Price chart

Candles and controls

• The terminal renders one chart using TradingView’s open-source Lightweight Charts library. Traded candles, forecast candles, drawings and indicator lines share the chart’s coordinate APIs. Binance USDⓈ-M kline streams update the forming traded candle while the tab is visible, with screen refreshes capped at twice per second. The /fapi/v1/klines endpoint seeds and reconciles history every 15 seconds and supports polling during a stream outage. Native intervals load up to 1,000 base candles. Custom intervals backfill bounded older history, up to 12,000 source candles, to support roughly 220 selected candles where the exchange has enough history. A failed exchange request retains previously received candles with an offline label, or uses an explicitly labelled candles.json historical snapshot when that exact interval is available. Missing data shows an error instead of fabricated candles.
• Presets: 1m, 5m, 15m, 1h, 4h, 1D and 1w. Enter a custom interval from 1 to 10,080 whole minutes. Custom intervals aggregate OHLC and base volume from the largest supported exchange interval that divides the requested duration; buckets align to UTC epoch time. An incomplete first bucket or a bucket with missing base candles is dropped. Custom backfill reuses retained older candles during refresh. Very large custom intervals can exceed the 12,000-source-candle budget; explicit history-limit feedback explains unavailable indicators. No missing bars are fabricated.
• Drag to pan, scroll or pinch to zoom, drag the price scale to adjust it, and reset the view from the toolbar. Expand fills the screen; use the exit button or Escape to return. Switching markets or intervals fits recent candles on the time axis and restores automatic price scaling. Subsequent feed refreshes preserve your manual view.
• Chart settings offer candles, OHLC bars or a close line, a linear/logarithmic price scale, grid lines and candle-volume visibility.

Drawings and indicators

• Draw a horizontal price level with one point, or a trend line, price/date range or fixed-range volume profile with two points. The toolbar names the active tool and the next placement step. Escape cancels an unfinished drawing. Select and drag a drawing to move it; Delete or the trash button removes it. Undo and redo recover creation, movement and deletion, with Ctrl/Cmd Z and Ctrl/Cmd Shift Z shortcuts while the chart has focus. The last 50 changes are kept for this chart session; changing market or interval or reloading resets that undo history. Drawings save per market and interval, up to 100. Saved drawings can also be removed from the chart settings dialog.
• The fixed-range volume profile is an approximation: each candle’s base volume is distributed uniformly over its high–low range into 24 bins. It is not exchange tick-by-tick volume-at-price. The price/date range shows percentage change and elapsed candle intervals.
• Indicators: SMA and SMA-seeded EMA, each with a configurable 2–200 candle period; Bollinger bands using SMA20 ±2 population standard deviations; session VWAP using volume-weighted (high + low + close) / 3, resetting at midnight UTC. VWAP is available only on intraday intervals; its first loaded session may be partial.
• Strategy toggles add only current entry-reference, take-profit and stop-loss pixels to the existing chart. They do not change market, interval, candles, pan, zoom, price scale, saved drawings, indicators or forecast preferences. No markers, shading, new candle data or extra indicator series are added. Existing hollow or dashed forecasts remain independently labeled forecast (predicted). Current strategy levels never represent a predicted price path. No qualifying setup means no strategy drawings. See Strategy setups for support, provisional entry rules, freshness and historical research.
• This is the open-source Lightweight Charts library with PerpIntel tools, not TradingView Advanced Charts. It does not include Pine Script, the complete TradingView drawing catalog, order execution, or TradingView account synchronization.

Forecast and price levels

The joint model produces genuine hourly scenarios starting at the completed five-minute input cutoff. The summary labels median return and up-probability as issued statistics from the original model input; retained remaining paths are not relabelled as a new full-horizon forecast. The default dashed median marks predicted hour-end closes; a separate shaded p10–p90 close range shows uncertainty. Optional hollow candles illustrate weighted median OHLC with within-candle excursions; their wicks are not prediction intervals. The forecast always follows the latest traded price: its line, candles and close range share one live anchor, without changing archived model output or evaluation.

The forecast overlay is enabled by default with the one-hour horizon. BTC, ETH and SOL share a joint observation/forecast pipeline, refreshed each completed five-minute close. With Candles selected, each body matches the selected chart interval, traded candle width and price scale. Whole-hour multiples aggregate complete hourly scenarios. Sub-hour candles are clearly labelled interpolations of an illustrative intrahour path, not finer-resolution model predictions. Incompatible larger custom intervals and remaining partial intervals use a close line. The exchange time axis keeps its cadence. Near-50% probabilities are labelled No clear tilt. Four-hour and 24-hour projections are explicitly exploratory; their directional skill is not established. Forecast style and horizon settings remain available; the legend identifies the automatic live anchor.

