14,295

AMFI scheme / plan / option rows

The market is bigger than it looks.

Skip to the research
Chapter two · Identity

2,156

Real distinct open-ended products

Most of those rows are the same fund.

AMFI publishes one line per scheme, per plan, per option. HDFC Flexi Cap appears four times — Direct and Regular, Growth and IDCW — and it is one fund, not four. Collapsing those duplicates is the only thing this step does; nothing is discarded for being small, new or unfamiliar.

A parallel scheme-code-keyed view of the same universe returns 2,037. A product group with no single resolvable Direct + Growth representative is excluded rather than assigned a guessed identity.

Chapter three · The structural screen

474

Passed · from 521 category-mapped

Three questions we can answer from real data.

Can we classify it? Does it have a real, current NAV? Is its identity unambiguous? Only five of Prospera's nine categories can be resolved from AMFI's own taxonomy — Flexi Cap, Mid Cap, Small Cap, Debt and Arbitrage. Pharma, Defence, Infrastructure and Energy cannot be sized from this data at all, and the system reports that rather than guessing.

And four checks we deliberately did not run
  • AUM FloorAMFI's NAV file carries no AUM field. Known only for the 56 researched funds.
  • Minimum Track RecordAMFI's NAV file carries no inception date.
  • AUM Trend ScreenNeeds AUM at current / 3M / 12M — not tracked for the broader universe.
  • Expense Ratio ScreenAMFI's NAV file carries no expense-ratio field.
Chapter four · Real market data

19,51,830

Dated NAV records held in-house

Real trailing return, per unit of realized volatility.

Candidates are ranked on real dated NAV history alone — a minimum of 100 real observations and 180 real elapsed days, with the window capped at 365. 459 funds are rankable; 15 are set aside for insufficient history, kept in the table with the specific reason rather than scored with an invented number.

Two things this is not

Not a Sharpe ratio — we have no defensible risk-free rate source, so we do not use the name.

Not a “1Y” window — dense full-market history begins 7 Oct 2025, 225 of the ~250 trading days a real one-year series needs. Each fund’s real elapsed window (~323 days) is reported instead.

Chapter five · The candidate pool

425

Active research candidates

Not every fund gets the same amount of attention.

Research capacity is finite, so pool slots are allocated in proportion to each category's own maximum portfolio allocation — the same category weights the portfolio-construction layer uses. A naive global cutoff was tested live and filled 147 of a 150-slot pool with ultra-low-volatility overnight and debt funds.

  • DEB279
  • FC35
  • ARB31
  • MC26
  • SC25
  • PH13
  • INF13
  • ENE3
Now we look under the hood

Thirteen named qualities. Every weight published.

A fund stops being a dot here. Each candidate gets a mark out of ten on thirteen specific things. Each of the thirteen carries a fixed importance weight. The weights sum to exactly 100, never change between funds, and are never adjusted after seeing the answer.

The thirteen scoring parameters, sized by published weightFMQPQTCONVALSTWFMTDSPOPPMOMAMQAUTERSTHS13PARAMETERSWEIGHTS SUM TO 100
FMQ

Fund Manager Quality

12%Analyst judgment
Measures
Investment philosophy, communication quality, and style drift.
Why it matters
The single highest-weighted parameter in the model. In active management, the manager is the product.
Evidence used
Manager identity and history, published commentary, stated approach versus the actual portfolio.
Real coverage
457 / 457 scored. `Fund Manager` evidence exists for 85 schemes.
Chapter seven · Judgment and formula

77%

Of every score is analyst judgment

We are not going to pretend a machine decided.

Data becomes human interpretation becomes a decision. Nine of the thirteen parameters — 77% of total weight — are analyst judgment, by the methodology's own design. Four are pure arithmetic: FMT, AMQ, AUT and ERS. We think judgment is the right way to research a fund, and we publish the split rather than burying it.

77%
23%

Even the 23% that is formula-computed rests on evidence we hold for only part of the market: AMQ is covered 457/457, FMT and ERS for roughly 45 schemes each, AUT for the 56 researched funds only.

Chapter eight · Evidence

6,487

Sourced, dated, objective data points across 457 schemes

Evidence, score and view are three separate layers.

An evidence row can never contain a score. A score row can never contain a narrative view. The separation is enforced at the database schema level, not by convention — so evidence never quietly becomes a recommendation.

Evidence · 6,487 rows

Append-only, each row carrying field name, value, source, source URL, the date the source states it is “as on”, retrieval timestamp and a confidence flag.

Score · 4,113 rows

Nine manual parameters × 457 schemes, exactly complete. Every change ever made is mirrored to an append-only audit trail of 10,797 rows.

View · 18 recommendations

Analyst narrative, thesis, stance, allocation cap and Investment Committee sign-off.

