Every NEPSE trader eventually faces the same fork in the road: do you buy the dip, betting prices snap back to the mean, or do you ride the trend, betting momentum persists? Opinions are cheap and loud in every trading group. Data is quieter, and far more honest.
So I ran the test properly. I backtested 19 rule-based trading systems on NEPSE index daily data from 2003 to 2026, 5,333 trading sessions, spanning the 2008 and 2021 bull-market peaks, the crashes that followed, and the long sideways grinds between. The verdict is unambiguous: the NEPSE is a trend market. Momentum systems beat buy & hold with smaller drawdowns. Mean-reversion systems lost money. Here is the full breakdown.
How was the study done?
The study uses daily open, high, low, close, and volume data for the NEPSE index from 2003 through 2026, 5,333 sessions in total. Nineteen distinct rule-based systems were coded and tested across the entire window, spanning the major strategy families: trend-following (moving-average crosses, Donchian breakouts, Bollinger breakouts, RSI trend-riding), mean-reversion (oversold RSI bounces, stochastic signals, support-level buying), and candlestick-pattern systems (engulfing patterns, hammers traded in isolation).
Every system was measured on the same yardstick: compound annual growth rate (CAGR), maximum drawdown, and consistency across sub-periods. The benchmark throughout is a simple buy & hold of the index, which returned 12.87% CAGR over the test window, a strong hurdle that reflects just how powerful Nepal's long-run equity compounding has been.
What was the headline result?
The winners: momentum and breakout systems
Four systems cleared the buy & hold benchmark, all of them trend or momentum approaches:
- RSI trend-ride, 15.59% CAGR, the best performer in the study. Instead of using RSI as an oversold/overbought oscillator, this system rides RSI strength: staying long while momentum confirms the trend and exiting when momentum breaks down.
- EMA cross, the classic fast/slow exponential moving-average crossover, long when the fast average sits above the slow one.
- Donchian breakout, buying new N-day highs, the purest expression of "the trend is your friend."
- Bollinger Band breakout, entering on volatility-expansion breakouts rather than fading the bands.
Critically, these systems didn't just return more, they did it with materially smaller drawdowns than buy & hold. The trend filter keeps you out of the market's worst stretches, which is where buy & hold silently bleeds.
The losers: mean-reversion
On the other side of the ledger, every mean-reversion family lost money over the 23-year window:
- Buying RSI below 30 (the textbook "oversold bounce")
- Stochastic oscillator reversal signals
- Buying at support levels
- Trading engulfing patterns or hammers in isolation
The NEPSE does not snap back, it runs. Buying dips into a falling Nepali market, or fading a strong rally, is statistically a losing proposition over full cycles.
Why does momentum work on the NEPSE?
The numbers point to a structural feature of Nepal's market that shows up clearly in the price data itself. The NEPSE moves in long, persistent phases, multi-year advances followed by equally persistent declines. In a market shaped by such extended regimes, the strategy that aligns with the regime wins, and the strategy that fights it gets run over.
Mean-reversion assumes prices oscillate around a stable fair value. But the NEPSE's "fair value" itself trends for years at a time. An RSI below 30 in a structural bear market isn't a bargain signal, it's a description of the weather. Conversely, breakouts and moving-average crosses keep you positioned with the dominant price trend instead of against it.
Does it hold across sectors?
I extended the work to the NEPSE sub-indices, eleven sector baskets, to check whether the index-level finding holds beneath the surface. It does, with sector-specific flavors. The primary system identified per sector:
| Sector | Primary system |
|---|---|
| Banking | Bollinger breakout |
| Development Bank | RSI trend-ride |
| Finance | Donchian breakout |
| Hotels | EMA cross |
| Hydropower | EMA cross |
| Life Insurance | Bollinger breakout |
| Non-Life Insurance | Donchian breakout |
| Microfinance | RSI trend-ride |
| Investment | SMA200 + SMA20/50 (paper-trade only) |
| Trading | EMA cross (illiquidity caveat) |
| Others | EMA cross (noisy-basket caveat) |
Every sector's leading system is a trend or breakout approach, not one mean-reversion system tops any sector. The seasonality analysis added another practical layer: a positive median July return across all eleven sectors (Banking averages +9.58% in July, positive in 96% of 23 years), Microfinance averages +9.21% in January (positive in all nine observed years), while August–September is broadly weak. Seasonality is a tendency, not a signal, but it rhymes with the momentum story.
A follow-up study across 15 individual liquid scrips, completed in September 2026, confirmed the index-level finding on single names: trend and momentum systems beat dip-buying on every one of the 15 scrips in risk-adjusted terms, while the two mean-reversion systems failed everywhere.
How does this become a trading plan?
A backtest is only useful if it becomes behavior. The study converts into a rules-based plan with hard guardrails:
- SMA200 regime filter, take long trend signals only when the index trades above its 200-day moving average. Below it: chop regime, stand aside or reduce size.
- Breakout and trend entries only, no dip-buying, no anticipating reversals.
- Risk 1% of capital per trade, with a maximum of three concurrent positions.
- Pause for one week after two consecutive losses, the circuit breaker against revenge trading.
- Never average down. Journal every trade and review every Sunday.
Where does the market stand now?
As of September 18, 2026, the NEPSE closed at 2,647, below its SMA200 near 2,690, squarely in CHOP regime. The study's prescription here is patience: wait for a decisive breakout above ~2,690 before committing real size. Discipline means doing nothing when the data says nothing qualifies.
What are the limitations?
Intellectual honesty requires the caveats. Backtests are simplified models of the past: real trading involves brokerage costs, slippage, and liquidity constraints that a rules engine doesn't feel. The study also did not model promoter-share lock-in expiries, a Nepal-specific supply overhang worth watching separately. And the golden rule stands: past performance does not guarantee future results, regimes change, and any system must be re-validated as new data arrives. This article is research and education, not investment advice.
But within those limits, 23 years and 5,333 sessions deliver a clear message for anyone trading the NEPSE: stop fading the trend. Start following it, with rules, with risk limits, and with the patience to wait for your setup.
This article is research and education, not investment advice. Past performance does not guarantee future results.