STOCK FUTURES — OI MATRIX
(ΔOI vs yesterday close)
FUTURES VOLUME LEADERS
INTRADAY P&L
--
Waiting for position
🔥0 STREAK
NIFTY SPOT
--
INDIA VIX
--
DAY TOTAL
--
Realized + Unrealized
PORTFOLIO
₹5,00,000
Base Capital
TRADES
0
STATUS
IDLE
◆ INTRADAY POSITION
NO POSITION
WAITING FOR OPPORTUNITY
AI is analyzing the market
★ INTRADAY BRAIN
Scan #0
WAITING
Waiting for first brain scan...
ACTIVE TRIGGERS
☷ SCAN LOG
0 scans
TIME
ACTION
CONF
STRATEGY
SPOT
VIX
REASONING
◑ MARKET STATE
--
EMA 9
--
EMA 20
--
VWAP
--
PCR
--
MAX PAIN
--
HRS LEFT
--
Δ PORTFOLIO GREEKS
DELTA
--
GAMMA
--
THETA/DAY
--
VEGA
--
☰ EVENT LOG
0 events
--:--:--SYSDashboard initialized. Waiting for data...
▼ TRADE HISTORY
Today & History
NO TRADES YET
DAY P&L0
📈 PORTFOLIO GROWTH
0.00%
TOTAL P&L
--
WIN RATE
--
BEST
--
WORST
--
💡 TRADE LESSONS
Live Feedback
👻 SHADOW P&L — Actual vs What-If
LOADING SHADOW DATA...
📚 INTRADAY LEARNINGS
LOADING LEARNINGS...
Fetching AI trading insights
WEEKLY P&L
--
Loading
NIFTY SPOT
--
INDIA VIX
--
REALIZED
--
Closed trades
WEEKLY CAPITAL
₹20,00,000
Base ₹20L
TRADES
0
STATUS
IDLE
📈 OPEN POSITIONS
LOADING
LOADING...
🧠 WEEKLY BRAIN
LOADING...
📄 WEEKLY SCAN HISTORY
Time
Action
Conf
Spot
VIX
Reasoning
🌎 MARKET STATE
LOADING...
💰 TRADE HISTORY
LOADING...
📈 PORTFOLIO GROWTH
LOADING...
📋 EVENT LOG
LOADING...
📚 WEEKLY LEARNINGS
LOADING WEEKLY LEARNINGS...
Fetching weekly trading insights
MONTHLY P&L
--
Loading
NIFTY SPOT
--
INDIA VIX
--
REALIZED
--
Closed trades
MON CAPITAL
₹20,00,000
Base ₹20L
TRADES
0
📅 MONTHLY IC POSITIONS
LOADING
LOADING...
Monthly Iron Condor — 25-30 DTE theta capture
🧠 MONTHLY BRAIN
LOADING...
📄 MONTHLY SCAN HISTORY
Time
Action
Conf
Spot
VIX
Reasoning
🌎 MARKET STATE
LOADING...
💰 TRADE HISTORY
LOADING...
📈 PORTFOLIO GROWTH
LOADING...
📋 EVENT LOG
LOADING...
📚 MONTHLY LEARNINGS
MONTHLY IC LEARNINGS
Entry window: first 3 trading days of each month. Learnings appear after the first trade cycle.
SCANNING MARKETS
⚠ EXIT ALL POSITIONS
This will immediately close ALL open legs at market price. This action cannot be undone.
Wrong password
DECAY P&L
--
Waiting for 09:57
NIFTY SPOT
--
STATUS
IDLE
CE P&L
--
PE P&L
--
SL HITS
0
TRAILDISARMED
PEAK--
GIVEBACK--/--
Arms at peak ≥ ₹500/lot × lots. Exits if giveback exceeds max(₹100/lot × lots, 25% of peak).
PERFORMANCE · SINCE 8 MAY (DECAY EPOCH)
CLOSED TRADES
--
Since 8 May
WIN RATE
--
--
TOTAL P&L
--
--
MARGIN
--
--
SHARPE
--
Annualized
MAX DD
--
--
OPEN POSITIONS
--
--
⚡ DECAY POSITIONS
LIVEIDLE
⚡ DECAY MASTER
Sells ATM straddle + strangle at 09:57. SL 35% per leg.
📋 ACTIVITY LOG
No activity yet — bot starts at 09:57
📈 CUMULATIVE REALISED P&L — SINCE 8 MAY
—
No closed decay trades since 8 May yet.
📊 INTRADAY P&L CURVE — LIVE + SHADOWS
—
No tick data yet — chart populates after first 09:57 entry.
