Pine Script backtesting tools compared: wavealgo vs TradingView vs Backtrader

If you have written or bought a Pine Script strategy, there are three realistic ways to find out whether it holds up. Test it inside TradingView. Port it to Python and run Backtrader or QuantConnect. Or put it through the same fixed grid every script on the wavealgo leaderboard went through. The three answer different questions, and the usual mistake is asking one of them the question another was built for.

TradingView's Strategy Tester evaluates one symbol and one timeframe at a time. You can switch the chart to QQQ and re-run, then to EURUSD and re-run, but there is no fixed universe and no shared fee model, so the numbers you collect are not comparable to anyone else's — or to your own from last month, if a setting moved in between. The deeper problem is what the tool shows you by default: a strategy tuned on one chart looks its best on exactly the chart you are looking at.

Backtrader and QuantConnect give you the opposite trade-off. You own the execution model, the fee schedule, the slippage assumption and portfolio-level risk. You also have to port the Pine by hand, function by function, before anything runs at all. For a strategy you have already decided to take seriously, that work is the right work. For deciding whether to take it seriously, it is the wrong order of operations.

wavealgo compiles Pine directly — resin, an open-source Pine compiler under Apache-2.0 — and runs one grid that does not move between scripts: 6 markets (BTCUSD, ETHUSD, SPX500, QQQ, GOLD, EURUSD) × 3 timeframes (1H, 4H, 1D), out-of-sample only, 0.05% fee per side, next-bar-open fills, long-only, 100% of equity per entry, 100,000 initial capital. That fixity is the entire argument for it. Two scripts audited months apart are still on the same yardstick.

What comes out is an alpha score from 0 to 100 — edge up to 40 points, stability up to 25, a flaw penalty down to −25, reproduction up to 10 — and one of four verdicts. Verified requires a composite Sharpe of at least 0.9 with at least 12 of the 18 cells positive. Conditional means a composite Sharpe above zero with at least one cell clearing 0.9: an edge that exists somewhere and not everywhere. No alpha is everything else. Repaints is not a performance result at all — it is a static finding in the code, a repaint or a look-ahead the compiler can see without running a backtest.

The trade-off wavealgo makes is control. You cannot set your own fee schedule, inject a slippage model, or run portfolio-level correlations. A configurable grid would produce scores nobody could compare, which would defeat the purpose of publishing them. If your question is which broker fee schedule makes a strategy viable on your capital, Backtrader is the right tool and always was.

The playground is a separate thing and worth being explicit about: it compiles and runs Pine in your browser and draws the chart. No server, no account, and no score — scoring is an offline pipeline that produces the leaderboard, not something the playground does while you wait. Everything on the site is free; there are no paid tiers and no run quotas.

So: TradingView to develop and eyeball, wavealgo to see how a script does on a grid it did not get to choose, Backtrader once you have decided and need the execution model to be yours. Mostly they do not compete.