Strategy Backtesting with Pineify's Deep Report Features
Pineify's Backtest Deep Report analyzes exported TradingView trades. It can help you inspect how results vary across trades and periods, but it does not rerun the strategy, verify its signals or establish that the strategy will remain profitable. The calculations depend on the exported fields, starting capital and the report's metric definitions.
Prepare and Reconcile the CSV
Save the strategy version, symbol, chart type, timeframe, date range, commission, slippage and position settings before exporting the trade list. TradingView's strategy documentation explains the broker emulator and why historical orders are simulated fills.
Open Pineify's Backtest Deep Report and import the supported CSV. Set initial capital to match the source test; the current interface's default is 1,000,000 and should not be assumed correct for your account. Then compare trade count, net profit, currency and several individual entry/exit pairs with TradingView.
The current parser groups records by trade ID and takes the first matching entry and exit. Some missing numeric fields fall back to zero, and a missing currency can fall back to USD. Inspect partial exits, open trades and unusual exports carefully. A successfully loaded file is not proof that every record was interpreted correctly.
Local Analysis and Optional AI Analysis
CSV parsing and the dashboard's numerical calculations run in the browser. However, selecting AI Analysis sends derived metric summaries to Pineify's analysis endpoint. The payload includes metrics and summaries of monthly returns, rolling statistics and Monte Carlo results. It does not include the raw trade list in that request, but derived performance data still leaves the browser.
Therefore, the entire workflow should not be described as local-only or as data that never reaches a server. Review the current product's access and privacy information before using AI Analysis. This article does not establish a free entitlement or a promise about server retention.
Read Metrics According to Their Calculation
| Report value | Interpretation and current limitation |
|---|---|
| Win rate and profit factor | Frequency of profitable closed trades and gross profit divided by gross loss; neither describes the size of the next loss. |
| Sharpe ratio | Calculated from trade percentage returns and annualized using estimated trades per year. It is not a daily portfolio-return Sharpe ratio. |
| Sortino ratio | Uses the root mean square of negative trade returns in its denominator. This differs from definitions using all observations below a specified target. |
| SQN | Uses the mean and standard deviation of percentage trade returns. It does not certify statistical reliability or use independently supplied risk multiples. |
| VaR and CVaR | The lower fifth percentile of observed trade returns and the average of returns at or below that threshold. These are not maximum daily losses. |
| Calmar and CAGR | Depend on initial capital, the observed date interval and reconstructed equity. Short intervals can produce unstable annualized figures. |
The risk-free input defaults to zero. The current calculation subtracts the supplied value directly from mean trade return without converting an annual rate to a per-trade rate. This is an identified calculation limitation: entering an annual percentage should not be assumed to produce a correctly adjusted ratio. Zero denominators can also produce fallback values rather than meaningful estimates.
Treat Kelly-style sizing output as a sample-derived statistic, not an instruction to allocate that fraction. Estimation error, changing payoffs, leverage and correlated losses can invalidate the assumptions behind it.
Rolling Analysis Describes Recent Observations
The current rolling window contains 20 trades. Twenty trades can span very different calendar durations in active and inactive markets. Adjacent windows share most observations, so their points are not independent evidence.
Rolling Sharpe and Sortino are not annualized in the same way as the headline metrics. Compare like definitions before interpreting a difference. A decline describes the selected observations; it does not demonstrate an early-warning ability or specify when to stop a strategy.
Record a review rule before inspecting the chart. For example, investigate whether a change coincides with higher transaction costs, a different instrument mix or a specific market regime. Test any proposed adjustment on a later period that did not influence its design.
What the Monte Carlo Simulation Actually Does
The tool runs 1,000 bootstrap samples of the observed dollar profit/loss values, sampling with replacement. A trade outcome can appear several times or not at all in a sample. This is different from merely shuffling the same trades, and it is not a simulation of 1,000 new market price histories.
Each sampled dollar result is added to equity without recalculating position size. The model does not preserve the original sequence, volatility clustering or changing exposure. It cannot generate a loss type absent from the input data. Only the first 50 paths are retained for the displayed curves and path export.
There are also identified implementation limitations that matter when reading the labels:
- The reported risk of ruin counts paths whose final equity is zero or negative. A path that falls below zero and later recovers is not counted by that test.
- Fields named best-case return at 95% and 99% currently use the lower fifth and first percentiles of terminal equity. They should not be interpreted as upside estimates.
- Drawdown percentiles summarize this resampling model, not a guaranteed bound on future account losses.
Use these outputs as a sensitivity exercise and retain the input file and settings. They cannot establish a safe leverage level.
Use MFE and MAE to Form Testable Questions
Maximum favorable excursion describes the largest favorable move while a trade was open; maximum adverse excursion describes its largest adverse move. Check the chart's axis labels and units rather than inferring them from a point's position.
An excursion plot does not show what happened after an exit. A losing trade with a large favorable excursion may motivate an exit-rule experiment, but the plot alone cannot establish that a tighter stop or earlier target would have improved the entire strategy. Intrabar order and transaction costs also matter.
Export a Reproducible Review
The current Excel export includes multiple analysis sheets; its Monte Carlo path data contains the retained 50 paths, not all 1,000 simulations. Keep the original CSV alongside the workbook and record the capital input, side filter and analysis date.
A useful review ends with a specific unresolved question: which trades drive the result, whether costs change the conclusion, or whether a revised exit survives a later sample. Open Pineify to inspect the available report tools and current access requirements.
