AI Suggestions

AI Suggestions analyze your strategy or portfolio and tell you, in plain language, what works, what doesn't and how to improve. It's like having an expert reviewer read your results and flag the problems you might not notice on your own — especially the ones that make a backtest deceptively positive.

The analysis gives you an overall grade and risk level, a diagnosis of the strategy type, strengths and weaknesses, concrete recommendations, a read on risk and robustness, and a final verdict. Below, three worked examples show what it produces on very different strategies.

Example 1 — a critical strategy

A poorly defined hybrid (mean-reversion entry, trend-following exit): graded 2/10, risk Critical, negative Sharpe and a high stop-loss exit frequency. The overview flags the design contradictions; the recommendations propose concrete fixes, each to be validated out-of-sample.

Critical strategy: overview, strengths and weaknesses

Critical strategy: recommendations and details by area

Example 2 — an optimal trend-following strategy

A clean EMA-crossover trend-follower on US tech equities: graded 6/10, risk Optimal, excellent risk-adjusted returns. Here the AI covers the full analysis — overview, recommendations, risk metrics, temporal robustness, signals, risk management and behaviour per market regime.

Optimal strategy: overview, strengths and weaknesses

Optimal strategy: recommendations and validation caveat

Optimal strategy: performance details and risk metrics

Optimal strategy: temporal robustness and signals

Optimal strategy: risk management and market regime

Example 3 — a crypto portfolio

A concentrated 100% crypto portfolio (BTC/ETH/SOL): extraordinary returns but severe tail risk and no diversification. The AI reads the risk metrics, the allocation and contribution, and closes with a verdict on long-term sustainability.

Crypto portfolio: performance and risk metrics

Crypto portfolio: allocation, contribution and conclusion

How to use it. AI Suggestions are a tool to better understand your backtests and find weak points, not an oracle. Every suggested change must be validated out-of-sample (walk-forward), on data not used to build the strategy, before trusting a result.