Trading Strategy Optimizer for TradingView

Optimize any TradingView strategy against every Strategy Tester metric. Parallel backtests, Bayesian search, overfit checks.

ind the parameters your strategy should have been using — without clicking through the Settings dialog three hundred times. Trading Strategy Optimizer reads the Pine strategy already on your chart, sweeps its inputs, and ranks every result against whichever Strategy Tester metric you care about. It is built around two ideas: run more than one backtest at a time, and be honest about which results will not survive contact with new data. HOW IT IS FAST • Several backtests at once. Most optimizers run one, wait, run the next. This one runs your strategy in parallel and collects the results as they land. Measured on a live chart: 1.9× faster on two workers, close to linear. • Reads results, does not scrape them. Numbers come from the same report object the Strategy Tester renders from — exact figures, no HTML parsing, no number-format or language problems, and nothing breaks when TradingView redesigns a panel. • A Bayesian sampler, not a blind grid. TPE models the promising and unpromising regions of the search space separately and spends your cycles where they are worth spending. Seven methods in total: Bayesian TPE, annealing, differential evolution, random improvement, sequential improvement, random and brute force. • Repeat proposals are free. If the search suggests a point it already tested, the cached result is reused and your cycle budget is untouched. WHAT YOU CAN OPTIMIZE FOR Every field the Strategy Tester reports — net P&L, gross profit and loss, profit factor, max equity drawdown and run-up, Sharpe, Sortino, percent profitable, average and largest trade, average bars in trade, commission, margin — each of them for all trades, longs only or shorts only. Choose the greatest or the least of it, or write your own expression. Then filter: keep only candidates with at least 40 trades, at most 25% drawdown, a profit factor over 1.5 — any metric, any number of rules. Candidates that fail are still recorded and shown; they just cannot win. THE PART MOST OPTIMIZERS SKIP An optimizer that only maximises net profit will hand you a curve-fit every time. Four things push back: • A real out-of-sample test. Hold out the most recent part of your tested period — a quarter by default. The search never sees it; every candidate is then checked on it, and you see whether the recommended settings held up on data they were not chosen on. • Quality score, the default objective, is return per unit of drawdown, discounted by how straight the equity curve is, how many calendar windows closed green, whether the later trades held up, and how many trades there are — a handful of lucky trades cannot top the table. • Plateau check re-tests the winner's immediate neighbours and ranks by the neighbourhood. A setting that only works at exactly length=17 and falls apart at 16 and 18 gets demoted, and so does one whose neighbours fail your filters. • An overfit check across every candidate: do the ones that rank well in-sample also rank well on the held-out period? Near zero means the search was ranking noise, and it says so. Monte Carlo resampling of the trade sequence is there too, and it costs no extra backtests — one backtest returns the whole trade list, so all of this is computed locally. RESULTS THAT SAY WHAT TO DO When a run ends, a card shows the recommended settings under their full names, what they were before, how each check went in one plain sentence, and a verdict. Nothing is applied to your chart until you press Apply, and Restore original puts your own settings back. YOUR RANGES, YOUR SPREADSHEET Export Parameter, From, To, Step, Default and Priority as a CSV, edit it in Excel, upload it back. Results export in full — every candidate, every metric, long and short splits — as CSV or JSON. FREE AND PRO Free includes the complete metric list, the Bayesian sampler, brute force and random search, 100 cycles per run and one backtest at a time. It is a working optimizer, not a demo. Pro adds parallel backtests, unlimited cycles, all seven search methods, the Excel templates, result exports, the out-of-sample test, the plateau and Monte Carlo checks, and saved setups per strategy. US$10.89 per month or US$108.92 per year, sold through Gumroad; one licence key activates three browsers. REQUIREMENTS AND LIMITS • A Pine strategy on your chart. An indicator is not enough — the Strategy Tester has to be there. • Chrome 111 or newer. • Parallel backtests use extra indicator slots on the chart, so your TradingView subscription tier is the real ceiling: Basic 2, Essential 5, Plus 10, Premium 25. The extension uses as many as it is allowed and tells you the number it actually got. • Backtests cover whatever testing period the Strategy Tester is set to. Deep Backtesting and custom date ranges are TradingView features on paid plans. • When a run ends — finished, stopped or failed — your original parameters are put back and every temporary copy is removed. If the browser crashes mid-run, the next page load offers to finish that clean-up. TRADINGVIEW'S RULES — PLEASE READ TradingView's help centre says that automation and optimization through external software, scripts, bots and extensions are not allowed, and that it bans accounts for it — temporarily at first, permanently for repeat cases. This extension runs backtests on your chart for you, so using it carries that risk to your TradingView account. The panel repeats this before every run. This is an independent extension. It is not made by, affiliated with, endorsed by, or sponsored by TradingView Inc. "TradingView" is used only to say what this tool works with. All trademarks belong to their respective owners. Nothing here is financial advice. Backtested results are historical and do not predict future performance.