📅 P&L CALENDAR
No trades yet
💰 TRADE HISTORY
No trades yet
SHADOW P&L
--
points · selected config
TRADES
--
--
WIN RATE
--
--
AVG / TRADE
--
--
WORST DAY
--
premium SL caps the tail
PHASE
SHADOW
paper — no live orders
PAPERAM MR-Buy: spot moves ±40pt off the 09:20 anchor → BUY the decayed ATM leg · replayed nightly on the real 1-min option store (18:40)
🔴 PILOT LEDGER — LIVE + PAPER
No live trades yet — pilot arms weekdays from 09:21 (champion bracket, 1 lot).
📈 CUMULATIVE SHADOW P&L
—
No shadow trades yet.
💰 TRADE HISTORY
Loading…
Rows up to 03 Jul are seeded BACKFILL (research replay); the deploy decision counts FORWARD days only (target 40-60 incl a trending stretch). ₹/lot = points × 65. Cost 2pt/round-trip included.
BROOKS P&L
--
paper / live
NIFTY SPOT
--
live LTP
NIFTY FUT
--
tradeable
ALWAYS-IN
--
STATUS
IDLE
TRADES TODAY
0
SIGNALS TODAY
0
HALTED
no
--PID--up--last tick--next bar--caps--
ALWAYS-IN BIAS:----
📊 LAST 5-MIN BAR
Waiting for first bar close...
🔥 INDICATORS
FUT
Loading...
👀 SHADOW SIGNALS · THIS BAR
Waiting for first bar...
📈 ACTIVE BROOKS TRADE
FLAT
No open trade
All setups start disabled. Enable from the grid below to go live.
⚙ SETUPS · PER-STRATEGY PERFORMANCE
Loading...
📢 TODAY'S SIGNALS (incl. shadow)
No signals yet today
QUANTSTATS RISK TEARSHEETS
Risk Metrics · Equity Curve · Drawdowns
Regenerated automatically at EOD (15:45 IST). Manual regen runs ops/generate_tearsheet.py.
EQUITY HERO · ML STOCK-PICKING
LightGBM big-mover (label: +10% before -5% within 7d) · honest monthly point-in-time top-600 universe · Live exits: +10% TP / -5% SL via GTT, force-exit at 7 trading days
MODE: --updated --:--:--
Trained on bullforce (honest point-in-time top-600, 21-trading-day retrain), executed daily on droplet @ 15:15 IST same-day-close. Capital cap ₹5L. Orders tagged EQUITY_DISC_H10. Strategy: take names clearing the 98th-percentile model confidence (floor 3), enter LIMIT at +0.3% above LTP near the close, exit on TP/SL or after 7 trading days. Honest survivorship-corrected backtest: ~46% hit rate, seed-mean ~+80% CAGR / Sharpe 3.2 (single-path headline higher; see validation dossier).
CLOSED TRADES
--
Since deploy
WIN RATE
--
--
TOTAL P&L
--
--
CAPITAL
--
--
SHARPE
--
Annualized
MAX DD
--
--
OPEN POSITIONS
--
--
📈 PORTFOLIO GROWTH (CUMULATIVE P&L)
--
No EOD snapshots yet — first snapshot writes today @ 15:25 IST after positions exist.
🎯 TODAY'S PICKS --
--
#
SYMBOL
LTP
SCORE
CATEGORY
Loading…
📊 OPEN POSITIONS
--
A/D:--advancing--declining--flat
SYMBOL
QTY
AVG
LTP
DEPLOYED
SL
TP
P&L
DAYS
Loading…
🏆 BEST TRADE
No trades yet.
🔌 WORST TRADE
No trades yet.
📝 TRADE HISTORY
--
SYMBOL
ENTRY
EXIT
QTY
ENTRY ₹
EXIT ₹
P&L
REASON
No closed trades yet.
EQUITY RUNNER · v4 BIG-MOVE
CPCV-validated v4 big-move model (PBO 0.00) · 600-stock liquid universe · confidence ≥ p98 (τ=0.828) · Regime gate: skip when trend200=0 & breadth50<0.30 · Live exits: +10% TP / -5% SL via GTT, force-exit at 10 trading days
MODE: --updated --:--:--
Trained on bullforce (v4 big-move, data through 2026-05-15); picks generated on bullforce and SCP-pushed to the droplet, executed daily. DRY-RUN PILOT — capital cap ₹50k (separate pool from Hero/Star). Orders tagged EQUITY_RUNNER. Strategy: triple-barrier (+10% TP / -5% SL / 10-day horizon) on a 600-name liquid universe; enter only when classifier confidence ≥ p98 (τ=0.828) AND the regime gate passes (skip when trend200=0 & breadth50<0.30); max 20 concurrent, 5% per stock. CPCV validation: mean CAGR +26%, PBO 0.00 — the only config that passed combinatorial purged cross-validation cleanly.