  • 5,401 · AMFI centralized fund-detail API
  • 463 · Value Research
  • 457 · AMFI quarterly AAUM disclosure
  • 38 · LIC official factsheet
  • 41 · ABSL factsheet
  • 29 · SBI factsheet + TER file
  • 23 · HDFC factsheet
  • 17 · Mirae factsheet
  • 8 · Invesco factsheet
  • 8 · Aggregators (labelled, 0.12% of all evidence)
Chapter nine · Building a real score

81.9

Parag Parikh Flexi Cap Fund · FC-001

Thirteen marks, thirteen fixed weights, one score out of 100.

Each parameter's contribution is its raw mark out of ten multiplied by its published weight. The weights never change between funds and are never adjusted after seeing the answer.

Parag Parikh Flexi Cap Fund final score by parameter.
ParamRawWeightContribution
FMQ1012%12.0
PQT1011%11.0
CON1010%10.0
VAL910%9.0
STW79%6.3
FMT88%6.4
DSP98%7.2
OPP88%6.4
MOM67%4.2
AMQ36%1.8
AUT65%3.0
ERS74%2.8
THS92%1.8
Final Score81.9

Rating bands are fixed: ≥75 high conviction · ≥65 moderate buy · ≥55 watchlist · below that, avoid. Note AMQ at 3/10 — a genuinely good boutique scores badly on a parameter that measures the size of the fund house. That is a real limitation of the model, disclosed rather than tuned away.

Chapter ten · The five gates

Five thresholds that do not move.

A fund has to pass five checkpoints, in order, to become a recommendation. Not for a fund we like, and not for a fund house we work with.

  1. Gate 1 · Eligibilityeligibility

    AUM ≥ category floor · track record ≥ category minimum · SEBI clean

    Researched book (56)All 56 researched funds pass.

    V7 pool (425)Genuinely unconfirmed for the majority of the 425. Real AUM and track-record evidence does not exist for them — not passed, not failed.

  2. Gate 2 · Quant Screen≥ 45

    Final Score ≥ 45 · quant sub-score ≥ 20 / 50

    Researched book (56)0 of the 56 sit here; all clear it.

    V7 pool (425)69 of 425 clear it — 16.2%.

  3. Gate 3 · Top 10≥ 55

    Final Score ≥ 55 · Top 10 by score within category

    Researched book (56)3 of the 56 stop here.

    V7 pool (425)Zero of 425. The pool's best score is 52.7.

  4. Gate 4 · Top 5 Review≥ 62

    Final Score ≥ 62 · Top 5 in category · Thesis ≥ 6 · no active Watchlist flag

    Researched book (56)19 of the 56 stop here.

    V7 pool (425)Zero of 425.

  5. Gate 5 · Top 2 · IC≥ 70

    Final Score ≥ 70 · Top 2 in category · written IC narrative · primary/secondary designation

    Researched book (56)34 of the 56 reach it.

    V7 pool (425)Zero of 425. Every category's best candidate falls 17+ points short.

Chapter eleven · Two research layers

These are two systems, not one pipeline.

It would be a tidier story to draw one funnel from 14,295 straight through to 18. It would also be false.

Live today

Current published research book

  1. Fully scored56
  2. Ranked in category27
  3. Recommendations18

Scores 59.589.4, mean 72.72. All 18 recommendations on this site come from here.

Active, and not yet qualifying

V7 active research pipeline

  1. Candidate pool425
  2. Clear Gate 269
  3. Clear Gates 3–50

Same methodology, same thresholds. Scores 36.152.7, mean 43.3. Zero of the 425 currently qualify.

We scored 425 new candidates under exactly the same system that produced our current book. Not one beat what we already recommend. So the 18 stayed — but they stayed because we checked, not because we did not look.

Chapter twelve · Nine categories

We don't rank funds against the entire market.

Ranking is strictly within a category. A Debt fund scoring 89.4 does not displace a Flexi Cap fund scoring 81.9 — they are not competing for the same job in a portfolio. Each category also carries a maximum share of a portfolio it can take.

  • Debt30%
  • Flexi Cap25%
  • Mid Cap20%
  • Arbitrage20%
  • Small Cap15%
  • Infrastructure15%
  • Pharma & Healthcare12%
  • Defence12%
  • Energy10%
Chapter thirteen · The watchlist

Six funds are monitored — and two of them are still picks.

A flag is a live monitoring instruction, not a verdict. Our standing rule reads: “Any watchlisted fund excluded from Top 2 regardless of score. IC must formally remove to reinstate.” Two of the eighteen currently override it.

Quant Small Cap Fund

71
Quant MF · Small Cap · SEBI Risk

Upgrade conditionSEBI all-clear + score ≥ 72 for two months

JM Flexicap Fund

62
JM Financial MF · Flexi Cap · AUM Watch

Upgrade conditionAUM crosses ₹8,000 Cr organically

Quant Flexi Cap Fund

65
Quant MF · Flexi Cap · SEBI Risk

Upgrade conditionSEBI case closed + AMQ ≥ 8

HDFC Defence Fund

79.1
HDFC MF · Defence · Valuation Watch

Upgrade conditionNo upgrade — already Primary; monitor P/E only

IC overridePrimary pick, Defence. Retained as the category primary despite an active Valuation Watch flag. Capped at 12% allocation.

