Signal Summary reads the indicators already on your TradingView chart (EMA stack, RSI, Balance of Power, VWAP) and distils them into an at-a-glance intraday read. It does not place trades and never invents data — anything missing from the chart shows as n/a.

You choose the symbol. Type it above and press Load: the dashboard switches your TradingView chart to that symbol and reads it. It deliberately does not follow whichever chart tab happens to be fronted, so the panel never silently changes under you — and if the chart switch does not take, it says so rather than showing another symbol's numbers.

How to combine the metrics: start with Market Bias + Trend (direction), confirm with Signal Strength + Confidence (is the evidence lopsided or conflicted?), sanity-check Volume Confirmation and Risk Level, then read Entry Readiness as the verdict: Ready = aligned high-conviction setup, Wait = directional but not yet clean, Avoid = the Avoid Trade reason tells you why (conflict, chop, or chase risk).

Metric guide: Market/Long-vs-Short Bias — net directional vote of the on-chart signals · Signal Strength — how one-sided that vote is (0–100) · Confidence — High/Good/Fair from strength · Trend — EMA-stack structure (Up/Down/Sideways) · Volume Confirmation — rVol read if a volume study is present · Risk Level — elevated on RSI extremes / large moves · Best Setup — the play that fits the current read · Trade Quality — A/B/C grade · Entry Readiness — Ready / Wait / Avoid.

This is a decision aid, not a signal to act blindly — always confirm against price at your own levels and the TPO scanner’s fixed entry/SL/target plan before trading.

Top-down, not single-chart. The symbol is analysed on Monthly → Weekly → Daily → 4H from real OHLC history (NSE via the NSE data provider — 4H aggregated from 30-minute candles anchored to the 09:15 session; US via the US data provider). Higher timeframes set the primary trend, the Daily gives the setup, the 4H gives the trigger. A timeframe with no data is reported as missing, never guessed.

Stage Analysis (Weinstein). Each timeframe is classified Stage 1 Basing / 2 Advancing / 3 Topping / 4 Declining from the 30-period MA: its slope decides advancing vs declining, and the prior slope separates a flat MA that is basing from one that is topping.

Pattern detection is geometric, not narrative. Swing pivots (5-bar fractals) feed least-squares boundary lines; every tolerance is scaled in ATR so the same rules port across timeframes. Detected: rectangle · ascending/descending/symmetrical triangle · rising/falling wedge · ascending/descending channel · bull/bear flag · pennant · cup & handle · VCP · double top/bottom · head & shoulders and inverse · breakout · breakdown · breakout/breakdown retest · trend continuation · trend reversal.

Status · Confidence · Score. Status is Confirmed (boundary decisively broken in the pattern's direction), Failed (broken against it) or Developing. Confidence (%) is a weighted sum of that detector's own evidence — line fit R², touch count, symmetry, depth, volume dry-up, break distance in ATR. Technical Score (/10) is confidence adjusted for status and higher-timeframe alignment: a Daily bull flag inside a Weekly Stage 4 is scored down, the same flag inside a Weekly Stage 2 is scored up. Only detections at ≥55% confidence are shown — an empty table is an honest "no high-confidence structure".

Support/resistance are confluence clusters: swing pivots, the 30/50/200 MAs and pattern boundaries from every timeframe, weighted by timeframe (Monthly 4× … 4H 1×) and merged within 0.6 daily ATR. Analysis only — no orders are placed, and nothing here is a recommendation.

Minervini’s order of operations, applied literally. First the Trend Template — eight objective criteria that decide whether a stock is even a candidate. A stock that fails it is rejected outright, however pretty its base looks. Only then is the VCP measured, and only then is a trade plan produced.

RS Rating is a true percentile, not a proxy. One scan of the full universe ranks every liquid name by (2×3-month + 6-month + 1-year) performance / 4 — IBD’s construction, which double-weights the most recent quarter — and this stock’s percentile is its position in that list. NSE names are ranked against NSE and US names against US; cross-market percentiles would be meaningless. Names below the liquidity floor (market cap and daily turnover) are excluded from the ranking universe entirely — unfiltered, illiquid microcaps posting +2,900% annual returns would occupy the whole top decile and make every real name’s rating worthless.

