Blended Magnets Forecast

Read this before you use the numbers

These are modelled probabilities from backward-looking signals, not calibrated predictions. Real closes will land outside these bands regularly — and more often than the stated 5%, because the model assumes a volatility that does not change and returns that are log-normal, and real markets are neither.

• The bands are not forecasts of where price will go. They describe where this particular random model puts its paths, given today’s volatility and today’s options positioning. Both change constantly.
• Volatility is assumed constant. It is taken from the last 20 sessions and held fixed for the whole horizon. In practice volatility clusters and jumps, so the real distribution has much fatter tails than the cone shows. The 8% drawdown figure is consequently an underestimate.
• Positioning is frozen. Open interest is a snapshot from Aug 07, 2026, 17:45 ET. The simulation carries it forward as if dealers never re-hedge, roll, or take the other side — which they do, all day.
• The drift tilt is deliberately tiny. A moving-average crossover and a 20-day rate of change are not a forecasting edge, so they are capped at ±8% annualised — a fraction of one day’s noise over the whole horizon.
• Magnet bending is a heuristic. Blended 0.6/0.3/0.1 across gamma, vanna and charm, and capped at 0.3 sigma per day so positioning shapes the paths without driving them. The weights are a reasonable guess, not a fitted or backtested result.
• Nothing here knows about the world. No earnings, no data releases, no policy, no news. A single headline can invalidate the entire cone in a minute.

The Accuracy Log exists precisely because claims like these should be scored rather than trusted.