How Our Multi-Agent AI
Calculates Your Score
MicroStocks does not use a single model or a black-box formula. Every conviction score, risk rating, and AI verdict is the product of a structured debate between four specialized AI agents — each with a different analytical lens — moderated by a CIO-level synthesizer.
The AI Institutional Committee
Every analysis begins with three parallel "desk" agents running simultaneously, then a fourth CIO agent that synthesises their outputs into a final verdict. This mirrors how institutional investment committees operate — removing single-model bias.
Macro Desk
Macro-Economist
Key Inputs
- › Sector tailwinds / headwinds
- › RBI / Fed monetary policy stance
- › Global commodity price trends
- › FII / DII flow patterns
- › INR / USD currency risk
Output
Macro sentiment (BULLISH / NEUTRAL / BEARISH) + confidence score 0–100
Quant Desk
Algorithmic Trader
Key Inputs
- › RSI (14-day), MACD signal
- › 20-day & 50-day SMA alignment
- › Volume momentum (5/20-day)
- › 52-week high/low positioning
- › Relative strength vs Nifty
Output
Quant sentiment + technical regime classification
Fundamental Desk
Value Investor
Key Inputs
- › P/E ratio vs. sector peers
- › ROE, Debt/Equity, EBITDA margin
- › Revenue and earnings growth CAGR
- › Promoter pledging %
- › Order book visibility
Output
Fundamental sentiment + margin-of-safety assessment
CIO / Risk Manager
Chief Investment Officer
Key Inputs
- › All 3 desk outputs
- › Inter-agent agreement score
- › Stop-loss invalidation levels
- › Portfolio allocation ceiling
- › Macro risk overlay
Output
Final verdict, conviction score 0–100, max allocation %
Data Source Waterfall
We fetch financial data from multiple sources in priority order, falling back gracefully to ensure every ticker gets coverage even during rate limits or outages.
Price, volume, fundamentals, history
Technical indicators, earnings, global macro
Real-time headlines for sentiment scoring
Agent reasoning, synthesis, natural language verdict
Rule-based verdict when LLM APIs are unavailable
Score Definitions
Conviction Score
0 – 100
Overall committee confidence. ≥70 = high conviction long. ≤30 = avoid or hedge. Derived from the CIO agent after weighing desk agreement.
Micro-Score
0 – 10
Normalised composite score blending fundamentals (40%), momentum (30%), and news sentiment (30%). Used for relative ranking within a sector.
Risk Score
0 – 10
Higher = higher risk. Factors: beta, debt/equity, promoter pledging, ASM/GSM status, and tail-risk from news. ≥7 requires confirmation before entry.
Data Freshness & Caching
AI verdicts are computationally expensive — each full committee analysis involves 4 sequential/parallel LLM calls and data from up to 5 external sources. To ensure sub-second page loads (critical for mobile users and Googlebot), we use a two-tier caching strategy:
- Redis L1 cache (TTL: 1 hour):Verdict results are stored immediately after computation. Any subsequent request in the same hour returns the cached result in <50ms.
- Background pre-warmer (every 4 hours): A scheduled job proactively re-computes verdicts for our top 50 tracked tickers before the cache expires, so cache misses essentially never occur for popular stocks.
- Stale-while-revalidate (24h): The CDN edge layer serves stale verdicts for up to 24 hours while a fresh computation is triggered in the background — preserving page speed even during revalidation cycles.
Price data and technical indicators refresh every 5 minutes during market hours. News sentiment scores refresh every 3 minutes.
Important Disclaimer
All analysis produced by MicroStocks' AI Committee is for informational and educational purposes only. It does not constitute financial advice, investment recommendations, or a solicitation to buy or sell any security. MicroStocks.in is not registered with SEBI as an investment advisor. Always consult a qualified financial professional before making investment decisions. Past performance of any model or score does not guarantee future results.
Ready to see these scores in action?