PE Valuation Zones on Nifty: Eight Variants Tested, Eight Beat the SIP, Coarser Zones Led

Lineage: The first valuation signal in the series rather than a price signal, and the first to vary how graded the response to that signal is.

The premise

Every study before this one timed deployment on price, which says nothing about what the index is worth. The PE ratio does, at least at extremes, which makes it the first signal in this series with a defensible reason to work. Whether that survives conversion into a deployment rule is the question here. Both strategies read the index PE against its own rolling average, and differ only in how many states the reading may produce.

How the strategies work

Strategy 10 sorts each month into one of three zones. PE below 0.85 times its moving average makes the month cheap and deploys the entire bank. PE at or below the average without clearing that discount makes it fair and deploys only the current contribution, which is a plain SIP. Anything higher is expensive and deploys nothing. The 0.85 discount is the whole strategy, and the sweep varies only the lookback of the average, across 3, 6, 12, 24 and 36 months.

Strategy 11 divides PE by that same average and sorts the ratio into four zones. A reading under 0.80 deploys the whole bank. Under 1.00 the rule deploys the contribution plus half the bank, which is the graded middle response a two-state rule cannot express. Under 1.20 the contribution alone deploys, and above that nothing does. Its sweep varies the window across 24, 36 and 48 months, giving eight variants between the two strategies. The ratio is formed in monthly space before the series is forward-filled, at Strategy 11.py:223, so a live PE is never paired with a stale average.

The setup

Both share the bank framework of the Nifty 50 era: ₹10,000 credited on the first trading day of each month, the undeployed balance compounding at 5% annualised, a 0.03% cost, and every unit held permanently. Nothing is sold, so no gain is realised and no capital gains tax applies. The benchmark is a capital-matched monthly SIP on the same schedule and cost: 11.9907% XIRR and ₹3.2641 crore on ₹36.70 lakh contributed across 367 months, a MOIC of 8.8941x.

One code-level caveat affects every Strategy 10 figure: the opening month's deployment amount is computed and never spent, because Strategy 10.py:437 sets it inside the i == 0 branch while the spend block at line 482 sits in the sibling branch.

The best combinations

Eight of eight beat the benchmark, the only unanimous result in this series and the least interesting thing in the table. The gap between the two families is the finding. Strategy 10's best variant finished ₹3.3668 crore for a MOIC of 9.174x against Strategy 11's 8.9777x, and Strategy 11's whole winning margin of ₹3.07 lakh is smaller than the one Strategy 10's fourth-placed variant produced.

Trade counts explain why. The three-zone rule deployed in 207 of 367 months against the four-zone rule's 326, which finished with an empty bank. A rule that almost always deploys is a SIP wearing a valuation label.

What the key parameters do

Zone counts from the workbooks make the mechanism explicit. Strategy 10's 36-month variant spent 53 months in the cheap zone, 155 in the fair zone and 159 in the expensive zone. Intended deployments therefore number 208, being 53 plus 155, against 207 executed, one short, which is the unspent opening contribution. Those 159 expensive months are the strategy: for 159 months of a thirty-year bull market it bought nothing, and accumulated the bank the 53 cheap months then spent in full.

Strategy 11's 48-month variant distributes very differently: 86 months warming up before its rolling average existed, 31 in the very cheap zone, 98 in the graded cheap zone, 111 in the fair zone and only 41 in the expensive one. The long window therefore costs more than seven years in which the rule is a SIP by default, and its expensive gate at 1.20 then lets it keep contributing through months the three-zone rule would sit out entirely.

Lookback length matters less than either, and for Strategy 10 it does not move in a straight line. The 36-month average ranks first and the 6-month second, ahead of both the 12 and 24-month settings, and only the 3-month average falls away sharply at ₹3.00 lakh. An average that short tracks PE too closely for the discount to open often, which shows up as the fewest trades, 185, and the largest idle bank, ₹4.29 lakh.

Where it fails

Neither family fails outright, which is a reason for caution: a sweep where every variant wins has not been stressed, and eight variants across two related rules is a small grid by the standards of this series. Strategy 10's edge rests on few decisions, its best variant making 53 full-bank commitments in thirty years, and the 0.85 discount was chosen rather than derived.

Strategy 11 fails in the narrower sense of failing to improve on the simpler rule. Its graded middle zone was the reason for building it, and 98 months landed there deploying half a bank each time. Committing half at a modest discount leaves the rest earning 5% while the index compounds near 12%, and leaves less to commit when a real discount arrives. Sophistication was added at the point where the edge came from decisiveness.

Verdict

The same valuation reading produced an edge worth ₹10.27 lakh under a rule with three states and ₹3.07 lakh under a rule with four, on the same window, costs and bank. Every one of the eight beat the SIP, so the signal is doing something real on this evidence, and the useful conclusion concerns the response rather than the signal. A valuation rule earns its keep by sitting out and then committing in full. Grading that response toward the middle produced a strategy which traded in 326 of 367 months and finished closer to the benchmark it was trying to beat. On this grid, run the coarser rule with a long lookback, and read the graded version as a measurement of what the finer one costs.

Appendix: assumptions and methodology

• Data. Nifty 50 daily open, high, low, close and PE from nifty50_research_master.parquet, 367 months, index level only.
• Benchmark. Monthly SIP, ₹10,000 on the first trading day of each month, 0.03% cost, no bank: 11.9907% XIRR, ₹3.2641 crore final, MOIC 8.8941x on ₹36.70 lakh contributed.
• Bank and tax. ₹10,000 credited monthly, compounding at 5% annualised until a zone releases it. No tax is modelled and none applies, because units are held permanently.
• Zones. Strategy 10: cheap below 0.85 times the rolling average, whole bank; fair at or below the average, one contribution; expensive otherwise, nothing. Lookbacks of 3, 6, 12, 24 and 36 months. Strategy 11: PE over its rolling average, formed in monthly space before the forward fill at Strategy 11.py:223, deploying the whole bank under 0.80, one contribution plus half the bank under 1.00, one contribution under 1.20, and nothing above. Windows of 24, 36 and 48 months, with months before the average exists treated as warm-up that contributes rather than waits.
• Metrics. XIRR, alpha, additional wealth, trades and idle bank. Sharpe reads 0.8274 and Sortino 1.6137 for Strategy 10's best variant, against 0.7788 and 1.5986 for Strategy 11's.
• Zone counts. From the Zone column of the STRATEGY_36M and STRATEGY_48 sheets of the two published workbooks, over the 367 month-start rows each populates. They are not in the SUMMARY sheets.
• Strategy 10's unspent first contribution. It costs one contribution's early deployment per variant, appears as 207 executed trades against 208 intended, and flips no variant's status.
• Not modelled or tested. Year-by-year returns, split-period figures, maximum drawdown, Calmar, brokerage beyond 0.03%, slippage, dividend reinvestment, a Nifty 500 version, any discount other than 0.85, any out-of-sample split, and whether the PE series is comparable across a window spanning changes in index composition and earnings-reporting convention.
• Withdrawn. The published version carried a third implementation, a continuous percentile dial numbered Strategy 16, reported as losing all 48 of its variants. No script or workbook for it survives anywhere in the tree, which holds Strategies 1 to 14 and nothing between 15 and 19, so every figure attributed to it was unverifiable. It was removed on 2026-09-24 rather than republished on numbers that cannot be checked.
• Withheld. Nothing. All eight variants, both zone structures and every threshold appear above, because the conclusion concerns the shape of the rule rather than any setting of it.