Quant Infrastructure Fund

70.1
Quant MF · Infrastructure · SEBI Risk

Upgrade conditionSEBI cleared = full secondary

IC overrideSecondary pick, Infrastructure. Maintained as the #2 pick on a satellite basis while the SEBI matter is open. Max 5% allocation only.

SBI Small Cap Fund

63.4
SBI MF · Small Cap · Downgraded

Upgrade condition5Y CAGR recovers above 17% for two consecutive months

Chapter fourteen · The narrowing

18

Recommendations

Thousands, hundreds, dozens, eighteen.

14,295 listings · 2,156 real products · 474 through the structural screen · 425 active candidates · 56 fully scored · 27 ranked · 18 recommendations, 9 primary and 9 secondary across 9 categories.

What comes out the other end

18

9 primary · 9 secondary · 9 categories

A first pick and a back-up, in each of nine categories.

Primary is intra-category rank one at the point of Committee review; secondary is rank two. Both cleared identical gates. Every one of these comes from the published research book — not from the 425-fund V7 pool, which has not yet promoted a single fund.

  1. FCFlexi CapCore HoldMax 25% allocation
    PrimaryParag Parikh Flexi Cap
    81.9
    SecondaryHDFC Flexi Cap
    74.8
  2. MCMid CapTactical OWMax 20% allocation
    PrimaryMotilal Oswal Midcap
    79.3
    SecondaryNippon India Growth
    78.3
  3. SCSmall CapSelectiveMax 15% allocation
    PrimaryNippon India Small Cap
    79.4
    SecondaryHDFC Small Cap
    73.1
  4. PHPharma & HealthcareModerateMax 12% allocation
    PrimaryMirae Asset Healthcare
    77.4
    SecondaryNippon India Pharma
    77
  5. DEFDefenceHigh ConvictionMax 12% allocation
    PrimaryHDFC Defence
    77.4
    SecondaryMotilal Nifty Defence Index
    71.8
  6. INFInfrastructureStructural OWMax 15% allocation
    PrimaryICICI Pru Infrastructure
    82.5
    SecondaryQuant Infrastructure
    71.7
  7. ENEEnergySelectiveMax 10% allocation
    PrimarySBI Energy Opportunities
    76.8
    SecondaryDSP Natural Resources New Energy
    72.8
  8. DEBDebtOverweightMax 30% allocation
    PrimaryICICI Pru Corporate Bond
    89.4
    SecondaryICICI Pru Liquid
    86.9
  9. ARBArbitrageLiquidity ToolMax 20% allocation
    PrimaryABSL Arbitrage
    82.7
    SecondaryInvesco India Arbitrage
    79.5

Scores are the published 13-parameter Final Score for each fund. They are comparable within a category and deliberately not across categories.

The part most research pages leave out

We don’t know everything.

Stated plainly, because a client who discovers one of these independently will discount everything else on this page.

01
We have not researched most of the market.
56 of 2,156 products carry a full Final Score. In Debt that is 5 of 365 — about 1%. A Debt recommendation is "the best of a deliberately small researched set", not "the best in India".
02
Four of nine categories have no denominator.
AMFI groups every sector and thematic fund into one bucket. We can say we researched five Pharma funds; we genuinely cannot say what fraction of the Pharma universe that is.
03
77% of every score is analyst judgment.
Nine of thirteen parameters are manual input, by the methodology's own design. No data feed can close that — it would require changing the methodology.
04
381 of 457 assessments are preliminary, not deep.
They are fact-anchored and category-templated, and each is labelled verbatim as a holistic estimate. Only 76 funds — 17% — have been individually researched against primary sources for every parameter.
05
Three of the four calculated parameters have thin evidence.
FMT: 45 schemes. ERS: 45 schemes. AUT: the 56 researched funds only. Only AMQ is fully covered at 457/457. The 'objective 23%' is less objectively sourced across the broad pool than across the legacy book.
06
Two recommendations override our own Watchlist rule.
HDFC Defence and Quant Infrastructure are current picks despite the rule that a flagged fund is excluded from the Top 2 regardless of score. Both carry written Committee rationale and a hard allocation cap. It is an override, and we describe it as one.
07
Our NAV history is not yet a full year.
Dense full-market history begins 7 Oct 2025 — 225 of the ~250 trading days a real one-year series needs. So nothing is labelled a "1Y return"; each fund's real elapsed window (~323 days) is reported instead.
08
The two research layers do not yet join.
All 18 recommendations come from the legacy researched book. Zero of the 425 V7 candidates qualify. The 18 have been verified against the broader pool — they stayed because we checked, not because we did not look.
  • 14,295listings.
  • 2,156real products.
  • 425active candidates.
  • 13parameters.
  • 5gates.
  • 18decisions.

This is how Prospera researches.

Figures on this page are a point-in-time snapshot of the Research Engine v7 methodology, verified 31 August 2026 against the production database, the source workbook and the live application.