QQQ · Intraday outlook
Connecting automatically…
Data, timestamps and chart notes
WebSocket · automatic updates
Minute chart and frequencies in similar past sessions. Not an option-profit forecast.
Blue: completed-minute closes. Dashed levels: barriers from the calculation price, not from the live bid/ask midpoint. Labels show the chance of observing each level within the selected horizon. A touch and the opposite touch may both occur.
Live chart: 04:00–20:00 New York time. Amber is premarket, blue is the main session, purple is after-hours; the main-session boundary follows the exchange calendar, including early closes. Only completed-minute closes are shown; gaps are not filled. Historical replay remains main-session only. Percent zero is the verified opening price; before that is available, the exact previous trading-session close is used and labeled. Without a verified reference, only dollars are shown. Extended hours do not enter the current probability model.
Live transport: one shared Massive WebSocket stream feeds the server; this page receives pushed quotes and completed-minute updates through the website WebSocket. No HTTP polling in Live mode. REST is used for initial snapshots and minute-history recovery, not for ongoing quote refreshes; an initial REST quote is labeled separately. Quotes do not modify the model’s minute-close inputs. Massive · WebSocket quotes · WebSocket · minute OHLC
Waiting for scenario data
Connection and updates are automatic. No refresh button is needed.
These three scenarios are mutually exclusive and sum to 100% before rounding. Events are observed at minute closes; intraminute touches are not included. Estimates are historical frequencies, not guaranteed future probabilities.
Touches in either order
- ↑
- —
- ↓
- —
- Both barriers
- —
At the end of the horizon
- Above current +0.05%
- —
- Below current −0.05%
- —
- Within ±0.05%
- —
Reversal to the session open
—Has this helped on later days?
Walk-forward · no future informationLower Brier is better. The score concerns all three first-barrier scenarios, not trading return. Each day has equal weight after averaging its tested times (10:00/12:00/14:00). A small score advantage does not prove a stable edge.
By year and probability calibration
| Year | Days / cases | Analogue Brier | Baseline Brier |
|---|
First-UP probability bins: mean predicted percentage versus actual frequency in later observations. Several observations can belong to one day; they are not independent days.
| Bin | Days / cases | Predicted | Observed |
|---|
Closest historical examples
Same instrument and session length, same minute, earlier dates only. Selection never uses how those days ended.
| Date | Move from open at that time | Feature distance |
|---|
Method, coverage and limitations
Four observed features: movement since open, movement over the last 15/30 minutes, and realized volatility so far. Up to 80 nearest past days, at least 40 matches out of at least 160 prior sessions. Rare states are left without an estimate. The model starts at 10:00 ET.
30/60-minute horizons are exact. If too little session time remains, choose “Until RTH close”. All movement levels refer to the latest completed-minute close. Return to open is estimated only after at least 0.1% excursion and among at least 30 similar days with the same excursion direction.
Quote midpoint and its timestamp are separate from the minute used for scenarios. No forecast is issued outside RTH, on delayed/error feeds, with missing minutes, or with more than two minutes of missing price progress. The graph and quote timestamps remain visible.
Intervals beside frequencies are descriptive 95% Wilson intervals, not guarantees of prediction accuracy. Serial dependence and market changes can widen practical uncertainty. Replay validation also excludes the replay date and all later dates.
The training archive is a dated, verified publication; live prices refresh independently. This is not an automatically retrained trading model. Public redistribution of quotes must comply with your data-provider license.
Massive · OHLC · Massive · bid/ask · Probability calibration