Technical analysis for NEPSE: candlestick patterns, indicators, and whether they work

Technical analysis (प्राविधिक विश्लेषण, टेक्निकल एनालाइसिस) is the study of past prices and trading volume, drawn as charts, to judge which way a price is likely to move and when to buy or sell. It reads only the market's own record, never the company's accounts.

Fundamental analysis (आधारभूत विश्लेषण) does the opposite. It reads profit, assets, debt and dividends to judge what a share is worth, and buys when the price is below that; the EPS, P/E and book value guide teaches it. Many Nepali investors use both: fundamentals to choose what to own, charts to choose when.

It rests on ideas usually traced to Charles Dow around 1900: the price reflects what is known, trends tend to continue, and crowds repeat themselves. The counter-argument is the efficient-market hypothesis (कुशल बजार सिद्धान्त): a pattern that reliably made money would be traded away. The NEPSE research, further down, lands between the two.

Every chart starts from four prices for each period: the open (खुल्ने मूल्य), the high (उच्चतम मूल्य), the low (न्यूनतम मूल्य) and the close (बन्द मूल्य), written OHLC. The picture draws NEPSE's last 12 sessions three ways.

• Line chart (लाइन चार्ट): joins the closes; good for years of history, blind to each session.
• Bar chart, or OHLC (बार चार्ट): a line from the low to the high, a tick left at the open and right at the close.
• Candlestick chart (क्यान्डलस्टिक चार्ट): the same four prices, with a thick body between open and close so rising and falling sessions stand out. Most charting sites use it.
• Timeframe (समय अवधि, टाइमफ्रेम): one candle can be a day, a week or a month. Most NEPSE analysis uses daily and weekly candles.

A candlestick (क्यान्डलस्टिक) has two parts. The body (बडी) is the box between the open and the close. The wicks or shadows (बाती, विक) are the thin lines reaching the high and the low. A close above the open is drawn in one colour, below it in another.

Read the NEPSE index on 30 October 2026. It opened at 2,693.55, rose as high as 2,712.99, dipped to 2,681.58 and closed at 2,690.15: a falling candle with a body of only 3.40 points and an upper wick of 19.44. Buyers pushed it up and sellers pushed it back. The session before was the opposite: it opened at 2,660.24 and closed at 2,695.60, a hair below its high, a long rising body with almost no wicks.

• A long body: one side was in control from the open to the close.
• A small body (a spinning top, स्पिनिङ टप): the open and close were close together; neither side won.
• A long upper wick: the price went up and was sold back down.
• A long lower wick: the price fell and was bought back up.
• A gap (ग्याप): the candle opens well away from the previous close. On NEPSE gaps come from the pre-open session, 10:30 to 10:45.

A candlestick pattern is a shape of one, two or three candles that traders read as a hint of what comes next. A reversal (रिभर्सल, दिशा परिवर्तन) is said to end a trend, a continuation (निरन्तरता) to extend it. Bullish (बुलिस, तेजी) means pointing up, bearish (बियरिस, मन्दी) pointing down.

Where a pattern appears matters as much as its shape: a hammer and a hanging man are the same candle, one after a fall, the other after a rise. So each drawing shows the trend before it in grey. The names are what Nepali traders say; the table adds what each means.

Less than the cheat sheets suggest. In Bulkowski's tests the patterns that did turn the trend did so 53% to 82% of the time, and some went the other way: after the inverted hammer, said to be bullish, prices kept falling 65% of the time; after the hanging man, said to be bearish, they kept rising 59% of the time.

And a 60% reversal rate is not a 60% chance of profit. It counts which way the price broke out, not how far it went, and nothing is taken off for charges. The bearish engulfing pattern reversed 79% of the time but ranks 91 of 103 overall, because by his measure the falls that followed were small.

Peer-reviewed tests are harsher. Marshall, Young and Rose tested candlestick strategies on the Dow Jones shares, 1992 to 2002, against random price series and found no value in them (Journal of Banking & Finance, 2006). In Taiwan's 50 largest shares, 2002 to 2008, three of six two-day patterns were profitable, all bullish (Review of Financial Economics, 2012).

