Mohan's Dashboard

1. How this dashboard is put together

Four independent analysis pillars — QoQ Fundamentals, 3-Month Price Action, Options positioning, and ICT/Smart-Money structure — get computed for every Nifty 50 stock. Two different tabs blend them differently:

That's the Trade Shortlist / Compare composite score — it deliberately has no ICT input. The Main Dashboard's overall score is a separate, equal-weighted blend of Fundamentals, Technical and ICT instead (no options input) — see section 2. The two scores are allowed to disagree on the same stock; each answers a different question.

Fundamentals get the largest composite weight because they change slowly and are hardest to fake; price action is weighted almost as high because it's what actually pays or costs you; options positioning gets the smallest weight because it's a short-dated, sentiment-driven overlay, not a standalone reason to trade.

Every score in every tab is on a common 0–100 scale so they can be compared and blended directly. Scores are built from linear ramps: below a lower bound the component scores 0, above an upper bound it scores full marks, and in between it's a straight-line interpolation. Where an input is missing, most ramps award ~40% of the points rather than 0 — a gap in data is treated as "unknown", not "bad".

Tabs at a glance

2. Main Dashboard tab

The landing tab. It blends three of the four pillars — QoQ Fundamentals, 3-month Technical (price action) and ICT structure — into a fourth, Overall score, then charts the universe on top of a filterable shortlist table.

Overall score

A simple, equal-weighted average — deliberately not the same formula as the Trade Shortlist's composite (45% fundamentals + 40% momentum + 15% options). Overall score has no options input at all and gives ICT structure a full third of the say, so the two scores can rank the same stock differently. Both are legitimate; they're answering different questions ("what does the options market think", composite, vs. "does the chart's institutional structure agree with the fundamentals", overall).

Charts

• Sector mix — ICT bias — every sector with at least one scanned stock, as a horizontal bar split around a centre line: bullish count to the right (green), bearish to the left (red); a sector with only neutral reads shows no bar. Ranked by how many stocks were scanned in that sector.
• ICT bias mix — how many stocks currently read bullish / bearish / neutral on the ICT Shortlist tab's bias, across the whole universe.
• Average overall score by sector — the mean Overall score per sector (top 10), drawn as horizontal bars, colour-coded the same way as every other score bar on the dashboard.

All three are computed from the same scan that populates the table below — there's no separate data pull for the charts. The table itself still respects the shared search/sector filter bar at the top of the page.

Evaluate

A one-click, mechanical run of the full top-down workflow, both directions at once: Nifty & Bank Nifty index bias (reuses the Nifty & BankNifty tab's index-pulse read) → sector rotation (sectors with ≥2 Nifty 50 constituents, ranked by their constituents' average Overall score — the top 2 feed the long screen, the bottom 2 feed the short screen) → a universe screen for long ideas and, separately, a mirrored screen for short ideas → a full per-candidate deep read (fundamentals, price action, liquidity, ICT, options) → multi-system confluence (how many of the 5 Institutional Picks strategies and how many Trading Systems families confirm it) → the precise ICT Strategies trigger (the consolidated multi-model opinion from that tab) → the composite Trade Shortlist direction/levels. This is the same 8-step process as the /stock-focus Claude Code slash command (.claude/commands/stock-focus.md), reimplemented as fixed rules instead of an AI judgment call — no API key, no LLM call, no per-click cost, same "mechanical formula" contract as every other score on this dashboard.

The short screen is the long screen's exact mirror against section 1's fundamental-score ramp: a primary short needs "Weak" fundamentals (<40) the same distance below the neutral midpoint as a primary long's "Strong" (≥70) is above it. Within each side, candidates are ranked sector-in-rotation (or out-of-favour, for shorts) first, then by how far the composite score sits from neutral, capped at 3 primary + 2 borderline per side. Each candidate's headline verdict (bullish / bearish / neutral) starts from the composite Trade Shortlist direction, then gets downgraded to neutral whenever the ICT Strategies trigger disagrees with it, or has live setups but no consolidated edge either way — that disagreement is surfaced as an explicit warning on the card, not averaged into a single number. The entry zone/stop/targets shown are the ICT trigger's levels when it agrees with the composite direction (the more precise, timing-aware read), and the composite's own ATR-based levels otherwise.

Checklist

Next to Evaluate, the Checklist button (GET /api/checklist/pdf, app/analysis/checklist.py) runs a fresh Evaluate and fills the dashboard's own paper template — trade-setup-checklist.pdf's 7 sections (Seasonality & sector strength, Fundamentals, Price structure, Liquidity sweep, ICT confirmation, Options data, Risk & trade shortlist) — into a downloadable PDF, one page per long/short candidate.

Every checkbox traces back to a field this dashboard already computes: a green check means the data genuinely supports that line (with the exact number shown underneath in small print), a red cross means the data contradicts it. Anything this app doesn't track at all — an earnings/corporate- action calendar, promoter-holding changes, filed-debt trend, position sizing against your own account, whether you've actually logged the idea — is left an empty box marked "verify manually" rather than guessed at. The NOTES section at the bottom of each page is auto-filled from that candidate's own Trade Shortlist rationale and warning flags.

PDF — written insights report

Next to Checklist, the PDF button (GET /api/insights/pdf, app/analysis/insights_pdf.py) downloads this dashboard's current reading as a written document rather than a form to tick. Part A walks the four Market-metrics cards section by section — Options & derivatives, Fundamental analysis, Sentiment & market structure, Technical indicators — for both indices, Nifty 50 first then Bank Nifty, so you get the full set instead of whichever one the on-screen toggle happens to be on. Part B is the whole Evaluate result: index bias, the sector-rotation table, then one page per long/short candidate with its screen, ICT and options read, multi-system confluence, both sets of trade levels, rationale and every warning.

Because a printed page has none of the screen's colours, bars and tooltips, each section opens with a plain explanation of what it measures, each verdict prints the exact per-metric vote behind it ("IV percentile neutral / PCR bearish / OI Calls-Puts bearish / Max pain bullish"), and each number carries a line on what that particular reading implies. The verdict rules are the same ones the browser runs on the same payloads, so the PDF can't disagree with the card it came from; where a feed genuinely isn't available (FII/DII flows, greeks without a Kite session) it says so rather than printing a placeholder. It runs its own fresh Evaluate server-side, so it works whether or not you pressed Evaluate first.

