Executive Summary
This study measures how often a lump-sum holding in a Large Cap Direct Growth scheme ended with a negative annualised return, across 5 holding horizons: 1Y, 3Y, 5Y, 7Y and 10Y. Each observation is a daily-start, point-to-point rolling window. No SIP cash flows are modelled.
Key findings from the internally validated dataset:
- Loss probability falls from 16.44% at 1Y (116,493 daily rolling windows across 33 Direct Growth schemes) to 1.47% at 3Y, 0.16% at 5Y, and 0.0% at 7Y and 10Y.
- The zero-loss horizon, the shortest tested horizon with 0.0% loss probability, is 7 years. In the 54,106 daily rolling windows across 24 Direct Growth schemes at 7Y, no window ended negative.
- The weakest 7Y outcome was +4.21% CAGR (Taurus Large Cap Fund, 23 Mar 2013 to 23 Mar 2020). The weakest 10Y outcome was +8.03% CAGR.
- The 5Y residual loss sits in 2015 start dates (1.55% of windows), with a worst 5Y outcome of -3.43%.
The 7Y and 10Y results rest on a limited set of start dates, and the universe contains only currently active schemes. The findings describe observed history and do not predict future returns.
Methodology
| Parameter | Specification |
|---|---|
| Source | nav_history daily NAV |
| Plan selection | Direct plan, Growth option, one series per scheme |
| Return metric | Annualised CAGR per window |
| Window step | Daily (calendar), NAV forward-filled over non-trading days |
| Data floor | 2013-01-01 (Direct plans launched Jan 2013) |
| Window start dates | From 2013-01-01 to each horizon's data_end |
| Dropped windows (gap) | 0 |
| Excluded schemes | 13 (no Direct Growth plan) |
| Inflation threshold | 6.0% (RBI CPI upper tolerance band) |
Universe and survivorship bias. The universe is currently active schemes only. Closed or merged schemes are excluded, which introduces survivorship bias. A universe that omits discontinued schemes can show a different loss profile from one that includes them. The data cannot size that difference.
Category label. The label reflects the current SEBI classification, and earlier windows use today's membership. The category_since date in the snapshot is 2017-10-06. This study makes no claim about the category's history before that date, and none about any period before the 2013-01-01 data floor.
Window counts. Each window_count below is the number of daily N-year rolling windows pooled across the stated number of Direct Growth schemes. The scheme count is a separate figure and is not a window count. Returns are lump-sum and point-to-point. Tax is not modelled.
Loss Probability by Holding Horizon
| Horizon | Daily rolling windows | Direct Growth schemes | Window data_end | Loss probability | Inflation-beat probability (6.0%) |
|---|---|---|---|---|---|
| 1Y | 116,493 | 33 | 2025-10-01 | 16.44% | 66.02% |
| 3Y | 93,372 | 30 | 2023-10-01 | 1.47% | 93.65% |
| 5Y | 72,375 | 27 | 2021-10-01 | 0.16% | 95.96% |
| 7Y | 54,106 | 24 | 2019-10-01 | 0.0% | 99.89% |
| 10Y | 28,952 | 22 | 2016-10-01 | 0.0% | 100.0% |
Loss probability declines at each successive tested horizon. The first horizon with 0.0% loss is 7Y, so zero_loss_horizon_years is 7. Other holding periods were not tested, so the study makes no claim about them.
Return Distribution by Horizon (annualised CAGR, %)
| Horizon | 5th pct | 25th pct | Median | 75th pct | 95th pct | Std dev |
|---|---|---|---|---|---|---|
| 1Y | -6.09 | 2.9 | 11.49 | 24.88 | 50.17 | 17.42 |
| 3Y | 4.71 | 11.38 | 14.46 | 17.46 | 22.69 | 5.31 |
| 5Y | 6.59 | 11.15 | 14.02 | 16.83 | 21.18 | 4.43 |
| 7Y | 9.78 | 12.22 | 13.56 | 14.84 | 16.74 | 2.07 |
| 10Y | 11.13 | 12.82 | 13.98 | 15.13 | 16.79 | 1.73 |
The median stays between 11.49% and 14.46% across horizons. The standard deviation narrows from 17.42 at 1Y to 1.73 at 10Y, so the spread of outcomes tightens as the holding period lengthens in the observed windows.