In Nepal, a 2016 survey of 710 investors, 29 brokers and 11 DPs found candlesticks the least used of 9 tools (Vaidya, 2018). A 2023 NepJOL study of candlesticks on the NEPSE index in the falling market of December 2022 to May 2023 found "mixed results": some very short-term help, and sometimes the wrong direction.

• Find the trend first: a reversal pattern needs something to reverse.
• Wait for confirmation, the next candle closing the pattern's way. A hammer followed by a lower close has failed.
• Location and volume: a pattern at support or resistance, on heavy turnover, means more than one mid-range on a quiet day.
• Decide where you are wrong before you buy, such as below the hammer's low.

Support (सपोर्ट, आधार तह) is a price level where falls have stopped before, because buyers stepped in. Resistance (रेजिस्टेन्स, अवरोध तह) is a level where rises have stopped, because sellers did. They are zones, not exact lines, and they form for human reasons: people who missed a price wait to buy there again, and people who bought at a top wait to get their money back.

NEPSE's daily chart from 3 August 2026 shows both. Near 2,700 the index turned down twice: 3 Aug's high was 2,705.46, and on 30 Oct it touched 2,712.99 but closed at 2,690.15, a long upper wick, then 2,657.50 the next session. Near 2,500 it found buyers: the low was 2,501.49 on 1 Sep. Turnover peaked on the late-October rise: Rs 8.89 bn on 30 Oct, the most since 16 April 2026.

• Broken support often becomes resistance, and broken resistance support.
• The more often a level has held, and the more turnover there, the more it is watched and the bigger the move when it breaks.
• Round numbers (2,500 for the index, Rs 500 for a share) act as levels because people place orders there.
• A break needs a close beyond the level, not just a wick; false breaks are common.

A trend (प्रवृत्ति, ट्रेन्ड) is the direction the price is taking over the timeframe you are looking at. An uptrend (उकालो) makes higher highs and higher lows, a downtrend (ओरालो) lower ones, and a sideways market (साइडवेज) stays in a range. A trend line (ट्रेन्ड लाइन) joins two or more lows in an uptrend, or highs in a downtrend.

On NEPSE's monthly chart the big trends are plain: up from 1,202 in May 2020 to 3,080 in July 2021, then down to 1,850 by May 2023, 40% lower, then years of sideways range. "NEPSE is in an uptrend" means nothing without the timeframe.

Volume (भोल्युम, कारोबार परिमाण) is how many shares changed hands, in kitta. Turnover (कारोबार रकम) is the rupee value traded, the market number in the news. Both say how many people took part in a move.

A move on heavy volume has more behind it than the same move on a quiet day; a break above resistance on thin turnover is suspect. In the 57 sessions drawn above, NEPSE's turnover averaged Rs 4.76 bn a day; on 29 Oct and 30 Oct it was Rs 8.57 bn and Rs 8.89 bn.

In Vaidya's 2016 survey, volume indicators were the second most used of 9 tools. On-balance volume (OBV) adds a day's volume when the price closes up and takes it away when it closes down.

A moving average (मुभिङ एभरेज, चलायमान औसत) is the average of the last few closes, worked out again each period, so it smooths the noise. The simple moving average (SMA) adds the last n closes and divides by n: NEPSE's 12-month SMA at the end of October 2026 was the average of the twelve month-end closes from November 2025, 2,676.71.

The exponential moving average (EMA, एक्सपोनेन्सियल) gives recent closes more weight: EMA today = EMA yesterday + k × (close today − EMA yesterday), where k = 2 ÷ (n + 1). For a 12-period EMA, k is 15.4%: each new close pulls the average that far towards it. EMAs turn sooner than SMAs, and give more false turns.

Traders watch whether the price is above its average, which way the average slopes, and crossovers. A golden cross (गोल्डेन क्रस) is a shorter average crossing above a longer one; a death cross (डेथ क्रस) the reverse. The classic daily pair is the 50-day and 200-day; 200 NEPSE sessions is about ten months.