It's the slowest button here — a cold run does a full index pulse, a universe fundamentals pass, a structure-event sweep and a complete Evaluate before a page is drawn — so instead of a featureless spinner the button starts a background job (POST /api/insights/pdf/start) and polls it, counting off the five real phases it's working through (Index pulse → Universe fundamentals → Structure events → Evaluate → Rendering) with the elapsed seconds alongside. It's a genuine position in a known sequence, not a made-up percentage: the label only advances when the server reports that phase has started. On a warm cache the whole thing finishes in seconds; on a cold one it's a couple of minutes. GET /api/insights/pdf still renders the same document in one blocking call if you'd rather hit it directly.

Chart Insight

Below Evaluate, the Chart Insight panel (GET /api/chart-insight/{symbol}?interval=, app/analysis/chart_insight.py) answers a narrower question than Evaluate: not "which stocks to trade" but "what just happened, on this one symbol, on this one candle". Pick any Nifty 50 stock, or NIFTY/BANKNIFTY for the index itself, and a candle interval — 5-Minute, 15-Minute, 30-Minute, 1-Hour, 4-Hour, Daily, Weekly or Monthly — and it reads the most recently closed candle at that interval:

• Price change — that one candle's open→close move in points and %, and its high–low range as a % of the open.
• Volume — that candle's volume vs the average of the 20 candles immediately before it (not including it). Genuinely up ≥50% is flagged as a real expansion; down ≤−40% is flagged as thin participation, worth discounting the price move.
• Market structure — the same ICT engine (section 9) run on real intraday candles at the chosen interval instead of the usual 3-month daily window: trend, premium/ discount zone, and the latest confirmed BOS/CHoCH/MSS.
• Options — PCR, ATM IV, max pain and today's futures OI change from the same Sensibull/ Kite read the Options tab uses (section 5).

Same rules as the /chart-insight Claude Code slash command (.claude/commands/chart-insight.md), which runs this same mechanical read first and then layers a judgment call on top — drawing on the ict-smc-analysis, classical-technical-analysis and nifty-sentiment-scoring skills to say whether the structure/volume/options reads actually agree, the way this fixed-rule version can't.

3. QoQ Fundamentals tab

Pulls the last few reported quarters (yfinance, with an optional enrichment source filling gaps) and compares the latest quarter sequentially (QoQ, vs the immediately prior quarter) and annually (YoY, vs the same quarter a year ago, matched 330–400 days back).

Fundamental score — 7 weighted components summing to 100

ramp(value, lo, hi, points) is the shared linear scale used everywhere on this dashboard: value ≤ lo scores 0, value ≥ hi scores the full points, and anything in between is a straight-line interpolation. A missing value scores 0.4 × points — neutral, not punitive.

Consistency is the share of the last 4 sequential quarters where the metric rose vs the quarter before it (e.g. profit consistency of 75% = profit grew in 3 of the last 4 quarter-on-quarter steps). It rewards steady compounding over one lucky quarter.

Verdict thresholds

Red flags (any that apply are shown as warning chips)

• PAT fell more than 15% QoQ
• Revenue fell more than 5% QoQ
• Net margin compressed more than 1.5pp QoQ
• Latest quarter was loss-making
• Revenue up >3% but profit down >3% ("cost pressure")
• TTM PE above 70 ("rich valuation")
• Latest reported quarter is more than 130 days old (stale data)

4. Price Action (3M) tab

Shows the last 63 trading sessions (~3 months) of daily candles. Indicators are computed over a longer 400-day padded history first, so lookback windows like the 200-day EMA are already valid at the left edge of the visible 3-month window, then the view is sliced down to 63 sessions.

Trend classification

• If the 20- or 50-day SMA isn't available yet → neutral.
• "Strong" means ADX(14) ≥ 25 (a genuinely trending tape, not just drifting).
• Price above SMA20 above SMA50, and above the 200-EMA → strong uptrend (or plain uptrend if not "strong").
• Price below SMA20 below SMA50, and below the 200-EMA → strong downtrend (or plain downtrend).
• Otherwise, if price is above SMA50 → recovering.
• Otherwise → sideways / weak.

Support / resistance are pivot levels from the last 20 sessions: resistance is the nearest high above the current close, support is the nearest low below it.

Momentum score — 6 weighted components summing to 100

Trend points: strong uptrend 30 · uptrend 24 · recovering 16 · neutral 10 · sideways/weak 8 · downtrend 4 · strong downtrend 0.

RSI posture rewards the 50–70 "healthy trend" zone the most, and penalises both extremes: 50–70 → 15pts (full marks) · 40–50 → 9 · 70–80 → 10 (strong but extended) · >80 → 4 (overbought) · <40 → 3 (weak) · missing → 7.

5. Options tab

Two data layers, both optional: a no-login Sensibull summary (ATM IV, IV percentile, PCR, max pain, futures basis) always tried first, and a full strike-by-strike chain with solved implied vols and Black-Scholes greeks when a Kite Connect session is available.

Options score — starts neutral at 50, then nudged up/down (clamped 0–100)

• PCR: ≥1.3 → +12 · 1.0–1.3 → +6 · ≤0.6 → −12 · 0.6–0.8 → −6
• Spot vs max pain gap: >+1.5% → +6 · <−1.5% → −6
• Futures basis: >0.35% (premium, long build-up) → +8 · <−0.2% (discount, shorts) → −8
• Futures OI + price combo (when |OI change| > 1%): OI↑ & price↑ → +10 (long build-up) · OI↑ & price↓ → −10 (short build-up) · OI↓ & price↑ → +4 (short covering) · OI↓ & price↓ → −4 (long unwinding)
• IV percentile: <30 (cheap options, tailwind for buyers) → +5 · >80 (rich) → −5
• IV skew (25-delta put IV − 25-delta call IV) > 4 → −4 (steep put skew = downside demand)

Options bias

Each adjustment above also casts a directional vote (+1/−1). Bullish if the net vote ≥ 2 and the score ≥ 58. Bearish if the net vote ≤ −2 and the score ≤ 42. Otherwise neutral.

Strategy ideas

Chosen from bias (bullish / bearish / neutral) crossed with whether IV percentile is rich (>70) or cheap (<35): rich IV favours selling defined-risk spreads (Bull Put / Bear Call / Iron Condor), cheap IV favours buying (Bull Call / Bear Put spreads, or an outright Long Call/Strangle when very cheap and enough time remains). Every idea is risk-defined — the engine never suggests a naked short option. A "max-pain pull" note is added whenever spot is more than 3% away from max pain, as context only.

