Verdict
Rejected
Higher headline return, but better than the standard version in only 4 of 10 years with no improvement in drawdown.
- Back-test CAGR 20.8% vs 16.0% for v4; maximum drawdown -19.6% vs -18.6%.
- Gain is concentrated in 2021 (+93%) and 2023 (+61%); worse than v4 in 2016, 2018, 2020, 2022, 2024 and 2025.
- Fails the reference design's own acceptance rule (higher median return AND no worse drawdown).
- Random-rank null averaged 11.7%; real ranking beat all 30.
Promotion gates
20.8% vs 16.0%.
-19.6% vs -18.6%.
4 of 10 years.
30 of 30.
Explain the strategy
Meridian Adaptive is Meridian with a smarter 'weather forecast'. Besides the market trend it also looks at how many shares are rising and how jumpy the market is, then decides how much money to invest and in how many shares. In good weather it invests more and in more shares; in bad weather it holds cash. It produced a bigger headline return, but only beat the simpler Meridian in 4 of 10 years and did not reduce the worst fall, so it was rejected.
Report
Robustness and scenarios
| Scenario | CAGR | Max drawdown | Sharpe | Index CAGR | Trades |
|---|---|---|---|---|---|
| Base case | 20.8% | -19.6% | 1.60 | 13.3% | 687 |
| Costs x2 | 20.0% | -19.3% | 1.53 | 13.3% | 690 |
| Start +1 month | 21.4% | -18.8% | 1.62 | 14.1% | 693 |
| Start +3 months | 21.6% | -18.8% | 1.62 | 14.2% | 694 |
| Start +5 months | 21.6% | -18.7% | 1.61 | 13.8% | 679 |
| From 2018 (better coverage) | 17.5% | -19.1% | 1.37 | 12.0% | 562 |
| From 2021 (coverage >= 80%) | 29.8% | -15.5% | 1.88 | 15.3% | 401 |
| Top 10 names | 20.8% | -19.6% | 1.60 | 13.3% | 687 |
| Top 20 names | 20.8% | -19.6% | 1.60 | 13.3% | 687 |
| Lookback 252d | 21.6% | -18.0% | 1.58 | 13.3% | 737 |
| Skip 21d | 18.3% | -18.7% | 1.44 | 13.3% | 712 |
| Slippage 0.5% per side | 17.9% | -19.9% | 1.40 | 13.3% | 703 |
| Slippage 1% per side | 16.7% | -20.1% | 1.29 | 13.3% | 688 |
| Slippage 2% per side | 1.5% | -20.9% | 0.20 | 13.3% | 536 |
Random-rank null (30 seeds)
Same stops, regime rules and costs, but stocks picked at random. Mean 11.7% (sd 2.1%, max 14.8%) vs the real ranking at 20.8%.
Alpha versus the index
Monthly alpha 1.29% (15.5% a year, HAC t = 2.88), beta 0.32.
Rules and parameters
Rules
- Universe: stocks that were Nifty 500 members on each date (point-in-time), average daily traded value at least ₹1 crore, quality label not EXCLUDE.
- Signal (month-end close): return from 182 to 7 calendar days ago divided by the stock's weekly volatility over the same window (risk-adjusted momentum).
- Regime (weekly Nifty 500): STRONG = close above 40-week EMA and SMA; MODERATE = above SMA only; WEAK = below. Invested ceiling 80% / 60% / 0%.
- Entry: buy the top 15 ranked names the next trading day at the open, at most 4% of equity each, only while invested value is under the regime ceiling.
- Exit: ATR(14) stop at 2.5× below entry, trailed at 5× ATR below the highest close; forced exit after 364 days; 30-day cooldown after any exit; winners above 12% of equity are trimmed to 8%.
- Portfolio risk: no new buys when drawdown ≥18% (up to 90 days); liquidate and halt for 126 days at ≥28% drawdown.
- Cash earns an estimated RBI-repo-based rate. Costs: brokerage, STT, exchange, SEBI, stamp, DP and GST by era; slippage 0.15% plus market impact.
Parameters
- capital
- ₹5,00,000
- lookback / skip
- 182d / 7d
- top_n
- 15
- max per name
- 4%
- ceiling
- 80%
- ATR window
- 14
- initial stop
- 2.5×ATR
- trailing stop
- 5.0×ATR
- max hold
- 364d
- cooldown
- 30d
- base slippage
- 0.15%
- min ADV
- ₹1 crore
Instrument
NSE cash equities (Nifty 500 point-in-time) · capital ₹500,000
Known limitations
- Past members that left the index and have no price history are missing, which flatters 2016–2020 especially.
- Split and bonus adjustment uses a price-ratio check, not a corporate-action table.
- Benchmark is a price index (no dividends); the strategy is price-only too.
- Tax on short-term gains (median hold 45 days) is not deducted.
- Estimated repo-rate cash yield and cost schedule; verify against contract notes.
Logs and raw data
Downloads appear once a data window is loaded.
Files are plain CSV so results can be re-computed independently. Trade dates are IST sessions; returns are net of modelled costs and slippage.