CLOSED TRADES
--
Since deploy
WIN RATE
--
--
TOTAL P&L
--
--
CAPITAL
--
--
SHARPE
--
Annualized
MAX DD
--
--
OPEN POSITIONS
--
--
📈 PORTFOLIO GROWTH (CUMULATIVE P&L)
--
No EOD snapshots yet — first snapshot writes today @ 15:25 IST after positions exist.
🎯 TODAY'S PICKS --
--
#
SYMBOL
LTP
SCORE
CATEGORY
Loading…
📊 OPEN POSITIONS
--
A/D:--advancing--declining--flat
SYMBOL
QTY
AVG
LTP
DEPLOYED
SL
TP
P&L
DAYS
Loading…
🏆 BEST TRADE
No trades yet.
🔌 WORST TRADE
No trades yet.
📝 TRADE HISTORY
--
SYMBOL
ENTRY
EXIT
QTY
ENTRY ₹
EXIT ₹
P&L
REASON
No closed trades yet.
🛡️ HEDGE (PAPER) no real orders · live hedge dropped 2026-08-10 at ₹−44,082
loading…
arm
legs
premium paid
recovered
mark (bid)
cum P&L
% of cap
no paper runs yet
Paper fills: buys at ask, sells at bid, marks at bid (liquidation value). Arms: ex-live 3% OTM daily re-target vs the dd_hedge_study 8% OTM fixed monthly. Real-hedge final for reference: ₹−44,082 (−0.88% of cap, 2026-07-23 → 08-10).
Search spaces v1–v7 (sampled in order, dedup by config hash)
v1 core
Barrier-geometry grid: TP 6–20% × SL 3–10% × hold 5–20d × tau p97/98/99 × floor 0/3/5. Close entry, fixed-barrier exit. The label IS the bet — each geometry retrains its own 47-model walk-forward bank. ● champion-v2 (tp15/sl5/h7) found here via targeted sweep
v2 entry
Entry-time axis: 09:30 / 10:15 / 13:00 / 14:30 vs same-day close. Outcomes resolved on 15-min bars (34M-row intraday moat) — no daily-bar guessing about which barrier hit first.
v3 exit
Exit-style axis: 8%-trail-off-peak (let winners run) and partial take-profit, vs fixed barriers. Trailing variants showed +56% mean CAGR historically but extreme variance — gate decides.
v4 overlay
Calendar-timing gates: Monday-entry (weekend-gap momentum, 56% vs 41% precision single-split) and turn-of-month (institutional/SIP flows). Tested as GATES, never as model features.
v5 universe
Universe axis: top-300 (concentration in liquidity) and absolute ₹22cr-turnover screen, vs the default rank-based monthly PIT top-600.
Short-horizon fast moves: TP 3/5/8% × SL 2/3/4% × hold 1–3 days. Different turnover regime — cost-fragility is the expected killer; the 1% CPCV gate does the filtering.
Deep checks — what a candidate must survive before touching capital
1 per-year decay
Calendar-year CAGR decomposition on the honest universe. The edge must hold in RECENT years, not just the 2022-24 smallcap cycle. (Champion-v2 passed: 2025 +55% / 2026 +31% vs incumbent +30/+18.)
2 cost curve
Portfolio + CPCV per-trade at 0.55% / 1.0% / 1.5% round-trip. Explicit Zerodha CNC floor is ~0.29% before slippage — thin-edge high-turnover configs die here by design.
3 seed stability
Full walk-forward bank retrained on alternative LGB seeds — SYMMETRIC: the incumbent is seed-stressed too. Compare seed-means, not lucky single paths. (v2: +93.6/3.25 seed-mean vs incumbent +84.5/3.18.)
4 basis + timing
Survivorship: honest monthly-PIT universe only (today's-600 inflated the old headline ~21 CAGR pts). Execution match: close-entry sim for close-entry live (the next-open gap was worth 65 pts).
5 human gate
Winner’s-curse discount on targeted sweeps (best-of-N is inflated), then shadow / pilot scale, then a DELIBERATE user decision — the factory alerts, it never self-promotes.