6. Trade Shortlist & Compare tabs

These are the two tabs that combine all three pillars. Compare additionally shows the full 50-stock table side by side with independently-ranked Top-Buy / Top-Sell leaderboards.

Composite score

If fundamentals failed to load, fundamental_score defaults to a neutral 45. If options weren't requested or are unavailable, options_score defaults to a neutral 50 — a missing pillar pulls the composite toward neutral rather than zeroing it out.

Direction (Call)

• Long: momentum score ≥ 55 (bullish chart) and fundamental score ≥ 55 (good books).
• Short: momentum score ≤ 38 (bearish chart) and fundamental score ≤ 40 (bad books).
• Watch: everything else — e.g. a strong chart running ahead of weak fundamentals, or solid books whose chart hasn't turned up yet.

Conviction

Only long/short picks can reach high or medium; everything else is low.

• High: composite ≥ 66 and options positioning doesn't contradict the call (a long needs options bias ≠ bearish; a short needs options bias ≠ bullish).
• Medium: composite ≥ 56.
• Low: everything else.

Entry / stop / target levels (ATR-based, long or short only)

ATR is the 14-day Average True Range from the Price Action tab; support/resistance are the same 20-session pivot levels shown there. These are mechanical reference points for sizing risk, not price predictions.

Rationale & warnings

The bullet list on every card is auto-composed from real numbers elsewhere on the dashboard: the latest QoQ revenue/profit/margin change, the YoY profit change, profit consistency %, the chart's trend plus its 3-month return / relative strength / RSI / ADX, how price reacted the day after the last quarterly result, and (when available) the options read — PCR, ATM IV, futures basis and max pain. Warning chips combine the fundamentals red flags above with two chart-side checks: ATR% > 4% ("high daily range") and volume ratio < 0.6 ("volume drying up").

Ranking rules — why Trade Shortlist and Compare's Top Buy/Sell can differ

• Trade Shortlist ("Top N" picks) sorts the whole scan by direction priority first (long and short both outrank watch), then by composite score, and truncates to N — so it's one interleaved list mixing long and short names.
• Compare › Top Buy / Top Sell are two separate leaderboards: Top Buy takes only direction = long picks sorted purely by composite score; Top Sell takes only direction = short picks sorted purely by composite score. Each is ranked 1..N on its own, independent of the other side. Seeing fewer than N names on one side just means fewer stocks cleared that side's bar today — the rest are sitting in "watch".

7. ICT / Smart-Money-Concepts Shortlist tab

Reads the same 3-month daily candles as the Price Action tab — no extra data fetch — through an ICT (Inner Circle Trader) / Smart Money Concepts lens: market structure, premium/discount, resting liquidity, order blocks, fair value gaps and the Optimal Trade Entry zone.

Swing points & market structure

A swing high/low is detected with a standard 5-candle fractal: a candle whose high (or low) is strictly the most extreme within 2 candles on either side. The engine then walks the candles chronologically, tracking the most recent unbroken swing high and swing low:

• Break of Structure (BOS): price closes beyond the tracked swing point in the direction the trend is already moving — continuation.
• Change of Character (CHoCH): price closes beyond the tracked swing point against the prevailing trend, with no displacement candle backing the break — the first, tentative sign of a reversal.
• Market Structure Shift (MSS): the same reversal break, but confirmed by a displacement candle on (or the bar immediately after) the break — the stronger, confirmed reversal signal.

Whichever direction just broke becomes the new trend, and the engine starts tracking the next fresh swing point on watch.

Displacement

Validates that a break is a genuine institutional move, not a wick-driven liquidity sweep:

Checked on the breaking candle itself, or the candle right after it.

Dealing range & Premium/Discount

The current dealing range is simply [most recent swing low, most recent swing high]. The zone is where price sits inside that range:

Liquidity pools & sweeps

Every swing high is a pocket of resting buy-side liquidity (BSL) — buy-stop orders from short sellers and breakout buyers. Every swing low is sell-side liquidity (SSL) — sell-stop orders from long holders. Swing points within 0.25% of each other are clustered into one pool (an "equal highs/lows" magnet with a touch count > 1). A sweep is a candle whose wick pierces a pool but whose close snaps back on the other side — a stop-run, not a real break. The dashboard reads the most recent sweep as directional fuel: a swept low (SSL) is read as bullish (liquidity grabbed to fill longs), a swept high (BSL) as bearish.

Fair Value Gaps (FVG)

A 3-candle imbalance: candle 1's wick and candle 3's wick don't overlap, leaving a gap in candle 2.

"Filled" tracks how far price has since retraced into the gap (as a % of its height); a gap is filled once price has traded all the way through it.

Order Blocks & Breaker Blocks

For every displaced BOS/MSS, the engine searches backward (up to 6 candles) for the last opposite-coloured candle before the impulsive move — its full high–low range is the order block. It's flagged fvg_confirmed if an FVG formed within ~6 days in the same direction, and bos_confirmed/MSS accordingly. If price later trades back through the block, it's mitigated; if price then closes fully through the far side, the block has failed and flips into a breaker — a former support zone that now acts as resistance, or vice versa.

Optimal Trade Entry (OTE)

The 61.8%–79% Fibonacci retracement of the current dealing range — the zone ICT treats as the highest-probability entry on a pullback.

ICT score & bias — starts neutral at 50, clamped 0–100

• Structure trend: bullish +25 · bearish −25
• Displacement on the last break, same direction as trend: +10 / −10
• Zone: discount +10 · premium −10
• Last liquidity sweep: sell-side (bullish read) +15 · buy-side (bearish read) −15
• Freshest unmitigated order block: bullish +10 · bearish −10

Bias: ≥62 → bullish · ≤38 → bearish · otherwise neutral. Entry/stop/target levels are only produced for a bullish or bearish bias, using the OTE zone as the entry, the far side of the dealing range (ATR-buffered) as the stop, and the nearest opposing liquidity pool as target 1 — the same ATR (14-day, from the Price Action tab) that powers the Trade Shortlist's levels.

8. Structure Events tab

A day-based feed of every Break of Structure, Change of Character, Market Structure Shift and diagonal trendline break fired across all 50 stocks — the same market-structure walk as the ICT Shortlist tab above, but exposing every event in the 3-month window (up to the most recent 30 BOS/CHoCH/MSS events and 10 trendline breaks per stock) instead of collapsing it down to just the latest one.

