Risk, allocation, and the benchmark that decides the verdict
14 October 2026 · 131 traded positions, 13 April to 13 October, marked to market daily · realised and open, from the live book
The headline depends entirely on which index you hold it against, and the two answers point opposite ways. Against the Nifty 50 this book returned +9.38% while the index fell 4.70%, for an annualised alpha of +16.19%. Against the NIFTY Smallcap 100 — which is where 41% of the equity exposure actually sat — the index returned +16.02% and the book underperformed it by 6.6 points, with alpha collapsing to +3.04% and the information ratio turning negative at −0.91. Both statements describe the same six months.
The same book against three benchmarks
| measure | Nifty 50 | Midcap 150 | Smallcap 100 |
|---|---|---|---|
| benchmark return | −4.70% | +3.08% | +16.02% |
| book return | +9.38% | +9.38% | +9.38% |
| beta | 0.257 | 0.439 | 0.372 |
| alpha, annualised | +16.19% | +11.97% | +3.04% |
| correlation | 0.270 | 0.534 | 0.511 |
| r² | 0.073 | 0.285 | 0.261 |
| tracking error | 14.02% | 12.27% | 13.77% |
| information ratio | +1.977 | +0.950 | −0.905 |
| benchmark volatility | 11.88% | 13.74% | 15.53% |
Everything that does not reference a benchmark is the same in all three columns, because it is a property of the book alone: total return +9.38%, annualised 19.80%, standard deviation 11.30%, downside deviation 8.59%, maximum drawdown 8.24%, Sharpe 1.18, Sortino 1.55, Calmar 2.40. The risk-free rate is 6.50% and is an input, not a measurement; every Sharpe, Sortino and alpha above moves with it.
Why the Smallcap column is the honest one
A benchmark is a claim about what the book could have done instead. The Nifty 50 is only the right claim if large caps were the alternative, and they were not: nine positions out of 116 were large caps and they carried a tenth of the exposure. The allocation says plainly where the money went.
| tier | positions | share of positions | deployed | share of exposure | realised | win rate |
|---|---|---|---|---|---|---|
| large | 9 | 7.8% | ₹17.8L | 10.0% | +11,291 | 33.3% |
| mid | 19 | 16.4% | ₹33.8L | 16.0% | +72,969 | 36.8% |
| small | 38 | 32.8% | ₹64.3L | 40.9% | +367,435 | 52.6% |
| micro | 20 | 17.2% | ₹39.9L | 19.9% | +180,682 | 55.0% |
| nano / unlisted | 30 | 25.9% | ₹47.6L | 13.2% | −152,610 | 16.7% |
| total, equity book | 116 | 100% | ₹2.03Cr | 100% | +479,766 | — |
Tiers are NSE’s own constituent lists — Nifty 100, Midcap 150, Smallcap 250, Microcap 250 — not a threshold invented here. A company in none of them is beyond roughly the 750th by market capitalisation and is shown as nano / unlisted rather than guessed at. Membership is current while the trades are historical; these lists rebalance twice a year and a name that changed tier inside the window is counted in its present tier throughout.
Small and micro together are 61% of exposure and produced ₹5.48L of the ₹4.80L realised — more than all of it, because the nano tier lost money. That is why the Smallcap 100 is the benchmark this book has to answer to.
The book was 11% invested
Across all exposure including funds, the equity book was 11.0% and ETFs were 89.0%, almost entirely idle cash swept into the parking instrument. Fifteen fund positions carried ₹2.60Cr of deployment and lost ₹58,462.
This is the same fact as the low beta, seen twice. A book that holds cash most of the time cannot move with an index, and cash has a beta of zero. The 0.257 against the Nifty is not a hedging achievement; it is a measure of how rarely the gate let anything through. It also means none of the three comparisons above is like-for-like in either direction: a book that is 11% invested should not be expected to match a fully invested index, and the alpha it does show is earned on a small fraction of the capital.
The tier that lost money
Nano / unlisted took 26% of all positions and 13% of exposure, won 16.7% of the time, and lost ₹1.53L. Every other tier is profitable, and small and micro both win more than half the time. It is the clearest tier-level signal in the six months, and it is a selection result rather than a market one: these are companies outside the largest 750, where the scanner’s relative-strength and earnings inputs are thinnest and the daily ranges are widest.
This sits beside a separate observation from the same week: the gate set contains no trend filter at all. G1 and G2 cap extension above a moving average to keep stops sizeable, and both pass a stock trading below its averages — the gate text says so in words, “BELOW the 10MA, which this gate does not refuse”. Nothing in the seven gates references the 50-day average.
What these numbers are not
- Not a verified statement. This is the service’s own bookkeeping, marked against NSE closes. The account’s broker statement is the only figure independent of this reconstruction.
- The risk-free rate is supplied, not observed. 6.50% annual. Sharpe, Sortino and every alpha are conditional on it.
- r² is low in all three columns — 0.07, 0.29, 0.26. The benchmarks explain between 7% and 29% of this book’s movement, so alpha and beta are weakly determined and should be read as descriptive rather than as a decomposition.
- 126 sessions is a small sample. These are descriptive statistics, not estimates with tight intervals.
- Ten position-sessions had no close and were carried at their previous mark out of 1,807. Volatility is understated to that extent.
- Cancelled triggers are excluded. 26 of them, inside the window. They bought nothing, so they carry no risk and no return; counting them would dilute every per-trade figure with rows that never happened. The commonly quoted 157 includes them.
Two errors made producing this, and corrected
Both are recorded because a number that was wrong and plausible is worth more as a warning than as a deleted draft.
- The first equity curve was 43% stale. The price fetch was passed a count of sessions where it wanted calendar days, returning 94 trading days of closes for a 126-session window. A curve that does not move cannot correlate with one that does, so beta came out at 0.082 and r² at 0.007 — artefacts of the construction, and precisely the failure the command was written to avoid.
- The first allocation reported “unlisted-tier: 90.4% of exposure”. It was arithmetically correct and read as the exact opposite of the truth: LIQUIDCASE, NIFTYBEES, GOLDBEES and ten other funds are absent from equity constituent lists because they are not companies, and the largest is where idle cash is parked by design.
The number to quote
For 13 April to 13 October 2026, over 131 traded positions on a book that was 11% invested: +9.38% returned against −4.70% for the Nifty 50 and +16.02% for the Smallcap 100, at a standard deviation of 11.30%, an 8.24% maximum drawdown and a Sharpe of 1.18. It beat the index it does not resemble and lost to the one it does.