CHRONOSDAILY
Daily Market Intelligence Model
ChronosDaily is the daily market intelligence model powering IndQuant's Market Pulse. It is a six-layer analytical pipeline designed to deliver a structured, calibrated directional view on Nifty 50 before every trading session, during it, and after it closes.
Model: Chronos Daily | Horizon: 12 hrs | Markets: NSE/BSE | Assets: Nifty 50, Sensex
What is ChronosDaily?
ChronosDaily processes over 60 market signals every trading session and produces a structured directional assessment of the Nifty 50. Each output includes a directional label (Bullish, Cautiously Bullish, Neutral, Cautiously Bearish, or Bearish), a confidence grade, a set of supporting drivers, support and resistance levels derived from the options chain, and a curated stock watchlist scored by a five-dimensional model.
ChronosDaily is not a language model wrapper. The directional call is computed deterministically by a rule-based scoring engine. The language model component performs macro synthesis and report writing after all quantitative signals have been computed. This separation is a deliberate architectural decision, not a convention. It means the directional call is reproducible, auditable, and consistent across sessions regardless of narrative context.
What ChronosDaily Produces
ChronosDaily operates across four report modes timed to the NSE trading clock: Pre-Market Pulse, Midday Pulse, Post-Market Wrap, and Weekend Pulse. The appropriate mode is selected automatically based on current IST time and the NSE holiday calendar.
Architecture Overview
ChronosDaily is structured as a sequential multi-agent pipeline. Each layer has a distinct analytical mandate and passes enriched context forward to the next. No layer operates without the outputs of all preceding layers. This structure exists because daily market intelligence is not a single inference problem. Data collection, macro interpretation, options decoding, sector rotation classification, directional scoring, and narrative synthesis each require different reasoning strategies. Combining them into a single model produces a system that handles none of them well and makes calibration of individual components impossible.
Layer 1: Market Data Ingestion
Signal sourcing across 60+ live feeds.
Assembles a market snapshot before analysis via parallel pulls. Includes Indian indices (Nifty 50, Bank Nifty, nine sector indices), global benchmarks (S&P 500, NASDAQ, FTSE, DAX, Nikkei), GIFT Nifty futures for pre-open gap signals, India VIX for volatility momentum, and FII/DII institutional flows with rolling three-session momentum checks.
Layer 2: Options Intelligence
Decoding conviction from the Nifty chain.
Analyzes four indicators: PCR as a signal classifier, Gamma Exposure (GEX) to classify market maker regimes (long/short/neutral), IV Skew to detect institutional fear versus retail complacency, and Max Pain strikes alongside OI walls to define realistic support and resistance constraints.
Layer 3: News and Sentiment
Structured classification via LLM.
Tags significant news across relevance, directional implication, sector specificity, and time horizon. Segregates global and domestic triggers (e.g., geopolitical oil impacts vs. RBI actions) into structured feeds for the macro synthesis layer rather than raw, undifferentiated text.
Layer 4: Sector Rotation
Detecting institutional regime shifts.
Classifies sessions into five regimes: Risk-On (offensive pull), Risk-Off (defensive rotations), Broadened (wide participation), Narrowing (concentration risk), or Neutral. Covers Banking, IT, Auto, Metal, and Realty against Pharma, FMCG, Energy, and Infrastructure to align watchlists with prevailing themes.
Layer 5: Direction and Confidence
Deterministic engine with ML modulation.
Computes direction via rule-based code across seven dimensions (Global, VIX, FII, Breadth, Technicals, Options, Rotation). Applies quality gates to issue Neutral on low-conviction days. A parallel ML ensemble provides probability estimates to modulate confidence grades without overriding the core rules.
Layer 6: Macro Synthesis and Narrative Engine
This is where ChronosDaily turns structured signals into a usable daily brief. After the direction engine finalises its call, a frontier‑grade large language model takes over – but only as a writer, not a forecaster. It receives a fully structured snapshot from Layers 1–5 and converts it into a 120–160 word macro narrative, explaining why the call was made, which signals mattered most, and how global and domestic forces are interacting. ​ The model cannot invent extra data or override the rules engine. It is hard‑wired to stay inside the quantitative context it’s given. Global and domestic triggers are listed separately, helping readers see whether today’s setup is being driven more by external risk (US markets, oil, FX, geopolitics) or local dynamics (RBI policy, domestic flows, earnings). ​ In practice, Layer 6 turns a dense multi‑factor state into something a human can scan in under a minute: a clean story, grounded in numbers, that explains the directional label, confidence grade, and key risks without hiding behind jargon or vague sentiment. ​ ​ ​
Stock Watchlist Selection
Each ChronosDaily report includes 4–5 stocks selected by the Composite Signal Score (CSS), a five‑factor model applied uniformly across the entire universe. Stocks are ranked within the current market regime using: relative strength versus Nifty over 5 days, volume versus the 20‑day average, technical setup quality (breakouts, support bounces, SMA alignment), sector alignment with the prevailing rotation regime, and RSI positioning versus statistically significant extremes.
The model is regime‑aware. In trending markets it favours strength and breakout candidates; in downtrends it prioritises oversold names at well‑defined support; in range‑bound conditions it looks for RSI extremes with confirming volume spikes. Sector diversification is enforced so that no more than two stocks from a single sector appear in any one report.
When Models LIke this Go Wrong
ChronosDaily's accuracy quality gate widens the Neutral zone during high-uncertainty periods because the model's predictive power degrades when tail risks dominate. Surprise central bank interventions, unexpected geopolitical escalations, and major earnings events in heavyweight stocks can override systematic signals. The system is designed to recognise the boundaries of its own confidence and communicate them transparently rather than issue a directional label that is not supported by the evidence.
Design Principles
Probabilistic, Not Deterministic
Markets are distributions of outcomes, not single-point forecasts. ChronosDaily is designed to describe where the weight of evidence lies, not to promise a specific closing level. Direction labels and confidence grades express probability under the current market state, grounded in backtested thresholds rather than opinion.
Rules First, Narrative Second
Every directional call is produced by a deterministic rules engine before any language model touches the output. The LLM cannot override or soften a signal because the surrounding story feels persuasive. This separation protects the system from narrative drift, recency bias, and sentiment-driven overrides during stressed markets.
Transparent, Auditable, and Calibrated
ChronosDaily logs every call, every contributing signal, and the subsequent market outcome. This makes the model auditable and continuously calibratable. When a signal stops adding value, its weight is reduced; when regimes change, thresholds are retuned against fresh data. The aim is not perfection on any single day, but systematic usefulness and statistical honesty across many sessions.
Intended Users
- ChronosDaily is designed for investors who want to understand the reasoning behind a directional call, not just receive the call itself. The intended users are:
- Professional traders and HNI investors who require institutional-grade daily analysis without maintaining a dedicated research function
- Systematic traders who want a structured, quantified directional signal with explicit confidence levels that can sit alongside or feed into proprietary models
- Active equity investors making 5 to 15 trades per month who need a consistent, data-backed starting point for daily session bias
- Wealth managers and advisors who require a well-documented analytical framework to anchor morning client communications
Note: ChronosDaily is a research instrument. It does not produce buy/sell signals, but rather provides the 'environmental map' upon which such decisions are made.
EnduraTrend and ChronosDaily: Complementary Instruments
Think of ChronosDaily as your tactical radar (noting the current speed and weather conditions) and EnduraTrend as your long-range navigational chart (predicting the destination and the oceanic currents). We recommend using EnduraTrend to set your strategic exposure and ChronosDaily to manage your tactical entries and exits within that strategy.