The picture uses month-end closes with the 12- and 24-month averages, which crossed 4 times in ten years. The death cross of September 2022 came when the index was already 40% below its July 2021 month-end high. That is the price of smoothing: a moving average always lags.

A rule can be tested: write it down and run it over past prices. Here three common rules run over NEPSE's month-end closes from December 2017, the first month all can be worked out, to October 2026. Each starts with Rs 1,00,000, buys the index as if it were a share at the month-end close when its rule says "in" and sells when it says "out", paying today's commission, SEBON fee, DP charge and capital gains tax on every trade; cash sits idle while out.

• Every rule ended below simply holding. Of the 10-month rule's 15 trips into the market, two made money, the biggest 109%, and 12 lost 2% to 10% each, before costs; it went back in at the last month-end. A trend rule is meant to lose small and win big; here the small losses came too often.
• Selling in falls did not always soften them. Two of the rules had a worse fall from a high than holding's 40%: being whipsawed out at lows and back in at highs compounds.
• Another decade, another answer: on daily closes from 2012 to 2022, before costs, Karki and co-authors' moving-average crossover earned 21.33% a year against 16.09% for holding.

The RSI (relative strength index, सापेक्ष शक्ति सूचकांक) compares the size of recent rises with recent falls, on a scale of 0 to 100. J. Welles Wilder published it in 1978 with 14 periods, and called a reading above 70 overbought (ओभरबट, अत्यधिक खरिद) and below 30 oversold (ओभरसोल्ड, अत्यधिक बिक्री).

RSI = 100 − 100 ÷ (1 + RS), where RS = average gain ÷ average loss over the period. Worked on NEPSE's last 15 month-end closes, which give 14 changes, September 2025 to October 2026: the rises add up to 622.91 points and the falls to 682.59. The average gain is 622.91 ÷ 14 = 44.49, the average loss 48.76. RS = 0.9126, and RSI = 100 − 100 ÷ 1.9126 = 47.7.

Chart sites show a different number, because after the first 14 periods Wilder smooths the averages instead of starting again: new average gain = (previous average × 13 + this gain) ÷ 14. Run that way from 2016, NEPSE's monthly RSI for October 2026 is 53.2; the daily 14-day RSI on 3 November 2026 was 60.0. Worked arithmetic, not forecasts.

• Overbought is not a sell signal. NEPSE's monthly RSI stayed above 70 for 10 months in a row, November 2020 to August 2021, while the index rose another 54% to its top. Selling at the first reading above 70 missed most of the boom.
• It works best in sideways markets; in a strong trend the readings sit at one extreme.
• Divergence (डाइभर्जेन्स): the price makes a new high and RSI a lower one, said to show the rise tiring. It can happen several times before a top.
• On NEPSE the RSI rule did worst of the five Karki and co-authors tested, 2012 to 2022: it lost 4.81% a year before costs while holding made 16.09%. A period shorter than 14 reacts faster and misleads more often.

MACD (moving average convergence divergence, एमएसीडी) follows the gap between a fast and a slow exponential average; Gerald Appel devised it in the late 1970s.

• The MACD line = the 12-day EMA − the 26-day EMA.
• The signal line (सिग्नल लाइन) = the 9-day EMA of the MACD line.
• The histogram (हिस्टोग्राम) = MACD − signal, drawn as bars around zero.
• Reading it: MACD above zero means the shorter-term trend is up. MACD crossing above its signal line is read as bullish, below as bearish.

Bollinger Bands (बोलिन्जर ब्यान्ड), devised by John Bollinger in the early 1980s, draw a 20-day simple average with a band two standard deviations (मानक विचलन) above and below. Standard deviation measures how widely closes spread, so the bands widen when the market is jumpy and narrow when calm.

On 3 November 2026 the 20-day average of NEPSE's closes was 2,618.93 and their standard deviation 38.05, so the bands ran from 2,542.83 to 2,695.04; the index closed at 2,664.13. Of 241 sessions, 25 (10%) closed outside the bands.