Trendline breaks

A distinct signal from a BOS/CHoCH/MSS break, which is always a horizontal swing-level break. A trendline break is diagonal: connect two consecutive lower swing highs into a descending resistance line, or two consecutive higher swing lows into an ascending support line, extrapolate it forward, and flag the first candle whose close trades through it.

Lookback & the bullish/bearish split

The Lookback selector controls how far back the full event-log table reaches (by calendar date). The Bullish today / Bearish today panels always show only the events dated on the single most recent session date seen across the universe, regardless of the lookback selected, so the headline split stays a clean "what fired today" read even when the table below is showing a wider window.

Highlights: sector shift, order block, FVG

Every row/card also surfaces three highlight signals so a reversal isn't read in isolation:

• Sector shift: flagged when 2 or more stocks in the same sector bucket fire a same-direction CHoCH/MSS on the same date — a broader, sector-wide reversal rather than one stock's isolated break. This dashboard has no separate tradable sector-index price feed (only NIFTY/BANKNIFTY are fetched as indices — see the Nifty & Bank Nifty Pulse tab), so this is a breadth read built from the sector's own Nifty 50 constituents' structure events, not a distinct sector-index candle series.
• OB bullish/bearish: that stock's current freshest unmitigated order block (from the same ICT read as the ICT Shortlist tab) — independent of this specific event's date.
• FVG bullish/bearish: that stock's current nearest unfilled Fair Value Gap — same caveat, it's the stock's present state, not dated to the event.

The "Sector-wide shifts" panel above the bullish/bearish split lists every sector that cleared the 2-stock bar on the latest session, with the count of reversing stocks out of that sector's total Nifty 50 constituents.

9. ICT Strategy Picker tab

Same 3-month daily candles, same ICT engine as the shortlist above — but instead of one blended score, this tab lets you pick a single symbol (any Nifty 50 stock, or the NIFTY 50 / Bank Nifty index itself, off ^NSEI/^NSEBANK) and any combination of named ICT entry models from ICT_Complete_Trading_Strategy_Playbook.md Section 10, each evaluated independently, then combined into one consolidated opinion.

The seven selectable strategies

Every strategy's entry/stop/target levels reuse the same ATR-buffered-stop / nearest-liquidity-pool-target logic as the ICT Shortlist tab (ict._levels), just anchored to that strategy's own zone (an order block, an FVG, the OTE zone, or a plain ATR band around the current price) instead of always the OTE zone.

Consolidated opinion

Only the strategies you've ticked are evaluated. Among the ones with a live setup ("applicable"), each contributes its own confidence (0–100) as a directional vote:

The displayed entry/stop/target is taken from the single highest-confidence applicable strategy that agrees with the overall bias (falling back to the ICT Shortlist's own OTE-based levels if none of your selected strategies produced one). If none of your selected strategies has a live setup right now, the opinion stays neutral and a warning says so — that's a valid, honest "stand aside" answer, not a failure.

10. Liquidity tab

Answers a different question from every other tab: not "should I buy/sell this" but "can I actually get in and out of a position in this without moving the price". It is deliberately kept out of the composite/overall scores everywhere else — a great fundamental+technical setup in a name you can't size into isn't actionable, but conflating the two would hide that.

Four independent signals, each optional/degradable

ADTV & turnover velocity

ADTV is the plain-English "how much of this trades per day, in rupees". Turnover velocity normalises that by company size — a ₹500cr/day stock with a ₹5,00,000cr market cap turns over far less of itself daily than the same ₹500cr/day on a ₹50,000cr company.

Options chain depth

Tagged Deep / Adequate / Moderate / Light against fixed thresholds (25L / 6L / 1.5L average OI) — informational, shown alongside the score rather than folded into a single number silently.

Bid-ask spread

Top-of-book spread from Kite's market depth: (best ask − best bid) / best ask × 100. Unavailable without a live Kite Connect session (python -m scripts.kite_login) — every other signal still works without it.

ICT liquidity-pool proximity — shown, not scored

Distance from last price to the nearest untested resting liquidity pool from the ICT tab (section 7), as a % of price. This answers "where might a stop-run happen", not "can I get filled" — a different question from execution liquidity, so it's surfaced as context (and drawn on the detail-view chart as dashed/solid horizontal lines) rather than mixed into the 0–100 score.

Liquidity score & tier — weights sum to 100 when every signal is available

A missing signal scores ~40% of its own weight (same "unknown, not bad" convention as every other ramp on this dashboard) rather than zeroing the score — e.g. without a Kite session, the spread component defaults to 4 of its 10 points instead of 0.

Where it shows up

Its own league table + Most Liquid / Thinnest-in-peer-group leaderboards on this tab; a compact Liquidity column on the Main Dashboard, Trade Shortlist cards, Compare and ICT Shortlist; and the shared filter bar's Min liquidity control (section 10) works on every one of those. The Nifty & Bank Nifty Pulse tab (section 9) also carries a Liquidity read, with ADTV/turnover velocity excluded — see that section for why.

11. Nifty & Bank Nifty Pulse tab

The only tab that looks at the index itself — NIFTY and BANKNIFTY — rather than the 50 underlying stocks. It reuses the exact same pillars as the rest of the dashboard, just pointed at the index's own daily candles and its own F&O chain instead of a constituent stock's.

Price action & ICT

Identical engines to the Price Action and ICT Shortlist tabs (same fixed 3-month / 63-session window, same trend classification, same market-structure/liquidity/OTE logic), run against the index's own candles — ^NSEI for Nifty 50, ^NSEBANK for Bank Nifty — instead of a stock's. See sections 4 and 7 for the exact formulas.

Average OHLC (3-month)

The mean of each of open, high, low and close across the same fixed 3-month (63-session) candle window used everywhere else — a smoothed read of the range, distinct from the single most-recent session's OHLC and from the high_3m/low_3m extremes below.

Informational only — like the pivot ladder below, it does not feed any score.

Range & pivot ladder (3-month)

Two figures struck from the same fixed 3-month (63-session) high/low/close used everywhere else on this dashboard — deliberately not the previous day's candle, so they stay consistent with the range every other index-pulse figure is read from.

Classic floor-trader pivot ladder, with H/L/C being the 3-month high, 3-month low, and latest close:

Informational only — the pivot ladder does not feed any score.

