We have made a major upgrade to the SAFE engine.

Until now, price touch probabilities (+2%, +5%, +/-10% in N days) were calculated assuming an “average” distribution of returns.

Now, SAFE explicitly models market regimes and the transitions between them.

What this means in practice

• The market is no longer treated as “the same every day”.
• A regime (range, down fast, low vol, etc.) can persist or change with a certain probability.
• Simulations respect this dynamic, not just raw statistics.

Validation

We validated the upgrade strictly statistically, using Brier score on events (touch / no-touch), not PnL.

Result: better calibrated probabilities, especially for rare events and longer horizons.

What stays the same

• The daily output looks the same.
• The interpretation remains the same.

But what is behind it is now a process model, not just a distribution.

Small change in numbers. Big change in meaning.

Also, the model was retrained, therefore small adjustments will be visible in the chart.

Stay SAFE!