Extremes by Horizon
| Horizon | Weakest window (CAGR) | Scheme and dates | Strongest window (CAGR) |
|---|---|---|---|
| 1Y | -37.19% | HDFC Large Cap Fund, 2019-03-23 to 2020-03-23 | +98.07% (SBI Large Cap Fund, 2020-03-23 to 2021-03-23) |
| 3Y | -8.17% | Taurus Large Cap Fund, 2017-03-23 to 2020-03-23 | +35.43% (Nippon India Large Cap Fund, 2020-03-23 to 2023-03-23) |
| 5Y | -3.43% | Taurus Large Cap Fund, 2015-03-23 to 2020-03-23 | +31.58% (Nippon India Large Cap Fund, 2020-03-23 to 2025-03-23) |
| 7Y | +4.21% | Taurus Large Cap Fund, 2013-03-23 to 2020-03-23 | +20.31% (Mirae Asset Large Cap Fund, 2014-02-15 to 2021-02-15) |
| 10Y | +8.03% | Taurus Large Cap Fund, 2015-03-03 to 2025-03-03 | +19.24% (Mirae Asset Large Cap Fund, 2013-08-28 to 2023-08-28) |
The weakest 1Y, 3Y, 5Y and 7Y windows all end on 2020-03-23. The data shows this clustering of end dates but cannot show why it occurs.
Where the Residual Loss Sits
Start-year breakdowns show how concentrated the non-zero loss figures are.
3Y. Loss probability is 0.0% for every start year except 2017, where it is 16.61% (8,258 windows that start year).
5Y. Loss probability is 0.0% for every start year except 2015, where it is 1.55% (7,665 windows that start year). The worst 5Y outcome is -3.43%.
1Y. Loss probability varies widely by start year. It is 40.22% for 2015 starts and 45.18% for 2019 starts. It is 0.11% for 2016 starts and 0.0% for 2020 and 2023 starts. A single 1Y loss figure of 16.44% therefore pools very different start-year experiences.
7Y and 10Y. Every start year in the data shows 0.0% loss. 7Y windows start from 2013 to 2019 (2019 has 6,499 windows). 10Y windows start from 2013 to 2016 (2016 has 5,983 windows).
Interpretation
- Sequence of horizons. Loss probability declines at each tested step: 16.44%, 1.47%, 0.16%, 0.0%, 0.0%.
- Tail narrowing. The 5th percentile moves from -6.09% at 1Y to 9.78% at 7Y and 11.13% at 10Y. The 7Y weakest outcome of +4.21% is above zero, but the margin is modest compared with the median of 13.56%.
- Inflation context. Against the 6.0% threshold, the share of windows beating it rises from 66.02% at 1Y to 99.89% at 7Y and 100.0% at 10Y.
- Dependence on start dates. The 5Y and 3Y residual losses concentrate in specific start years (2015 and 2017). Different start dates outside the data window could produce different results.
Limitations
- Survivorship bias. Only currently active schemes are included. Closed or merged schemes are excluded.
- Short history. The data floor is 2013-01-01. The 7Y windows start between 2013-01-01 and 2019-10-01, and the 10Y windows between 2013-01-01 and 2016-10-01. Neighbouring windows overlap heavily, so the 7Y and 10Y results rest on a limited set of start dates.
- Pooling. Windows are pooled across schemes and are not independent observations. The scheme count differs by horizon (33 at 1Y down to 22 at 10Y).
- Scope. Results are lump-sum and point-to-point. They say nothing about SIP outcomes, taxes, or other holding periods.
- No forecast. A 0.0% observed loss probability describes the sample only. It does not show that a loss cannot occur in the future.
Disclaimer
This is educational content, not investment advice; consult a SEBI-registered investment adviser. Historical distributions do not predict future returns.
Data Integrity Statement
The figures in this study are drawn from the FundSageAI Quant Snapshot generated on 2026-10-05, based on daily NAV history for Direct Growth plans of currently active Large Cap schemes. The data is internally validated. Universe, plan selection, data floor and survivorship-bias limits are stated in the Methodology section.