Backtesting Volatility-Adjusted Momentum with a Gold Regime Filter (2018 - 2026)

Investing Strategies
1

Hi everyone,

I have been studying momentum strategies in the Indian context (specifically typically Nifty 200/500 momentum), but the primary risk that concerns me is the massive drawdowns during regime shifts (e.g., 2018 midcap crash, 2020 Covid).

I wanted to test a hypothesis: Can reallocation to gold using a dynamic “Regime Filter” reduce drawdown in a Smallcap portfolio and improve CAGR?

backtest_with_app_details
backtest_with_app_details1833×1034 116 KB

Existing backtests often suffer from survivorship bias or lack the ability to model “Asset Class Switching” (Equity → Gold) easily. So, I coded a custom backtesting engine to simulate this using Point-in-Time data.

The Strategy Logic: Instead of pure price momentum, I tested a Volatility-Adjusted Ranking:

config
config1268×236 24.3 KB

  • Ranking Formula: (Weighted Returns 6m/3m/12m) / (Volatility)

  • Hypothesis: High-beta stocks that jump up on low volume are penalized. Consistent compounders are rewarded.

  • Regime Filter: If Nifty 50 breaks Weekly Supertrend → Exit 100% Equity → Enter 100% Gold (GOLDBEES).

The Backtest Results (2018 – Feb 2026):

  • Universe: Nifty MidSmallcap 400 + Next 50

  • CAGR: ~38% (Gross)

  • Max Drawdown: -20.9% (vs Nifty -34%)

  • Sharpe Ratio: 1.51

Curious to hear thoughts from the community. What other risk metrics or stress tests should I look at to validate this further?

Note: This is a raw backtest using EOD prices. I haven’t accounted for taxes or slippage yet, which would reduce the real-world CAGR, but the relative risk reduction from the Gold hedge seems significant.

3 Likes
2
  1. I hope, you used the index constituents list that was effective on each day in history.
    1. If yes, did you refresh the constituents on 1st Feb and 1st Aug (when the upcoming changes are announced)?
    2. or did you refresh the constituents at the end of 31st March and 30th September (when the changes come into effect)?
  2. Thanks for revealing that you did not account for taxes and slippages.
    1. But what’s the point of running non-realistic tests?
  3. Eight years backtest is not enough. You could test from 8th March 2007 onwards, when both of these indices and GOLDBEES were available.
    1. I tested a similar strategy that produced 54% CAGR in the last 10 years but only 12% CAGR in the last 20 years after taxes, brokerage and a little slippage.
3

Hi Abishek,

Thank you for sharing the your working/ proven model with 38% CAGR. it’s really very good returns. Can you please share more details on how you filter shares from nifty 200 and how many shares you invest at time and holding time. which tools are referring. It would be great if you provide more details