Options

Same engine as the Options tab (section 5), run against the NIFTY / BANKNIFTY index F&O chain instead of a stock's. Strategy ideas are generated the same way, cross-checked against the index's own momentum score as the bias hint. Any strategy with concrete BUY/SELL legs gets the same editable "Order legs — edit before creating the virtual order" table as an Evaluate/Find_Dir card (section 13a's Portfolio tab is where it ends up once closed) — each leg's live/estimated LTP, IV, delta and OI, editable quantity, and a Create virtual order button that records it to data/backtest.json under the index symbol (NIFTY/BANKNIFTY) itself, same as any stock's paper order.

Fundamentals — a breadth proxy, not the index's own result

Neither NIFTY nor BANKNIFTY files a quarterly result — there's no P&L to read QoQ. Instead this panel aggregates the QoQ fundamental score (section 3) of the Nifty 50 constituents that make up the index:

• Nifty 50: average, and breadth stats (% with positive QoQ revenue/profit growth, count scoring ≥70 "strong" / <40 "weak"), across all successfully-scanned Nifty 50 constituents.
• Bank Nifty: the same aggregate, restricted to Nifty 50 constituents tagged sector "Bank". This dashboard's universe doesn't carry the official 12-stock Bank Nifty constituent list, so this is a sector-breadth proxy from whichever of those names are also in the Nifty 50 — read it as "how healthy are the large listed banks", not as the index's own financials.

Liquidity — options depth & ICT pool proximity only

Same engine as the Liquidity tab (section 8), run against the index's own candles and F&O chain, with one deliberate difference: ADTV and turnover velocity are excluded. Neither NIFTY nor BANKNIFTY is itself traded — only their futures/options are — so the "volume" yfinance reports on ^NSEI/^NSEBANK candles is a proxy figure, not shares traded, and multiplying it by the index level would produce a rupee-value ADTV that looks real but measures nothing. Those two signals fall back to the same "unknown, not bad" 40%-of-weight default used whenever a signal is unavailable elsewhere on this dashboard, rather than being fabricated from that number.

What's left — options chain depth (index F&O open interest) and distance to the nearest untested ICT liquidity pool — are meaningful at the index level and score exactly as described in section 8. Bid-ask spread also stays unavailable (Kite market depth is an equity-only signal in this dashboard).

This tab additionally splits that single "nearest pool" figure into two directional reads — buy-side liquidity (nearest untested pool above price, clustered from swing highs) and sell-side liquidity (nearest untested pool below price, clustered from swing lows), each as a signed % distance from the current index level. See the ict-smc-analysis skill / section 7 for what a liquidity pool is and how buy-side/sell-side clustering works; like the single-pool figure, these are informational and don't feed liquidity_score.

Overall score

A simple five-way average across every pillar computed for that index, including options and liquidity this time (unlike the Main Dashboard's Overall score in section 2, which only averages three). A missing/failed options or liquidity read defaults to a neutral 50, same convention as the composite score in section 6.

Prediction & bias

Three independent directional signals are read off the index and majority-voted:

• Price trend → bullish if strong uptrend / uptrend / recovering, bearish if strong downtrend / downtrend, else neutral.
• ICT bias — bullish / bearish / neutral, straight from the ICT engine (section 7).
• Options bias — bullish / bearish / neutral, straight from the options engine (section 5); neutral if the options read is unavailable.

Two or more bullish votes → overall bias bullish; two or more bearish → bearish; anything else (including a 1-1-1 split) → neutral. Confidence is (votes agreeing with the overall bias) / 3, shown as a percentage — e.g. 2 of 3 signals agreeing is 67%. This is a mechanical vote count, not a probability estimate.

Each index card shows this vote as a directional indicator badge (▲/▼/▬ + BULLISH/BEARISH/NEUTRAL + the confidence %) next to the index name, colour-matched to the rest of the dashboard's up/down/flat convention — a more prominent restatement of overall_bias and prediction_confidence, not a new calculation.

Sentiment score — Positive / Negative / Neutral

A second, differently-weighted read of the same index — a market-mood gauge, distinct from the trade-direction vote (Overall bias) above. The two are computed from overlapping inputs but can legitimately disagree: bias asks "which way should a trade lean", sentiment asks "how constructive does the tape look right now". It blends five categories drawn from the classic market-sentiment toolkit (price/trend, options positioning, breadth, volatility, market structure) — each scored 0-100, then combined:

Breadth is an advance/decline-style proxy, not an official index breadth feed:

Volatility maps the India VIX regime to a score using the standard inverse VIX–Nifty correlation (rising/elevated VIX = fear = bearish mood; low/falling VIX = confidence = bullish mood): low (VIX < 15) → 70, normal (15–20) → 50, elevated (> 20) → 25; unavailable → a neutral 50, same "unknown, not bad" convention used elsewhere on this dashboard.

The blended score maps to a label: Positive ≥60, Negative ≤40, otherwise Neutral. The card's breakdown panel shows each of the five category sub-scores individually so you can see what's driving the label, plus the raw PCR/RSI/VIX/breadth figures behind it.

Sentiment gauge (Sentiment tab)

The Sentiment tab (its own tab, reusing this same /api/index-pulse payload) renders sentiment_score as a semicircular gauge, canvas-drawn with no charting library — three coloured zones matching the label thresholds above (red 0–40 / amber 40–60 / green 60–100) with a needle at the current score. It is the single blended number from every category in this section, given a visual read at a glance instead of a bare figure.

IV & greeks — ATM (current price)

A separate info block reading the option leg nearest the index's current spot price — the ATM strike, where gamma/theta exposure is highest and the IV read is cleanest:

• ATM strike / ATM IV — always available from the Sensibull summary layer (section 5), same figures already shown in this tab's kv block above.
• Delta / Gamma / Theta / Vega (call and put) — only available when a live Kite option chain is connected (section 5's "layer 2"): Kite's per-strike LTPs feed the Black-Scholes solver (app/analysis/bs.py) for a solved IV, then greeks off that IV. Without Kite, this block shows ATM strike/IV only, with a note explaining why greeks are blank — same "degrade gracefully, don't fabricate" rule as every other Kite-only signal on this dashboard (e.g. bid-ask spread on the Liquidity tab).

These figures are informational — like the pivot ladder and average-OHLC blocks earlier in this section, they don't feed sentiment_score or options_positioning directly (that sub-score is options_score from section 5, which already factors ATM IV/skew when available).