VCP detection. Swings come from a significance-filtered ZigZag; each swing high paired with the following low is one contraction. The base is the last 2–6 contractions that together satisfy every structural rule: base at least 5 weeks, no deeper than 35%, the final pullback no more than 75% of the first, at most one widening step, and a final contraction of 12% or less. Volume must dry up into the last contraction — that is the actual signal, supply being exhausted.

The pivot is the point. Buy-stop above the final contraction’s high on at least 1.5× average volume; stop at the tighter of the final contraction low or a hard −7%; targets at 2R and 3R. A tight pivot is what makes the loss small when it fails — that is the whole reason for demanding a tight base.

Honest scope: this is the technical half of SEPA only. Minervini also requires earnings and sales acceleration and evidence of institutional sponsorship, for which this dashboard has no data. BUY-READY here means “the chart qualifies”, never “buy this”. Analysis only — no orders are placed.

Read this first. An algorithm produces one rule-valid count out of several. A wave count is an interpretation, not a measurement, and it re-labels as new bars print — with the ambiguity worst at the right-hand edge, which is exactly where you would want to trade. This tab therefore always shows alternate counts beside the primary one, and confidence is capped at 75%: a wave count should never present itself as being as certain as a moving average. Trade the invalidation level, not the label.

The three hard rules are absolute. A count breaking any one is discarded outright, never shown with a lower score: R1 wave 2 never retraces more than 100% of wave 1 · R2 wave 3 is never the shortest of waves 1, 3 and 5 · R3 wave 4 never enters wave 1’s price territory. Everything else — Fibonacci proportion, alternation, channelling, volume and momentum personality — is a guideline and only contributes to the score.

Where alternates come from. Swings are extracted with a ZigZag at four significance thresholds (1.0× to 3.8× ATR). Different thresholds legitimately resolve different degrees — which is precisely where genuine disagreement between analysts comes from. Every rule-valid reading is kept and ranked.

Ranking. Counts are scored on guideline conformance, on how much of the chart they actually explain (a count spanning three months beats one labelling the last fortnight), and with a structural preference for the five-wave impulse over a three-wave correction. Counts covering less than a quarter of the window are dropped as right-edge noise.

Degrees are nested by timeframe: Monthly = Primary ①②③ · Weekly = Intermediate (1)(2)(3) · Daily = Minor 1 2 3 · 4H = Minute (i)(ii)(iii). The Daily count is treated as the actionable one; the nesting table shows whether the degrees agree.

Scope (deliberate): impulses and simple corrections — zigzag, flat and contracting triangle. Diagonals, truncations and complex combinations (WXY, double and triple threes) are not counted. Where the structure is one of those, this reports no valid count rather than forcing a label it cannot justify — an empty result is an honest answer. Analysis only; no orders are placed.

Two stages, deliberately. Stage 1 is one server-side TradingView scanner request for the whole universe (~2,900 NSE / ~5,000 US primary stocks). It keeps only names in a genuine uptrend (close above the 50-day and 200-day MA, 6-month performance positive) that are already within 15% of their 52-week high, then ranks them by how close they are, their momentum and their relative volume. Stage 2 fetches real 4H/Daily/Weekly/Monthly OHLC for the top candidates and runs the same detectors as the Pattern Analysis tab — nothing here is approximated.

Stage 2 covers the whole shortlist — and what that costs. breakouts.candidates defaults to all, so every name surviving Stage 1 gets real OHLC and the real detectors; there is no arbitrary top-N. That is deliberately expensive: a fresh NSE symbol costs up to five NSE-provider requests, that provider's Cloudflare block trips well below the documented quota, and the deep pass runs sequentially through the shared rate limiter — so a cold full India scan can run for many minutes and may well be cut short by the provider.

Three things make that honest rather than merely slow. If the limiter trips, the scan stops and says so instead of presenting a partial result as complete. A stopped-early result is not cached, so pressing Scan again actually retries. And price history is cached per symbol, so a re-run pays nothing for names already fetched and walks further down the list — successive runs converge on full coverage. Watch deep-analysed in the header: that is your real coverage, not a promise. Complete scans are cached for 10 minutes per filter combination. Set breakouts.candidates to a number if you want to cap the pass while probing filters.

