Times. All Singapore time; the window is 08:00 to 08:00 SGT. The day-on-day move is against the previous session's close, so a weekend gap still shows even though it happened outside the window.

Charts. Bars are resampled from 1-minute data, each column aggregated separately, so the wicks are real. Levels are ratio-adjusted across contract rolls — read the shape, not the old prints. Spreads are the exception: unadjusted front-month closes. Indicators are textbook defaults, never tuned: Bollinger 20/2.0, MACD 12/26/9, RSI 14.

News. Picked by relevance, ordered by time, with a per-feed cap (x ≤ 10, newsapi ≤ 8, rss ≤ 12) relaxed only to fill the list. Timestamps are when the story was filed, not when the event happened — a piece filed today about last month's outage carries today's time, and wire items reaching us through an aggregator cannot be traced back to the publisher's own clock. Stories already carried in the past seven days are dropped. A dashed circle on the timeline means the source gave the hour only, so it carries no price line.

Value areas. The band holding 70% of the volume traded in the window named. It says where business was done, not where price goes next: across five years of both grades, being above, below or inside value says nothing about the next session's range.

Scope. A calibrated range forecast, the levels people watch, and what each is worth against a random walk. It does not publish a point target or directional call and does not give trading advice. A driver with no source here is named as not covered, never replaced by the nearest number we happen to hold.