• A touch of a band is not a signal, in Bollinger's own rules. A close above the upper band in a strong rise is strength, not a reason to sell.
• The squeeze (स्क्वीज): bands at their narrowest in months often come before a big move, in a direction the squeeze does not tell. NEPSE's 20-day bands were narrowest on 4 January 2026, 2.4% of the average from edge to edge, and widest on 26 March 2026, 16.9%.
• In Karki and co-authors' test a Bollinger rule traded 9 times in ten years for 18.05% a year before costs.

Chart sites offer dozens more, most of them other ways of measuring trend and momentum.

Textbook technical analysis was written for deep markets with short selling. NEPSE differs in ways that change what a chart shows.

• A 15% daily limit for each share (उतारचढाव सीमा) since 20 April 2026, 10% before, measured from about the previous close. Big news can take two or three sessions to reach the price, and a day stuck at the limit draws a candle with almost no body. The Himalayan Times also reported a 3% limit on each move from the prevailing price during continuous trading; check NEPSE's notice.
• Circuit breakers (सर्किट ब्रेकर): an index move of 5% in the first two hours halts trading for 15 minutes; 8% closes the day. Before April 2026 the market closed at 6%: in our 260 sessions from September 2025, the biggest moves from one close to the next were a fall of 6% on 18 September 2025 and a rise of 6.01% on 9 March 2026.
• Gaps come from the pre-open (10:30 to 10:45), where the price may move up to 5%: on 15 September 2026 the index opened 3.3% above the previous close.
• Monday to Friday since April 2026: the last Sunday session was 5 April 2026, the first Friday 10 April 2026. Indicators count sessions, so a Dashain closure is simply skipped.
• T+2, no intraday trading, no short selling. A share bought today reaches your demat two trading days later and only then can be sold; SEBON's September 2026 intraday concept paper is not in force. A sell signal can only mean selling what you hold, while many textbook tests count profits from falls too.
• Prices jump at book closure. NEPSE adjusts a share's price for bonus and right shares: after a 15% bonus, Nepal Lube Oil's Rs 277.40 became Rs 241.22 in September 2026, 13% lower with nothing lost. An unadjusted chart draws that as a crash through every support line; cash dividends are not adjusted. See the bonus and right share guide and the book closure guide.
• Thin trading. Many companies trade little on most days, so a handful of orders can draw a breakout or a golden cross. Stale prices in small shares also make sector indices look predictable when nobody could have traded at those prices.

Operator (अपरेटर) is the market's word for a person or group that moves a share's price on purpose, usually a small company's. SEBON's notices name the methods: circular trading (सर्कुलर ट्रेडिङ: selling to yourself or to friends to fake activity), cornering (buying up most of the shares on offer), insider trading (भित्री कारोबार: trading on news before it is public) and false orders.

On 15 June 2021 SEBON published 51 companies whose prices it found overvalued and linked to insider trading, circular trading and cornering. Its screen: a price up more than 300% in a year, a P/E above 100, negative EPS, and net worth below the issue price. Onlinekhabar reported that false orders and TMS's all-or-none option had been used to push prices up, and that SEBON had the option removed. In such a share the chart is the operator's drawing, and a breakout may be someone else's way out.

The floorsheet (फ्लोरसिट) is NEPSE's record of every trade: share, quantity, price, and the buying and selling brokers' numbers. Broker analysis reads it for who is buying. In 2078 BS SEBON told NEPSE to publish broker numbers only after trading ends, Nepal Live reported. A broker number is a firm with thousands of clients: heavy buying through one may be many small buyers, or one large one.

• NEPSE said in August 2025 that urging people to trade, or recommending buys and sells, on social media is illegal, and that it "neither advises anyone to buy or sell securities, nor guarantees returns".
• SEBON warned in October 2026 against deciding "solely on the basis of information, suggestions or promotional content" on social media, and to trade only through licensed firms.
• A group selling calls that shows only its winners is the oldest pattern on NEPSE: ask for every call, dated.

The honest answer: sometimes, before costs, and not reliably. The best NEPSE test, by Dipendra Karki and four co-authors (Quest Journal of Management and Social Sciences, 2023), ran five indicator rules over the NEPSE index and six sector indices, daily, from September 2012 to August 2022. Four of the five beat holding the index before costs; RSI lost money. But on 500 random price series built from NEPSE's own returns the rules did about as well: the bootstrap results "discard the predictability power of technical indicators", the authors wrote.