What "closest to breakout" means. Each bullish detection carries the level that must be cleared (pivot / trigger / neckline / resistance). Distance is that level relative to the last close: AT PIVOT ≤0.5% · NEAR ≤2% · APPROACHING ≤5%. Levels already cleared by more than 1% are dropped — the low-risk entry has passed.

Long-only by definition — a breakout is the level above price. Bearish structures (double top, head & shoulders, breakdown) are still reported per symbol on the Pattern Analysis tab. Only detections at ≥55% confidence qualify, so an empty table is an honest "nothing is close". Analysis only — no orders are placed.

Same two-stage pipeline as Breakout-Patterns (full-universe screen server-side, then real OHLC for the shortlist), with the pivot measured by a methodology engine instead of the chart-pattern detectors.

VCP (Minervini). Each candidate runs the full SEPA technical screen from the VCP Analysis tab — Trend Template (≥7/8 required), true percentile RS Rating, and a valid volatility contraction base. Only names with a real base and a pivot within range are listed; the breakout level is the final contraction's high, with the stop and 2R/3R targets from the same trade plan. VCP bases are measured on daily bars by construction, so the timeframe filter does not apply to this engine (the tab says so when you change it).

Elliott Wave. The wave count runs at your chosen timeframe's degree. Only a live up-impulse in wave 3 or wave 5 poses a breakout question, and the level to clear is the previous same-direction wave's extreme (the wave 1 high for a wave 3, the wave 3 high for a wave 5). Corrections, completed counts and down impulses have no upside trigger and are excluded. Every row carries its invalidation level — that, not the label, is the number to trade against.

Honest scope: VCP here is the technical half of SEPA only (no earnings or sponsorship data), and a wave count is one interpretation among several that re-labels as new bars print. Distance bands, full-shortlist Stage-2 coverage and the 10-minute cache for complete scans work exactly as on the Breakout-Patterns tab. Analysis only — no orders are placed.

The NSE data provider sits behind Cloudflare, whose IP-level block (error code 1015) trips well below the documented API quota and lasts minutes. Analysing a new NSE symbol costs up to 5 requests (30-minute, day, week, month, and today's intraday), so a handful of unfamiliar symbols in quick succession is what actually triggers it.

This tab tells you the cost before you pay it. Type a symbol and press Will this fetch? — it answers entirely from local state and never contacts a provider. Free means the analysis is served from cache and costs nothing.

How long things stay cached. Settled bars cannot change while the market is shut, so windows fetched after the close stay valid until the next session opens — a symbol you analysed yesterday is free the next morning. Weekly and monthly are held for 30 days because their recent periods are rebuilt exactly from daily (verified at 0.0000% error), so they never go stale.

The cache lives in data/history_cache.json and survives restarts. Pacing is set by the request-pacing keys in .env (see .env.example) and is shown below, so you can confirm what the server is really using.

Forward-test journal. Every plan the TPO scanners freeze is recorded automatically. A plan only counts as a trade once price actually traded into the entry zone (state reached VALID) — what a real trader could have filled. Plans that ran to target or stop without filling are tracked separately as missed (fill-rate), never in PF/WR.

Outcomes: WIN = Target-1 hit (realized R = planned R:R) · LOSS = stop hit (−1R) · SCRATCH = session expired, marked at the last price. Stats need n ≥ 20 closed trades before gates show pass/fail — until then they read insufficient sample, honestly.

India 1-minute backtest. The button below replays the most recent journaled India plans (testing.backtestLimit, default 100) against real the NSE data provider's 1-minute historical candles — exact fills, stop-outs and targets (conservative: if one candle touches both stop and target it counts as a loss). Each plan costs one NSE-provider request, so the run is deliberately bounded and rate-limited: an unbounded replay of the whole journal was both a multi-minute request the browser would abandon and the single largest source of upstream load. Raise the limit in config/markets.json if you want a wider sample. USA stays forward-test only — there is no free intraday history source, and this dashboard never fabricates data.