Meridian PAPER

Imagine that every month you look at the 500 biggest listed Indian companies and ask: which ones have been climbing the most steadily over the last six months? Meridian buys up to 15 of those 'steady climbers', puts at most 4 out of every 100 rupees into any one of them, and keeps part of the money in cash. If the overall market looks weak it holds more cash; if the market looks strong it invests more. If a share falls too far from its recent high, it is sold automatically, so one bad share cannot do much damage. It will often be wrong on individual shares (about 6 in 10 trades lose a little), but the winners can run a long way, and that is where the profit comes from.

Each month-end Meridian ranks stocks that were in the Nifty 500 on that date by risk-adjusted momentum: the price change from six months ago to one week ago, divided by how bumpy the stock has been. It buys the top 15 next morning at the open, up to 4% of equity each. How much it may invest depends on the Nifty 500 trend: up to 80% in a strong trend, 60% in a mixed one, nothing in a weak one. Every position has a stop that starts 2.5 ATRs (typical daily ranges) below the purchase price and trails 5 ATRs below the highest close. Positions are sold at the stop, after 364 days, or when trimmed. Costs and slippage are deducted in every number you see.

Signal: score = [CF(t-7d)/CF(t-182d) - 1] / (weekly return stdev x sqrt(52)), where CF is a split-adjusted close; at least 20 weekly observations and a 20-day average traded value of ₹1 crore are required. Membership is point-in-time using effective-dated Nifty 500 reconstitutions. Regime is the weekly Nifty 500 close relative to its 40-week EMA and SMA. Orders fill at the next open with slippage of 0.15% plus 5% of participation; stops are evaluated intraday and fill at min(open, stop). Winners above 12% of equity are trimmed to 8%. A 18% drawdown freezes new buys, 28% liquidates and halts for 126 days. The honest reading is in the gates: ranking beats a random-rank null by about 5 points a year, but 2016-2020 shows no edge and the top 10 trades give 63% of profit. Hold-out result: -5.2% vs -2.9% (index), maxDD -9.7%.

Treat this as a long-only, low-turnover trend/momentum sleeve with an exposure overlay, not as alpha in isolation: beta to the Nifty 500 is 0.25 at an average 37% invested, monthly alpha 1.0% (HAC t = 2.74, n = 116) with heavy-tailed payoffs (win rate 37%, 417 of 464 exits are stops, top-10 trades 63% of P&L). Fragilities to price in: (1) data coverage of true point-in-time members is only 68% in 2016 rising to 94% in 2026, so early years are biased upward by missing delisted losers; (2) split detection is ratio-based without a corporate-action table; (3) benchmark and strategy are price-only; (4) median hold of 45 days makes most gains short-term for tax; (5) capacity is not an issue at ₹5 lakh but the ADV filter and the 10% participation cap bind above roughly ₹2-3 crore. Audit path: trades, equity and the scenario grid are published as CSV; the sealed hold-out and the forward run are the only out-of-sample evidence, and the specification is frozen. Update 2026-10-18: the sealed hold-out (2025-10 to 2026-10) returned -5.2% against -2.9% for the index with a 9.7% drawdown (index 14.8%), a 30% win rate and negative gross P&L; with 67 trades the confidence interval is wide, but there is no out-of-sample support so far. No real capital is justified until the forward run adds evidence.