We bring publicly available strategies into one controlled research environment, so differences in results come from the strategy logic—not from different data, costs or backtest settings.
We research popular, publicly available strategies from online communities such as TradingView. The original author is credited on every strategy page. We select strategies because they are widely followed—not because their published backtests look profitable.
Each strategy is tested with fixed rules. We verify signal timing, indicator calculations and order execution before publishing its results.
Strategies use the same normalized market database, trading sessions and bar-building rules. A 15-minute result cannot gain an advantage from cleaner or differently aligned data than another 15-minute result.
Every standalone strategy is tested with the same published starting-capital convention and fixed parameters. Users cannot tune parameters or choose a flattering custom start date after seeing the result.
Net performance includes market-specific commission and slippage assumptions. We rank and compare net results; gross performance remains visible only to show how much trading friction changed the outcome.
The benchmark uses the same dates and fixed position quantity as the strategy; for equity-sized strategies it invests the initial capital. Fixed-quantity stock benchmarks can therefore be leveraged, and are not an unlevered investment of US$100,000 at the selected start date. Vs. buy & hold is strategy return minus benchmark return, in percentage points.
Starting accounting capital is US$100,000. Most engines use a constant position size: stocks use round(100,000 / the first 5-minute closing price in the full input dataset), calculated before the selected start-date filter. The resulting share count stays fixed. Futures use one contract; FX and crypto use one configured unit with the market multiplier.
This means a 2024 stock test can retain a share count based on 2021 prices. If the stock has risen, entry notional can exceed starting capital. Position size is not reduced as equity falls. These are fixed-exposure research simulations, not a guarantee of available buying power; financing, margin calls and forced liquidation are not simulated.
Example: the NVDA Hull Suite 15m test from 2024 records 7,686 shares. At US$0.017 per share round trip, the reference cost is US$130.662 (US$130.66 rounded). The detail page shows the recorded quantity and first-entry notional relative to starting capital.
Exceptions: Flawless Victory and Ichimoku + Hull + MACD use 100% of the engine’s current simulated equity at entry, so quantity varies. These engines run on the input history before trades are filtered to the selected start date. Their actual average recorded cost is shown separately from the base-quantity reference cost.
When equity reaches zero, subsequent simulated gains are not a valid return. Capital-depleted runs are listed last, full-period return and ratios are suppressed, and their detail equity chart is withheld. Comparison growth curves remain at zero after depletion. Raw research records are retained for audit; this display update does not recalculate historical backtests.
A custom cutoff preserves the published strategy decisions, position sizes and cost assumptions up to that date. A position still open at the cutoff is closed at the last available bar close, with the recorded round-trip cost. Equity, trade statistics and the benchmark are recalculated; headline returns are not simply copied from the full-period result.
On non-trading days the last available bar is used. Daily settlement prices are checked against the published benchmark before a dated result is produced. Stored trade amounts are rounded to cents, so very small rounding differences from a full engine replay are possible. The latest published result remains unchanged. Choosing a historical cutoff is not a forward test and does not eliminate selection bias.
Instead of collecting attractive screenshots from unrelated backtests, we test popular public strategies with one data standard, realistic costs, fixed choices and visible benchmarks across supported markets and five timeframes.
A backtest is still a historical simulation, not a prediction. Liquidity, financing, taxes, market impact and future regime changes can make live results materially different.