Option direction — day-by-day OI (this month)

A running, day-over-day read of Call vs Put open-interest change against the index's own price change — the classic long-buildup / short-buildup / short-covering / long-unwinding matrix, applied separately to calls and puts and then combined:

Each day's call-signal and put-signal are combined: if they agree, that's the day's signal; if one is neutral, the other wins; if calls and puts point opposite ways outright, the day reads neutral (no consensus). The tab's option_direction badge takes the majority of the last 3 logged days.

Snapshots are logged once per calendar day into the same SQLite cache every other tab already shares (app/cache.py, key oi_history:<NIFTY|BANKNIFTY>), with a ~400-day TTL so it behaves as a permanent log rather than an expiring cache entry. Re-running the scan later the same day updates today's row rather than duplicating it. The table shows the current calendar month, most recent day first, and starts empty and builds up one row per day — there is no bulk historical backfill, since neither Sensibull nor this dashboard's providers expose past-dated OI.

CMP strike-activity card

A second card under each index card, tracking option-chain activity at the strikes right around the current market price (CMP): the ATM strike plus one strike above and one below (offset configurable to ±1/±2/±3 strikes via the card's own selector), split into Call and Put rows. It reuses the same options chain already fetched for the card above it (OptionsReport.chain, section 5) rather than a separate data source.

The table only populates once a Kite session is connected (Dashboard Management → Settings) — OptionsReport.chain is empty in Sensibull-only mode, since Sensibull's free summary has no per-strike ladder. The card falls back to a message explaining this rather than showing empty/fabricated rows. The mini line chart below the table plots total CE OI vs total PE OI (summed across the three strikes) across the last 40 polls held in that browser tab; it starts empty and fills in as polls accumulate, same as the OI-buildup log's daily table above.

Polling happens every 15 seconds via GET /api/index-pulse/options/<NIFTY|BANKNIFTY> while the Nifty & BankNifty tab is open (it stops the moment you switch tabs), but the figures themselves only change as often as the shared 5-minute options cache TTL refreshes — this is near-real-time within that cache window, not a tick-by-tick feed.

12. Sector tab

Every requested NSE sectoral index, compared against its own same-calendar-month average over the trailing 5 years — not a flat average of every month in a 5-year window. If the current month is August, the average is specifically the average of the last 5 Augusts, compared against this August's own level so far. Powered by GET /api/sectors → app/analysis/sectors.py build_report(), run in parallel across a thread pool (same pattern as the main scan), one Yahoo Finance fetch per index.

Each index ticker is fetched verbatim from Yahoo Finance (yfinance.download(ticker, period="6y", interval="1d")) — unlike the per-stock provider, no .NS suffix is appended, since index tickers don't follow that convention consistently (some carry a ^ prefix, some don't). Daily closes are grouped by calendar year/month in-app (Yahoo's own monthly aggregation truncates history to a single row for several of these tickers even though the full daily series exists), then averaged per matching month. Results are cached for CACHE_TTL_SECTOR_HOURS (default 24h) since a 5-year seasonal average barely moves within a day.

The bar chart draws two bars per sector: grey for five_year_avg, green/red for current_price (green when at/above the average, red when below). Each row is scaled to its own max(five_year_avg, current_price) rather than one shared scale across all rows — sectors sit at wildly different absolute index levels (Bank Nifty in the tens of thousands, Nifty Realty in the hundreds), so a shared scale would flatten every low-level sector's bars to nothing. Bar length therefore only compares within a sector; the pct_vs_5y_avg label next to the current-month bar is what makes sectors comparable to each other.

Auto-preloaded on page open, same as every other scan-based tab — press Run scan to force a refresh.

Single-sector seasonal growth card (above the main table): pick any sector with a known ticker from the dropdown to see seven vertical bars, one per year (the current year back to 6 years prior) — each bar is that specific year's own current-calendar-month growth, from the month's first close to its last close. This is a different number from pct_vs_5y_avg above: that compares this month's average level against the 5-year average of that month; this card shows each year's own intra-month return in isolation, so you can see the seasonal trend behind the average rather than just the average itself. The current year's bar is necessarily partial (start-of-month to the latest available close, not a completed month) — it's marked with a dashed accent outline and a "(so far)" sub-label so it isn't read as a like-for-like completed-year figure. Backed by GET /api/sectors/<name>/monthly-growth → sectors.py build_monthly_growth(), fetched on demand when a sector is selected (not part of the main scan). A year with fewer than 2 trading sessions in that month shows "n/a" rather than a fabricated bar.

13. Backtest tab

A walk-forward simulation of the ICT Strategy Picker tab's own signal logic against a symbol's daily candle history. Press Backtest on any card there (the Top 5 bullish/bearish scan cards, or the single-symbol opinion card) to queue it here — the queue is localStorage-only, same as the watchlist, so it isn't shared across devices. Press Run on a queued row to execute it.

At every historical session, it rebuilds a PriceActionReport from only the trailing 63-session (price_action.WINDOW_DAYS) window ending at that session's close — i.e. only information that would actually have been available at the time — then re-runs the exact same ict.build_report → ict_strategies.evaluate_signals → consolidate pipeline the live tab uses. If the consolidated opinion is directional with concrete entry/stop/target1 levels, it opens a paper trade at that session's close and walks forward (up to Max hold bars) checking each day's high/low against the stop and target — a same-day hit on both is scored as the stop (conservative). No exit inside the window closes the trade at Max hold at that day's close ("timeout"). Trades never overlap: the next signal search starts after the prior trade's exit session.

Options P&L overlay (NIFTY / BANKNIFTY only)

Checking Translate to NIFTY/BANKNIFTY options P&L before running a queued index backtest overlays every trade with a buying-only options read — a long call (CE) for a bullish trade, a long put (PE) for a bearish one — priced with this app's existing Black-Scholes engine (app/analysis/bs.py, the same one the Options tab's IV solver uses). It is a deliberate approximation, not a real option-chain replay:

• Strike: nearest at-the-money strike to the entry spot price (NIFTY: 50-point step, BANKNIFTY: 100-point step).
• Volatility: the trailing 30-session realised volatility as of entry (falls back to a fixed 13% if there isn't enough price history yet) — used as a stand-in for market implied volatility, since this app has no historical IV series to replay against.
• Time to expiry: assumed 7 calendar days from entry (a weekly-expiry proxy), ticking down by the calendar days actually elapsed to the exit date, floored at 1 day so an exit is never priced at literal zero time value.