Studies of whether NEPSE is weak-form efficient (whether past prices say nothing about future ones) mostly find it is not. But predictable is not the same as profitable: a pattern can be too small to survive commission and tax, or come from small shares nobody could have bought at yesterday's price.

Every round trip, a buy and a sale, pays commission, the SEBON fee and a DP charge on both sides, and capital gains tax on any gain: 5% if you held the shares 365 days or less, 3.75% if longer, since 22 September 2026. On a Rs 1,00,000 trade (200 shares at Rs 500) the charges come to Rs 740 for the round trip, 0.74% of the money, so the price must reach Rs 503.72 before you are even.

A 3% gain makes Rs 3,000 on paper and keeps Rs 2,137.17 after charges and Rs 112.48 of tax. The more often you trade, the more goes to charges:

Traders who last do not try to be right more often; they lose small when wrong. A stop loss (स्टप लस) is the price at which you sell because the trade's idea has failed, chosen before you buy: just below the support, or the hammer's low. Position size (पोजिसन साइज) is the number of shares that makes a hit stop cost only a set part of your money, often 1–2%.

1. 1Set the risk. With Rs 2,00,000 of trading money and 1% at risk on one trade, the most you accept losing is Rs 2,000.
2. 2Set the stop where the chart idea fails. Buying at Rs 400 with support just under it, the stop goes 8% below, at Rs 368.
3. 3Size the trade. Each share loses about Rs 35.88 at the stop, charges included, so the largest multiple of 10 within the limit is 50 shares: Rs 20,100, 10% of the money.
4. 4If the stop is hit, selling 50 shares at Rs 368 brings back Rs 18,306: a loss of Rs 1,794, within the Rs 2,000 you set.
5. 5Write it down before you buy, sell when the price closes below the stop, and never move a stop further away.

• Reading a pattern without its trend. The same candle is a hammer after a fall and a hanging man after a rise.
• Stacking indicators that say the same thing: RSI, the stochastic and MACD all measure momentum.
• Changing the timeframe until the chart agrees with what you wanted to do.
• Selling only because RSI says overbought in a strong rise: NEPSE's monthly RSI sat above 70 for 10 months in 2020–21.
• Moving the stop, or averaging down when the trade goes wrong.
• Trading every week without counting charges and tax.
• Using an unadjusted chart across a bonus or right share book closure.
• Trusting a backtest that leaves out costs or counts short sales.
• Copying calls from groups, often from people who bought first.
• Forgetting the company: a chart cannot see losses, bad loans or a coming right issue. Read the fundamentals too.

Free charts of the index and every listed share, with the indicators above, are on several Nepali sites. We do not rank them; check that a site says whether its prices are adjusted for bonus and right shares.

• NEPSE (nepalstock.com): the official prices, indices, floorsheet and notices of rule changes.
• Your broker's TMS: live prices and market depth while you trade.
• ShareSansar, Merolagani, NepseAlpha, Chukul, ShareHub and others: charts, screens and news.
• Books in English most courses draw on: John Murphy's "Technical Analysis of the Financial Markets" (1999); Steve Nison's "Japanese Candlestick Charting Techniques" (1991); Wilder's "New Concepts in Technical Trading Systems" (1978); "Bollinger on Bollinger Bands" (2001); Bulkowski's encyclopedias for test results. Nepali share market books exist; we have not reviewed them.

1. 1Read single candles on NEPSE's weekly chart for a month.
2. 2Mark support, resistance and trend on three charts, without trading.
3. 3Add one moving average and RSI, and see what each said at past turns, wrong ones included.
4. 4Write a rule with entry, stop and size, and paper trade it (no money) for three months, recording every trade with charges.
5. 5Compare the results with simply holding the index over the same months.
6. 6Only then use small money, and keep long-term savings out of it, in a SIP or a mutual fund.

Nepali market writing mostly uses the English word in Devanagari (सपोर्ट, आरएसआई); the descriptive Nepali terms in brackets appear in some explainers. Both are understood.