Live champion — month-wise backtested returns (honest universe, 0.55% cost)
-
Single-seed walk-forward path; the seed-mean expectation is lower (see the PDF, section 6). Full methodology, validation and risk disclosures in the Strategy Brief.
Status pushed from the ML server every 30 min. All results on the honest point-in-time universe (survivorship-corrected). A candidate is NOT a promotion: deep checks + shadow precede any live change.
Option Buying Factory
Search for a NIFTY Buy-Side Edge
loading…
Shadow configs — live scoreboard (all / forward-only)
config
n
win%
avg pt
total pt
worst
fwd n
fwd win%
fwd avg
fwd total
Idea families searched & verdicts (2026-07-05 sweep)
directional open
Opening-features direction prediction. WF OOS hit 48.8% — coin flip. DEAD.
EMA slope stacks
EMA 5/10/15 bull/bear alignment. P(up) ≈ 0.50 both ways. DEAD.
3m→5m scalp
Momentum persistence 0.483; ≥20pt follow-through only 16%. DEAD.
buy the moving leg
Momentum buy ATM & 1-3 OTM: −22 to −28pt/trade. OTM never helps. DEAD.
TP-no-SL grids
79% win but −EV (4 big losses erase 15 wins); full 6×6 TP/SL grid all ±noise. DEAD.
pin-bars
Direction 48.8%, annSharpe −0.43, negative 3 of 4 years. DEAD.
PM-MR (12:00)
Approx said strongest leg (+4-5). REAL prices: −5.3/trade — afternoon theta on the decayed leg is 2.3× the approx. KILLED BY REAL REPLAY.
gap-fade G60
REAL prices: −7.5/trade — open IV-crush swamps delta (spot moved 25pt favourably, long PE still lost). KILLED BY REAL REPLAY.
ATR trigger + HH/HL veto
Approx Sharpe 10.7 vs 5.4 fixed — but real June replay WORSE. Approx-only; parked, not promoted.
Fisher/stoch + LGB gate
Oscillator hard filters kill the edge (n collapses). LGB gate flips sign under feature perturbation at n=19 — UNSTABLE, removed from the deploy path.
premium bracket
● SURVIVOR. Buy decayed leg, exit in premium space TP+12/SL−20/EOD: +7.4pt/trade, 79% win, worst −24, all 30 grid cells positive (plateau). Ungated, GTT-able.
armed trail
● VARIANT. SL−20 until +12, then trail 5 off peak: +9.2/trade, identical losing days & downside-dev, Sortino 0.45 vs 0.37. Needs runner days.
Honesty gates — what a buying config must survive
1 approx harness
3y walk-forward on 5-min spot+VIX with a delta/theta premium sim. HYPOTHESIS GENERATOR ONLY — its errors run optimistic (PM leg +4-5 approx vs −5.3 real).
2 real-price replay
Every config re-run on the 1-min option store with real premium fills + 2pt cost. Killed 2 of 3 approx-validated legs the same night they were proposed.
3 plateau check
A surviving cell must sit on a plateau (all 30 TP×SL cells positive), never a lone spike. %-of-entry targets failed here; absolute-point targets passed.
4 perturbation
Change an implementation detail (gate features, exit convention) — if the sign flips, it was never real. This retired the LGB confidence gate.
5 forward shadow
40-60 FORWARD paper days on real closes (backfill excluded), must include a trending stretch. Nightly cron 18:40 after the option-store sync.
6 human gate
The shadow reports; it never self-promotes. Live deployment is a deliberate user decision with lot-sized risk (₹/lot = pt × 65).
Last shadow trade (champion)
-
Status pushed daily 18:40 from the ML server after the option-store sync. Sell-side context: the same trigger’s NON-directional opening vol is what the Decay/IC books already monetise — this factory hunts the residual buy-side reversion. See FNO BUYER tab for the trade-by-trade shadow ledger.
⚡ SELL FACTORY nightly 19:30 · bar = random-time selling, not zero
loading…
Promoted cells — forward paper-shadow (verdict at 40 days)
cell
days
n
mean%
win%
In-sample leaderboard (top 10 by edge vs null)
cell
n
mean%
edge%
win%
p
folds
Gates: beats day-matched random-selling null p<.05 + ≥4/5 day folds → promoted to forward shadow; ≥40 shadow days → verdict-ready. Live promotion is a human decision that must beat the Decay bot’s economics.
Commentary
AI Market Commentary Feed
Live mirror of the NIFTY commentary bot's Telegram messages. Fires every 15 minutes during market hours.
Loading...
