Methodology
Each axis on the pentagon is scored 0 to 5. Below is what is measured and where the score came from. Update us if a number is wrong — every cell links to a public source we can re-check.
Native speed
whatThroughput on the 50,000-bar reference backtest, measured wall-clock from 'compiled' to 'final report'.
scoringPineForge: in-tree benchmarks at HEAD. TradingView: not user-controllable, scored 2 because runtime is shared and rate-limited. MQL5: native compiled MQL5 4. QuantConnect: LEAN cloud throughput 3. Backtrader: Python interpreted 1.
Source privacy
whatCan the seller ship a strategy without exposing its source to the buyer?
scoringPineForge: compiled .so binary, can be reverse-engineered with effort, 4. TradingView: invite-only scripts hide source from non-invitees, 2. MQL5: encrypted .ex5, considered the strongest in the market, 5. QuantConnect: source is shared by default, 1. Backtrader: pure Python, no protection, 0.
License control
whatHow many dimensions can the seller bound? Time, machine, broker, symbol, input ranges.
scoringPineForge: all five planned, 5. TradingView: invite list only, 1. MQL5: account-locked + time, 3. QuantConnect: subscription only, 1. Backtrader: nothing, 0.
OSS audit
whatCan the security or compliance team read the runtime that runs your money?
scoringPineForge: Apache-2.0 runtime, 5. TradingView: closed, 0. MQL5: closed, 0. QuantConnect: LEAN engine OSS, but cloud is closed, 3. Backtrader: MIT, 5.
Data freedom
whatCan you backtest on any OHLCV data — your CSV, alt data, custom timeframe — without the platform restricting symbols or data sources?
scoringPineForge: any OHLCV CSV/Parquet, any timeframe, fully offline, 5. TradingView: TV-listed symbols and broker integrations only, no custom data upload, 2. MQL5: broker-served history, custom imports limited to MT4/5 format, 3. QuantConnect: Lean data libraries plus custom universe API, 4. Backtrader: arbitrary DataFeed adapters, fully BYOD, 5.