Requesting the overlay against an individual stock symbol is silently ignored (a warning explains why) — it's only meaningful for the two index symbols this dashboard treats as options underlyings. Treat the resulting P&L as a directional shape check (does theta/time decay eat this edge before it plays out?), not an exact premium quote.

13a. Portfolio tab

Where a paper order from the Backtest tab goes once you're done tracking it live. Pressing Close & move to portfolio beside a Backtest-tab order marks every leg to its current price one last time (live Kite LTP, or a Black-Scholes estimate when Kite isn't connected — same pricing as the Backtest tab), tags the trade closed with a closed_at timestamp, and moves it out of data/backtest.json into data/portfolio.json (app/analysis/portfolio.py). Nothing is ever sent to Zerodha — this only rearranges which local JSON file the paper order lives in.

The tab shows two separate sections, each with its own summary row:

14. Scalping tab

1–5 minute intraday scalp setups for NIFTY, BANKNIFTY and FINNIFTY, built from nifty-banknifty-scalping-dashboard-prompt.md's Section 5 signal modules. Unlike every other tab, this one reads intraday candles (5-minute bars via yfinance, delayed) instead of the fixed 3-month daily window — each index's own session, refetched every time you press Run scan.

Session status

Read off the current IST clock (not the data), Section 2 of the spec:

• Pre-open — before 9:15.
• Opening (signals suppressed) — 9:15–9:20, deliberately blank to avoid the first-print volatility distortion.
• Live — 9:20–14:45.
• Live — position-limit snapshot window — 14:45–15:30, SEBI's intraday FutEq monitoring window; signals still fire, just flagged.
• Closed — after 15:30, weekends, or holidays; the panel shows the last completed session instead.

Regime banner — CPR width, virgin CPR, India VIX

Central Pivot Range, struck from the prior day's high/low/close:

CPR regime: narrow (favour breakout setups, high trend-day probability) when width / prev_close × 100 < 0.15%, else wide (favour mean-reversion setups). A relative threshold, not a fixed point count, so the same rule reproduces the original 35pt-NIFTY / 80pt-BANKNIFTY tuning and also works unchanged for a ₹250 Nifty 50 stock in the Nifty 50 filter below.

Virgin CPR: yesterday's CPR zone [BC, TC] is computed from the day before yesterday's OHLC and checked against yesterday's actual traded range. If yesterday's session never touched that zone, it's flagged as an untested magnet — price tends to gravitate back to it.

India VIX filter: current ^INDIAVIX close. Below 15 → low-IV regime, a banner warns of theta-decay chop risk, and every signal's confidence is multiplied by 0.8 before it can reach the confluence scorer.

Camarilla pivots

Also struck from the prior day's high/low/close, with the standard 1.1 multiplier:

VWAP suite (Section 5.3)

Session VWAP and running ±1 standard-deviation bands, recomputed from every 5-minute candle since the day's open, volume-weighted on typical price (H+L+C)/3. Suppressed for the first 30 minutes of the session (SD is unstable with too few prints). Three setups:

• VWAP Reclaim (long) — price dipped below VWAP, closed back above it, RSI(9) 38–58.
• VWAP Rejection (short) — rally into VWAP from below rejected with a bearish close, RSI(9) 45–65.
• VWAP Squeeze Breakout — multi-candle compression around VWAP resolved by a close outside the ±1SD band on >2× the 20-candle average volume.

Camarilla mean-reversion & breakout (Section 5.6)

• H3/L3 fade — a reversal candle at the H3 or L3 boundary with RSI(9) at an extreme (≥65 / ≤35) fades back toward the opposite boundary inside the band.
• H4/L4 breakout — a decisive close beyond H4 or L4 targets the H5/L5 Fibonacci extension.

Volume Profile / POC (Section 5.1)

Session candles are bucketed into 20 price bins by closing price (a close-price proxy — true volume profile needs tick data); the highest-volume bin's midpoint is the POC, and the tightest set of bins covering 70% of session volume gives the Value Area High/Low.

• POC Mean Reversion — price stretched ≥0.10% of price away from POC with a flat EMA9/EMA21 (no sustained directional flow) fades back toward POC.
• POC Migration — the session's first-half POC vs second-half POC has shifted ≥0.08% of price — treated as trend-day confirmation, traded with the shift.

Both thresholds are relative (% of price), same reasoning as the CPR-width threshold above — it keeps the setups meaningful whether they're running against the Nifty 50 index or an individual constituent stock.

EMA momentum & Opening Range Breakout (Section 5.4)

• EMA 9/21 cross on the 5-minute chart, confirmed by the triggering candle's volume being >1.3× the 20-candle average.
• 15-minute ORB — the first three 5-minute candles (9:15–9:30) set the opening range; a close beyond it on a >1.5× volume spike targets 1.5× the range extended past the broken level.

RSI throughout this tab uses a shortened RSI(9) lookback (not the 14-period used on the Price Action tab), matching the spec's instruction to use faster oscillators intraday.

Confluence scorer & Top Picks (Section 6)

Signals are grouped by module family — vwap, cpr_camarilla, volume_profile, ema_orb — and by direction. A pick needs ≥2 distinct confirming modules agreeing on the same direction; anything with only one confirming module lands in Watch instead, never Top Picks.

Module weights: VWAP 25%, CPR/Camarilla 20%, Volume Profile/POC 20%, EMA+ORB 15% (Order Flow's 20% is absent for lack of tick data, so weights are normalised over whichever modules actually fired). vix_multiplier is 0.8 in a low-IV regime, else 1.0.

Cost-aware breakeven check

Every candidate pick is netted against an estimated round-trip cost before it can appear in Top Picks — Section 4's post-2026 SEBI cost constants, expressed in index points for an ATM-option-equivalent scalp (this dashboard has no live option premium feed, so this is an approximation, not a broker quote):

STT (sell side, options) 0.15%, exchange transaction fee 0.03503%, GST 18% — re-verify these against the latest NSE/SEBI circular; they change by circular, not by code release. A pick is rejected (demoted to Watch) if cost_points exceeds 40% of the raw gross move, or if net points after cost is ≤0.

Position sizing

A simple per-trade risk sizer against a default ₹100,000 capital base and 1% max risk per trade (not yet user-configurable):

Lot sizes: NIFTY 65, BANKNIFTY 30, FINNIFTY 60 (post-2026 SEBI larger-lot regime) — also subject to periodic re-verification against live NSE circulars, not a permanent constant.

