AI Agent
The Backtest AI Agent brings the same analytical intelligence as the Trading Analytics AI Agent, but focused entirely on backtest data. It can analyse individual backtests, evaluate entire portfolios, detect overfitting, run Monte Carlo risk projections, and — most powerfully — construct optimal portfolio combinations with recommended weights and exclusions that can be imported directly into your Backtest Portfolios.
The agent automatically selects the appropriate Claude model based on task complexity, with the model label shown next to the response header.
Backtest Context
Before starting a conversation, select the analysis context from the Backtest Context dropdown:
Single Backtest — analysis focused on one selected backtest report.
Multi Backtests — select multiple backtests from the list in the left panel. Each entry shows the strategy name, symbols, date range, net profit, and profit factor — giving full context at a glance. The agent analyses all selected backtests simultaneously and can compare, rank, and reason across the entire dataset. The context counter at the bottom shows how many backtests are currently selected.
Portfolio Builder — a dedicated mode for constructing optimal portfolio combinations. Select two or more backtests, click Build Portfolio, and the agent will analyse their metrics, compute inter-strategy correlations, assess overfitting risk for each strategy, run Monte Carlo risk projections, and produce a structured recommendation — including which strategies to include, which to exclude, and what weights to assign to each.
What the AI Agent Can Do
Individual backtest analysis — a full structured review of a single backtest: statistical metrics, overfitting risk assessment (win rate suspicion, trade count adequacy, history quality, single-symbol exposure), Monte Carlo risk projection, and a plain-language verdict with flagged items.
Multi-strategy comparison — when multiple backtests are selected, the agent ranks them by risk-adjusted performance, identifies the strongest and weakest strategies, and explains the reasoning.
Overfitting detection — the agent flags statistical anomalies that are common indicators of curve-fitting: unusually high win rates, low trade counts relative to the number of parameters, short test periods, single-symbol dependence, and asymmetric risk profiles (many small wins with occasional large losses).
Monte Carlo risk analysis — the agent extracts trade parameters from the backtest history (Avg Win, Avg Loss, Win Rate, Trades/Month, Win:Loss Ratio) and uses them to project forward risk scenarios, producing a structured assessment of expected drawdown range and probability of ruin.
Portfolio construction (Portfolio Builder) — the agent analyses strategy correlations, evaluates each strategy's individual merit, and produces an optimal portfolio recommendation with specific weight assignments. Strategies that are highly correlated, overfitted, or underperforming are flagged for exclusion with clear reasoning.
Importing AI-Recommended Portfolios
When the AI Agent produces a portfolio recommendation in Portfolio Builder mode, you can import it directly into your Backtest Portfolios with a single click — without manually recreating the composition, weights, and exclusions. The imported portfolio appears immediately in your portfolio list, ready for further analysis, Monte Carlo simulation, or sharing.
Chat History
All previous backtest analysis sessions are saved and listed in the left panel under Chat History, grouped by recency. Each session is labelled with a summary of the query. Click + New to start a fresh conversation. Sessions are preserved independently so you can return to any previous analysis at any time.
The AI requests counter at the bottom of the panel shows monthly usage against your plan limit.