Log in
Account
Account
Tools

Study · · By Shashank

Does holding stocks from the start of an uptrend pay?

Verdict

RejectedBacktest, hypothetical, costs included

The best version made 19.2% a year against 28.1% for Birch on the same stocks. With an agreement score of 80 or more, it found only about 340 trades in 10 years.

On this page

The short answer

Is it worth holding a stock from the day our cycle models agree its uptrend has begun? No. Every version we tried trailed Birch on the same stocks and dates. Demanding stronger agreement picked slightly better trades, but far too few of them.

Our market-cycle page asks several research models whether each stock's uptrend has started. When most of them agree, we call it the start of an uptrend. How strongly they agree gives a score from 0 to 100.

So we held a stock from the day the models agree its uptrend has begun, until the agreement ends. We asked three things. Does it pay? Does a higher score pick better stocks? And which way of closing positions, and how wide a spread of the money, works best? The version listed on the site is Willow.

How did we test it?

  • The idea: take stocks from our usual list of most-traded NSE stocks whose uptrend has just been confirmed, at the next day's open. When there is little room, the highest scores go first.
  • Four score levels, from "any agreement" to only the strongest.
  • Four ways of closing a position:
    • when the agreement ends, with no stop-loss;
    • the same, with a stop-loss;
    • a trailing stop and nothing else;
    • when the price falls back below a medium-term average.
  • Five ways to spread the money, from 30 to 100 slots with the same money in each, and the shared risk-based sizing.
  • 80 combinations in all, from 2016 to October 2026. Every run starts with ₹10 lakh and includes Zerodha delivery charges, 0.1% slippage per side, and no trades on days a stock is locked at its circuit limit.
  • A luck test: for the three best runs, 200 repeats with the same rules, but with each day's picks drawn at random from stocks already in an uptrend that day.

What did we find?

Rejected. No combination came close to Birch on the same stocks and dates.

Portfolio2016–2021, per year2022–2026, per yearWorst fall
Birch31.8%23.4%27%
Best version: any score, close when agreement ends, 30 slots25.1%11.9%39%
Best with a score filter, plus a stop-loss, 50 slots21.4%11.2%28%
The same stocks, same money in each, no costs18.6%12.4%59% (2016–2021)
NIFTY 50016.1%7.5%38% (2016–2021)
  • A higher score picked slightly better trades, but far too few. The share of winning trades rose from 38% at any score to 43–44% at the higher levels. At the strictest level there were only about 340 trades in 10 years. Most of the money sat in cash, and the portfolio made 2–15% a year.
  • The luck test: the strictest runs beat random picks from stocks already in an uptrend on return for the risk taken (the Sharpe ratio), in both periods. They did not beat them on total growth, because they were mostly in cash. The test doesn't count the 80 combinations we tried, so even that result is generous.

What failed?

  • More slots did not help. With any score, growth stayed at 18–19% a year from 30 to 100 slots, with falls around 40%. With a score filter there were not enough new uptrends to fill the extra slots, so the money waited in cash.
  • Trailing stops were the worst way to close. Trades lasted about 30 days on average, and ordinary swings kept stopping them out.
  • Closing below a medium-term average had the smallest falls but did not hold up: 27.8% a year in 2016–2021, then 7.2% in 2022–2026.
  • Risk-based sizing gave tiny positions with the no-stop exits, leaving 43–70% of the money in cash.

What are the limits?

  • The price data holds only companies still listed today. Delisted stocks are missing, which flatters every result here.
  • The models behind the score were chosen with 2023–2026 in view. So the 2022–2026 column confirms the result more than it tests it cleanly.