Order Flow
NIFTY F&O
Live 5-level depth imbalance + tick-rule CVD on the index future, per-strike OI buildup, and volume-spike detection. CVD is tick-rule-approximated — Kite has no per-trade aggressor flag, so up-tick volume is read as buying, down-tick as selling.
Loading...
—
Cumulative Vol Delta
--
tick-rule · since 09:15
Depth Imbalance
--
bid vs ask · 5 levels
Spot
--
ATM --
Net OI Flow
--
Feed
--
WebSocket ticker
Depth of Market · future · 5-level
Loading depth…
Cumulative Volume Delta · intraday
OI Buildup · vs day-open · near ATM
CALLS (CE)
PUTS (PE)
Flow
ΔOI
OI
LTP
Strike
LTP
OI
ΔOI
Flow
Loading…
Volume Spikes · future · >3× avg
No spikes yet.
Market Structure
NIFTY
Net dealer GEX vs spot, computed from intraday option-chain OI (calls +, puts −). Above the flip = positive gamma (vol-suppressing, range-bound, IC-friendly); below = negative gamma (vol-amplifying, trending).
Loading...
Gamma Flip Level
--
Spot
--
Regime @ Spot
--
Spot vs Flip
--
Expected Move · straddle
--
SPAN Margin · 1-lot straddle
--
Bull / Bear Power · 5m EMA13
--
Ultimate Osc · 400-EMA (1m)
--
Lin Regression · 60×1m
--
VWAP · session
--
Net GEX vs Spot
── Flip│ Spot
OI Magnets
Heavy open-interest strikes pin price near expiry — walls = where writers defend.
Max Pain
--
Call Wall · Resistance
--
Put Wall · Support
--
PCR (OI)
--
Open Interest by Strike
█ Call OI█ Put OI── Max pain│ Spot
Intraday Trend
Each metric normalised to its own day-range — compare shape & timing, not absolute scale. Toggle lines:
Strike Picker · Buy-Side
--
Score = 50 + EV% (clamped ±50) − illiquidity drag (spread >2% NFO / >4% BFO) − chase/extended-entry (−15) − rich-IV vega drag (VIX>20, −5). Candidates ranked by Score among positive-EV strikes (★ = best to buy); EV/lot breaks ties. Vega shown per 1% IV move per share. Deterministic decision-support from the live chain — nothing is auto-traded.
Chain pulled live from Kite — prior-session OI + close before market open, real-time during the session — refreshed ~60s while this tab is open. Dealer-sign convention is the standard calls-long / puts-short heuristic; the flip is the zero-crossing of net GEX. OI magnets reflect the same chain: call wall = max CE OI (upper magnet), put wall = max PE OI (lower magnet), max pain = strike of least total writer payout.
Market Structure
SENSEX
Net dealer GEX vs spot, computed from intraday option-chain OI (calls +, puts −). Above the flip = positive gamma (vol-suppressing, range-bound, IC-friendly); below = negative gamma (vol-amplifying, trending).
Loading...
Gamma Flip Level
--
Spot
--
Regime @ Spot
--
Spot vs Flip
--
Expected Move · straddle
--
SPAN Margin · 1-lot straddle
--
Bull / Bear Power · 5m EMA13
--
Ultimate Osc · 400-EMA (1m)
--
Lin Regression · 60×1m
--
VWAP · session
--
Net GEX vs Spot
── Flip│ Spot
OI Magnets
Heavy open-interest strikes pin price near expiry — walls = where writers defend.
Max Pain
--
Call Wall · Resistance
--
Put Wall · Support
--
PCR (OI)
--
Open Interest by Strike
█ Call OI█ Put OI── Max pain│ Spot
Intraday Trend
Each metric normalised to its own day-range — compare shape & timing, not absolute scale. Toggle lines:
Strike Picker · Buy-Side
--
Score = 50 + EV% (clamped ±50) − illiquidity drag (spread >2% NFO / >4% BFO) − chase/extended-entry (−15) − rich-IV vega drag (VIX>20, −5). Candidates ranked by Score among positive-EV strikes (★ = best to buy); EV/lot breaks ties. Vega shown per 1% IV move per share. Deterministic decision-support from the live chain — nothing is auto-traded.
Chain pulled live from Kite — prior-session OI + close before market open, real-time during the session — refreshed ~60s while this tab is open. Dealer-sign convention is the standard calls-long / puts-short heuristic; the flip is the zero-crossing of net GEX. OI magnets reflect the same chain: call wall = max CE OI (upper magnet), put wall = max PE OI (lower magnet), max pain = strike of least total writer payout.