Quickfuse NO EDGE

Closed: no edge

After 186,480 tested combinations, no setup is profitable and robust at 2x costs with drawdown under 20%. The specification as written loses ₹268 a trade (t -4.9), no better than random entries (-₹240).

• Specification as supplied: 863 trades on 328 sessions (2024-12-03 to 2026-03-30, development window), win rate 23.8%, expectancy -₹267.6 a trade, total -₹2.31 lakh, profit factor 0.66, maximum drawdown about -₹2.34 lakh. Morning -₹170 a trade, afternoon -₹418. Random entries with the same exits and costs: mean -₹240 (95th percentile -₹155); the strategy beats 29% of them.
• Multi-combination search (owner request, capital ₹5,00,000, drawdown limit 20% = ₹1,00,000): 186,480 combinations = 67 signal variants in six families (pullback, opening-range breakout, momentum burst, VWAP reversion, trend continuation, narrow-CPR) x 5 time windows x 3 strike choices (ITM1, ATM, OTM1) x 240 exit settings, each priced at 1x, 2x and 3x slippage and charges. Ranked on 2024-12 to 2025-09 only; 2025-10 to 2026-03 was kept aside for checking.
• Selection-bias check: 1.6% of combinations were profitable on the ranking period at 2x with drawdown under 20%. The top 60 by that ranking made +₹57,700 on average there and -₹40,000 on the checking period; only 5% stayed positive. White's reality check across those 60: p = 0.57 (best t-statistic 0.86). Nothing is significant.
• Only the opening-range-breakout family beats a random-entry calibration: profitable on both halves at 1x in 16% of its combinations against 0.2% for random entries (2,400 random combinations), and at 2x in 591 combinations against none. Its best cell made +₹128,000 at 1x (drawdown 5.7%), +₹60,000 at 2x (10.4%) and -₹8,000 at 3x (17.7%) over 16 months, 176 trades, 10 of 16 months positive. That was chosen in hindsight, so it overstates the future.
• Walk-forward test of 'optimise, then trade' on that family (best configuration chosen on earlier data only, traded on the next quarter, three quarters): +₹30,800 at 1x (drawdown 15%, t 0.57), -₹18,800 at 2x (21%), -₹68,500 at 3x (27%). The chosen strike and parameters changed from quarter to quarter, which is a sign of noise rather than a stable edge.
• Round 2 (69,120 more combinations, same family, holds of up to 150 minutes to dilute costs): worse. Walk-forward -₹43,000 at 1x, -₹98,000 at 2x, -₹153,000 at 3x.
• Reading: before costs the ORB idea may carry about half a point of edge per trade, which is smaller than the friction. Profit at standard costs is possible in sample and disappears at 2x. The hold-out (2026-04-01 to 2026-10-16) stays sealed; opening it would only test a configuration that did not pass the development checks.
• Short side (added 2026-10-20). Quickfuse was never long-only: long signals bought calls and short signals bought puts. Over 863 trades the calls lost ₹262 a trade (433 trades, win rate 22.4%) and the puts lost ₹274 (430 trades, 25.1%); gross before friction was about -1.2 points on each side, so mirroring the signals would earn about +1 point before roughly 1.5 points of friction. The untested version was option SELLING, so it was run on the same 1-minute data (5 lakh capital, 1% of capital at risk, margin cap of 2 lots, 1x/2x/3x costs, development window only): 35,200 single-leg combinations (each of the 67 signal variants selling the option that profits from the signal's direction, at-the-money or one strike out, 5 windows, 64 stop/target settings) and 27 intraday short-straddle settings (sell the ATM call and put at 09:30, 09:45 or 13:45). Result: at 1x costs 1.1% of single-leg combinations were profitable on both halves; at 2x only 22 (0.1%) and none within a 20% drawdown; at 3x two. The 40 best on the first half made +₹21,000 on average there and -₹37,700 on the second half; none stayed positive. Short straddles lost in both halves at every cost level (median -₹54,000 to -₹84,000 on the first half at 2x), with drawdowns up to ₹1.0 lakh. Selling does not rescue the idea: the seller pays the same spread and charges, and loses when moves are larger than the premium collected.
• Capital re-test at ₹15 lakh (2026-10-20; drawdown limit ₹3 lakh, risk per trade scaled, margin cap 6 lots single-leg and 3 lots straddle). Buying grid (186,480 combinations): positive on both halves 6.8% at 1x, 1.0% at 2x, 0.07% (134) at 3x; 25 survive 3x within the drawdown limit, all one opening-range-breakout cluster of 91 and 65 trades earning ₹10,000 to ₹26,000 over the whole window; the 40 best on the first half made +₹2.35 lakh there and -₹0.78 lakh on the second half (8% stayed positive). Walk-forward re-selection of that family: +₹79,800 at 1x, -₹57,400 at 2x, -₹194,600 at 3x. Selling grid (35,227): 7 survive 2x within the limit, none at 3x; top-40 first-half picks +₹0.90 lakh then -₹0.95 lakh. Bigger capital dilutes the fixed ₹20 brokerage per order (more combinations clear 1x and 2x) but not the spread, so the conclusion is unchanged.
• Variants of the written specification: all five conditions required 51 trades, +₹122 a trade (t 0.41); pullback mandatory 149 trades, -₹85 (t -0.59); slippage 1.0 point -₹416; slippage 1.5 -₹565; targets of 15 or 30 points still negative. The paper run was stopped before its first session at the owner's instruction.