to take the logic, strip the specific parameter values, set realistic costs, and test it on instruments and periods the author did not use. What survives that is worth your attention; what does not was never a strategy, just a well-chosen chart.