SMA-distance trend study NO EDGE

Weak, mostly inconclusive: no threshold is a reliable trend signal

Best-supported: daily close 2-4% above the 10/20-SMA (sustained uptrend 29%-30% vs 20% baseline), but late and only weakly significant; downward signals mark rebounds.

• Data: Nifty 50 daily OHLC, NSE official end-of-day index files from 2013-01-01 and Yahoo Finance ^NSEI for 2007-09-17 to 2012-12-31; the sources agree to within 0.0005% on 3,382 overlapping days. 4,687 daily, 997 weekly and 229 monthly completed candles; null closes dropped, nothing interpolated.
• Daily upward signals: a first close ≥3% above the SMA was followed by a sustained uptrend (3% net move over 20 sessions, 70% persistence, no close back below the signal-time SMA) 29% of the time for the 10-SMA and 30% for the 20-SMA, against 20% for every day already above its average and 14% for every day. The improvement is positive in both halves of the history but its lower confidence bound is only about 6 points.
• It arrives late: before the median ≥3% daily up-signal the index had already risen about 4% in 10 sessions; the next 10 sessions add a median of about 1%. About 29% of 10-SMA signals and 12% of 20-SMA signals close back across the average within 5 sessions.
• Downward signals do not signal downtrends: the median 10-session return after a ≥3% close below the 10 or 20-SMA was +0.5% and +1.2% (rebounds), and only 18%-21% became sustained downtrends. Short trades after the signal mostly lost money after 0.10% costs.
• Larger distances mostly identify stretched moves: the sustained-trend rate rises only a few points from 1% to 3-4%, the run-up completed before the signal grows steadily, and forward returns do not rise with the threshold (the 5% cells rest on 13-30 independent episodes).
• Statistics: 300 tests of success rate against the unconditional baseline, exact binomial, Benjamini-Hochberg at 10%: 1 survives (next-day continuation after the first daily close ≥1% above the 10-day SMA: 64% vs 53%); 0 against the like-for-like baseline. Weekly results are mostly indistinguishable from baseline; monthly combinations have 6-22 signals and no conclusion is supported.
• Not trading advice and not a strategy: the executable view (next-open entry, 0.10% round trip, no leverage) shows long trades after up-signals earning roughly what a long trade started on any day earned. The report has the heatmaps, comparison charts, event studies, return distributions, annotated examples and the full tables.