Nightjar NO EDGE

Passes on Nifty, fails the Sensex replication: not paper-traded

A pre-registered finalist (sell within two sessions of expiry at 15:15, buy back next midday) made +₹871 a night at 2x costs on 50 sealed Nifty nights and +₹1,035 over all 196 nights, but on Sensex it was +₹368 then -₹331 (about zero overall). Rule set beforehand: both halves of Sensex positive. It failed, so no paper run.

• 1. The report: a Tradetron back-test of a Sensex OTM1 strangle sold and bought back the next day (6 sets, 2024-09 to 2026-09, ₹7.5 lakh): gross +₹8.53 lakh (+114%), Sharpe 3.4. It is gross: its own cost lab shows ₹4.96 lakh of costs (58% of gross). Its own exit table shows planned exits +₹44.9 lakh and 'repair' exits -₹36.3 lakh. The call leg made ₹7.6 lakh and the put leg ₹0.2 lakh. Real live fills were about 1.5% worse than the minute open (3.5% in the first 15 minutes).
• 2. This study: Nifty, 464 sessions (2024-12-03 to 2026-10-16), Dhan 1-minute weekly options, strangles with 8 entry times, 8 exits, 15 strike structures, 5 stops, 3 targets, 3 defined-risk structures and 29 filters (287,000 configurations). Slippage 1.5% of the premium per fill (3.5% in the first 15 minutes), STT 0.10% then 0.15% from April 2026, lot 75 then 65. About half of overnight exits use a Black-76 model price because the strike leaves Dhan's -2..+2 window.
• 3. Gross edge is real, net edge is thin: a plain OTM1 strangle every non-expiry night earns ₹637 to ₹965 a night per lot pair before costs on Nifty (₹726 to ₹949 on Sensex), +₹107 to +₹441 at 1x costs, and -₹422 to -₹84 at 2x.
• 4. The search does not hold up: at 2x costs 38% of Nifty configurations are positive in development but only 5.7% in both development and validation (1.1% at 3x); on Sensex 5.9% and 0.2%. The 40 best by development lost ₹1,634 (Nifty) and ₹1,160 (Sensex) a night in validation. The best development t-statistic (8.7) is far beyond luck, which fits a persistent regime, not an edge that carries over.
• 5. Direction explains the winners: put-only configurations won while the market rose (development) and lost in validation; call-only configurations did the reverse. Defined-risk wings lose ₹2,000 to ₹2,600 a night; volatility timing (implied over realised) is positive in development and negative in validation; stops and targets did not help.
• 6. The finalist (1-2 sessions to expiry; ATM call and the put 100 points lower; 15:15 to next 12:00; no stop): development +₹1,080 (99 nights), validation +₹1,114 (47), sealed +₹871 (50) per night at 2x; +₹461 sealed at 3x; worst night -₹41,185 (2 February 2026, 40 average winning nights); maximum drawdown for one lot ₹52,368 (3.5% of ₹15 lakh). It is short a call: it loses ₹3,327 for each 1% the market rises overnight, with an intercept of +₹984 a night (t 2.9).
• 7. Weak points: 60% of the finalist's exits use the model price; with model prices 10% worse for the seller the sealed result at 2x is +₹448 and at 20% worse +₹25. On Sensex it failed (first half +₹368, second half -₹331 at 2x). The tail rule allows only one lot.
• 8. Decision: no paper run (Sensex replication failed). A paper run that records real contract prices would settle the model-price question; the script is written but not scheduled. Dhan's data token must be valid each evening for it to run.