Nifty 50 filter — Top Bullish / Top Bearish

A second, separately-triggered scan on the same tab (Scan Nifty 50 button, not auto-loaded on page open since it's ~50 symbols' worth of intraday data). It runs the identical engine described above — CPR/Camarilla, VWAP suite, Volume Profile/POC, EMA+ORB, the confluence scorer and the net-of-cost check — against each Nifty 50 constituent stock's own 5-minute intraday candles, one report per stock, in parallel (same ThreadPoolExecutor pattern as the shared scan() used by every other tab).

Every stock's Top Picks (already vetted: ≥2 confirming modules, passing the cost check) are pooled across all 50 names and split by direction:

India VIX is fetched once per scan and reused for all 50 stocks (not refetched per symbol) — it's a single market-wide regime read, not something that varies stock to stock. There is no per-stock Watch list here; a stock with only single-module signals simply doesn't appear in either ranking.

15. Trading Systems & Strategies tab

Named, rule-based strategies built directly from this dashboard's skill references — classical technical analysis, ICT/Smart-Money-Concepts, Wyckoff, and value/growth fundamentals — each one a fixed checklist of conditions evaluated against the same fundamentals / 3-month price action / ICT / options bundle every other tab's scan() already computes. No extra data is fetched.

Every strategy has a bearish (short) mirror sharing the same family, so the tab splits each family into a Bullish leaderboard (left) and a Bearish leaderboard (right) — the same split2 layout as the ICT Shortlist and ICT Strategies tabs.

Two kinds of catalog entry:

• strategy — draws mainly on one or two skills (pure ICT, pure classical trend-following, pure value/growth, pure Wyckoff).
• system — a multi-pillar composite requiring several different skills' conditions to agree at the same time (fundamentals AND momentum AND ICT AND options) — the actual trading system you get by combining skills rather than picking one.

Each side of each family is ranked by its own key score (momentum, ICT, fundamental, or blended composite as appropriate, inverted on the bearish side so "further from the bullish ideal" ranks higher). Entry/stop/target levels on qualifying cards are the same ATR-derived reference points used on every other tab (see section 6) — not recommendations. Discretionary frameworks (ICT, Wyckoff) are encoded here as one consistent, reproducible reading of them, not the only valid one. GET /api/strategies/catalog returns the static rule definitions (12 entries: 6 families × bullish/bearish); GET /api/strategies/scan runs a chosen subset (or the whole catalog) across all 50 stocks and returns them grouped back into families.

16. Institutional Picks tab

Five named fundamental + technical confluence strategies modelled on how bank research desks, systematic/quant funds and veteran discretionary traders actually screen — each a StrategyDefinition/StrategyMatch pair (bullish and its bearish mirror) run off the same fundamentals/price-action/ICT bundle every other tab uses, filtered to its own Top 5 per side, then all 5 merged into one final combined Top 5 Bullish / Top 5 Bearish for a ~1-month swing trade — not a long-term buy-and-hold position. This is a separate catalog from the Trading Systems tab (section 14); the two don't share strategy IDs.

How the final combined Top 5 is ranked

Every stock any of the 5 strategies flagged (on a given side) is merged by symbol. Ranking is conviction first: how many of the 5 strategies agree on that stock and direction, tie-broken by the average condition-match % across the strategies it cleared — not any single strategy's own key score, since those aren't on a common scale across strategies (RSI, ICT score, relative-strength % and fundamental score all mean different things numerically).

Entry/stop/target levels reuse the same ATR-derived 2–6 week reference points as the Trade Shortlist (see section 6) — sized for the ~1-month horizon this tab targets, not a long-term position. GET /api/institutional/catalog returns the 10 static rule definitions (5 families × bullish/bearish); GET /api/institutional/scan runs all 5 across the Nifty 50 and returns each family's Top 5 per side plus the combined Top 5 Bullish/Bearish.

17. Filter bar (top of every tab)

All filtering runs client-side against data already fetched for that tab — changing a filter is instant and never re-triggers a live scan. The Main Dashboard also has four extra local filters of its own (see section 2); the Nifty & BankNifty and Watchlist tabs ignore this bar entirely — the former is always exactly two fixed rows (nothing to filter), the latter is a manually curated list, not a scored/filterable scan (see section 15).

18. Data sources & caveats

• Fundamentals: yfinance quarterly financials, optionally patched from a secondary enrichment source where a line item is missing.
• Prices: yfinance daily OHLCV candles.
• Options summary: Sensibull's public snapshot endpoint (no login required).
• Full option chain (strike ladder, live greeks): requires a Kite Connect session (python -m scripts.kite_login). Everything else works without it.
• ICT / Smart-Money-Concepts: pure computation over the same 3-month daily candles already fetched for Price Action — no separate data source, no extra API cost.
• Liquidity: no new data source — ADTV/turnover reuse Price Action's candles and Fundamentals' market cap, options depth reuses the Options tab's chain fetch, and the bid-ask spread signal reuses a Kite quote batch (one extra call per scan, only when Kite is connected).
• Every API response is cached locally (SQLite, time-limited) so repeat views are fast; "Run scan" / "Load league table" always does a fresh pass across all 50 stocks, which takes 30–90s cold and is much faster once warm.

19. Watchlist

The ☆ next to every stock symbol — in any table row or any shortlist card, on any tab — adds it to your watchlist; a filled ★ means it's already on there, and clicking it again removes it.

• Stored in this browser's localStorage only — nothing is sent to a server, it isn't synced between browsers or devices, and clearing site data clears it too.
• The Watchlist tab doesn't cache a snapshot: every time you open it (or hit Refresh), it re-fetches live fundamentals, 3-month price action and options data for each watched symbol via the same per-stock endpoint the detail drawers use, so the numbers are always current.
• The tab label shows a running count, e.g. "Watchlist (5)".
• Click a stock's symbol in the table to open a 3-month (~90 calendar day) advanced candle chart below the table — the same candlestick/volume/20&50-DMA chart used on the Price Action tab, fetched fresh via /api/price/{symbol}?markers=true.
• 10D / 30D / 90D Chg % columns are the percentage change from the close price 10, 30, and 90 trading sessions ago (capped at the earliest available candle within the fixed 63-session / 3-month window) to the current price.