The last successful forecast for the same market and model stays visible during background computation and retries, then is replaced atomically. A compact status identifies an unavailable, stale or updating forecast; technical input-price/timestamp prose is omitted from the chart. Existing exchange candles and a retained forecast can remain during a temporary feed outage with an offline label. A historical snapshot does not receive a current live forecast. Switching markets never draws the prior market’s prediction; a remembered result for the selected market can appear immediately. Retained output expires at its forecast horizon and is never relabelled as newly generated. Live anchoring changes the displayed reference price, not the inference timestamp. Manual pan and zoom are preserved during refresh and representation changes.

Previous-day levels are optional. They use the completed previous UTC day’s high, low and classic pivot (high + low + close) / 3 from Binance daily candles, refreshed every minute. Failed or incorrectly dated responses hide the levels.

Auto support and resistance

Turn on Auto support and resistance in Overlays & indicators. Support marks an area where traded price previously bounced; resistance marks an area where it stalled. Either can break. These are local chart references, not guaranteed reversal points, predictions or a validated trading edge. The option is off by default, saves in this browser, and affects only its own neutral areas. Solid lines label Support; dashed lines label Resistance. The chart view, forecast, strategies, alerts, source and other preferences remain unchanged.

Context recalculates from the selected chart candles: 1m uses 120 bars (2 hours); above 1m through 15m uses 96 bars (5m = 8 hours, 15m = 1 day); above 15m through 1h uses 72 bars (1h = 3 days); above 1h through 1D uses 60 bars (4h = 10 days, 1D = 60 days); larger intervals use 52 bars (1w = 52 weeks). Custom minute intervals use the same mapping. The Last … label states the actual window. Context ends at the most recent completed candle close, so levels stay stable while the current candle forms. Changing market or interval clears previous levels while the matching feed loads; no levels are borrowed from another interval.

A pivot low/high is confirmed after two right-hand candles close. Equal-price swing plateaus count once at their last candle, with stricter prices on both sides; entirely flat history never creates a level. Once a band exists, a retest within it also qualifies when its close and both right-hand closes reject outside the band. Retests must be separated by at least four candles and an intervening departure from the band, preventing continuous contact from becoming repeated tests. One confirmed pivot can publish a swing level, explicitly labeled swing rather than retested. Two or more distinct reactions label a retested level with its measured test count. The first pivot fixes center and half-width: 0.25 × simple ATR14, bounded between 0.01% and 0.35% of price. Later reactions do not drift or widen the band. Older candles warm ATR only; minimum history remains local-window bars + 14 + 2.

Areas replay causally in candle order. A completed close below a support band's lower edge or above a resistance band's upper edge retires it; intrabar wicks do not. A broken area is never silently flipped or revived: it needs a new local confirmed swing. Reactions age out when they leave the time window. At each completed close, choose up to two non-overlapping supports wholly below and two resistances wholly above that actual close, within six ATR14, capped at 3% of price through 15m, 5% through 1h, 8% through 4h, 15% through 1D and 30% above 1D. Both volatility distance and the percentage cap therefore follow the selected interval. Older or farther areas are omitted. Areas begin drawing at their first swing's confirmation close, never back at a pivot before it was known. Later retests preserve that original drawing anchor. Exchange history corrections and rolling-window expiry can revise the next calculation.

Only ready, online exchange candles with the selected market/interval key qualify. OHLC/volume values, continuity and UTC alignment (Monday for weeks) are validated; current-bar identity and 45-second receipt freshness are checked every second and on visibility changes. Loading, stale, offline, invalid or insufficient data clears these areas and shows a brief reason. The detector never reads forecast candles. Overlaps with visible previous-day references, strategy references, visible forecast-plan entry/stop/TP1/TP2 or manual horizontal levels are suppressed, with no changes to those tools. Nearby labels are omitted when compressed; the HTML summary retains numeric centers and measured reaction counts. Selection is nearest first; ties use more reactions, then the latest reaction time. These counts describe historical tests, not a success probability. No breakout alerts or trading actions are attached to this toggle.