How the forecasting engine works

Input vectors

Four data families feed each daily range. Weights are fixed for the session.

• Historical price record

  40%

  Rolling 20-, 50- and 200-session WTI closes anchor the centre of the range. Sessions closer to today carry more weight through an exponential decay.

• Momentum vector

  25%

  Rate of change across the last 5 and 20 sessions. Sustained directional momentum widens the range asymmetrically in the direction of travel.

• Realised volatility band

  20%

  Standard deviation of the last 20 daily returns sets the half-width of the published low/high band. Quiet tape means a tighter range.

• Inventory variance

  15%

  Official crude inventory levels compared against the five-year average. Larger-than-normal draws shift the range higher; builds shift it lower.

How a range is produced

  historical closes ──┐
  momentum vector  ───┼──▶  weighted centre  ──┐
  inventory variance ─┘                        ├──▶  published low / high
  realised volatility ───────▶  band width ────┘

Standing rules

• ›The range is published before the session opens and is never revised after the fact.
• ›Every published range is written to the public archive alongside the eventual close.
• ›A session counts as a hit only when the actual close falls inside the published band.
• ›No discretionary human override is applied to a published range.