Technical analysis is the study of price and volume history to make decisions about timing. It does not ask what a company is worth. It asks whether the market’s recent behaviour tells you anything useful about its next move, and it answers with rules specific enough to be wrong.
That last part is where most introductions stop short. There are a great many technical patterns, and they are not equally supported by evidence. This article covers the tools, and it is honest about which claims survive contact with data.
For the question of what a business is actually worth, see fundamental analysis. The two are complements, not rivals.
The three assumptions, and which one carries the weight
Technical analysis rests on three claims, usually stated together as though they were equally solid. They are not.
Price reflects available information. Uncontroversial in its weak form. Prices move on news, and by the time you read a headline the move has usually happened. This assumption is doing real work and almost nobody disputes it.
Prices move in trends. Partly supported. Momentum, the tendency of recent winners to keep winning over horizons of a few months, is one of the most widely documented effects in finance and it shows up across markets and decades. Note the horizon though. Momentum evidence is strongest over three to twelve months, which is not the timescale most chart traders operate on.
History repeats itself. This is the weak one, and it is the assumption most chart patterns quietly depend on. The argument is that recurring human psychology produces recurring price shapes. That is plausible, and it is also exactly the kind of claim that survives casually and fails formally, because a chart with enough history contains every shape you might look for. Treat pattern claims as hypotheses to test, not as received knowledge.
Being clear about which assumption you are leaning on tells you how much confidence a signal deserves.
Reading a candle
Before any pattern, the mechanics. A candlestick compresses four numbers for one period into one shape.
The choice of period matters more than beginners expect. A one-minute chart of a liquid NIFTY constituent and a weekly chart of the same stock are describing different processes with different noise characteristics. On NSE the regular equity session runs from 9:15 to 15:30, preceded by a pre-open call auction from 9:00, and the official market timings are worth knowing precisely because the first and last few minutes behave differently from the rest of the day.
Picking a timeframe
The honest way to choose is to work backwards from how often you can actually pay attention, then accept the consequences.
Intraday charts, from one to fifteen minutes, contain mostly noise. The signal-to-noise ratio is at its worst here, costs are incurred most often because you trade most often, and you are competing with participants who are faster than you by design. It is the hardest place to start and the most commonly recommended one.
Daily and weekly charts suit most people. Noise averages out, the momentum evidence discussed above actually applies at these horizons, and a decision every few days is compatible with having a job.
Monthly charts describe regime rather than trade timing. They are useful context and a poor entry trigger.
One rule holds across all of them: pick your timeframe before you look at the chart. Choosing the period on which a pattern appears clearest is fitting, not analysis.
Support and resistance, and what they actually are
Support is a level where buying has repeatedly been strong enough to stop a decline. Resistance is the mirror image. The useful question is why they exist at all, because “the chart bounced there before” is a description, not a mechanism.
Two mechanisms are real. The first is order clustering: traders place stops and limits at round numbers and at prior extremes, so resting liquidity genuinely concentrates there. The second is memory of pain. People who bought at a level and watched it fall often sell when it returns to break-even, which supplies real selling pressure at a specific price.
Treat these as zones rather than exact prices. A level that has to be precise to the rupee is a level you have fitted to the chart.
Indicators, grouped by what they actually measure
Most introductions list twenty indicators. That is unhelpful, because nearly all of them are transformations of the same two inputs. There are only four things they measure.
| Measures | Typical tools | What it is really doing |
|---|---|---|
| Trend | Moving averages, MACD | Smoothing price to suppress short-term noise |
| Momentum | RSI, rate of change | Comparing recent moves against their own recent history |
| Volatility | Bollinger Bands, ATR | Scaling by dispersion, so a move is judged relative to normal |
| Participation | Volume, OBV | Asking whether a move happened on real activity |
Once you see the grouping, the redundancy becomes obvious. Stacking three momentum oscillators on one chart does not give you three independent opinions. It gives you one opinion, drawn three times, and a false sense of confirmation.
The genuinely useful move is to combine across categories rather than within. A trend signal confirmed by participation is two different questions agreeing. Two momentum oscillators agreeing is one question asked twice.
How to tell whether an indicator actually works
This is the section most introductions skip, and it is the one that separates a trader from someone reading tea leaves with better software.
An indicator makes a claim. Write it down precisely enough to test: buy NIFTY when the 50-day moving average crosses above the 200-day, exit on the reverse cross. That is now a rule that produces a specific set of trades over any historical period, and you can ask what it earned after costs.
Then apply the discipline from quantitative analysis: adjust your data for corporate actions before computing anything, build the universe as it existed at the time rather than as it exists now, subtract realistic transaction costs, and keep a holdout period you look at once. Most indicator rules that feel obviously correct do not survive that process. Learning which ones do is worth more than memorising another forty patterns.
Five numbers tell you most of what you need to know about a rule.
- Number of trades. Fewer than about thirty and you have an anecdote, whatever the return looks like.
- Return after costs, not before. Multiply your annual turnover by a realistic round-trip cost and subtract it.
- Maximum drawdown, because it decides whether you would actually have held on through the bad stretch. A rule you abandon at the worst moment returns nothing.
- Hit rate together with average win over average loss. Either one alone is misleading. A rule can be right a third of the time and still be excellent.
- Performance in the holdout period you set aside and did not tune on.
Then ask the question that kills most technical rules: did it beat simply buying and holding the same index over the same period, after costs and after tax? A signal that generates forty trades a year to underperform a decision you could have made once is not a strategy. It is an expensive hobby, and it is worth finding that out on a spreadsheet rather than over three years of live trading.
Two failure modes are worth naming because they are so easy to fall into.
Parameter shopping. If a 14-period RSI does not work, try 9. Then 21. Then 12. Somewhere in that search you will find a number that looks excellent on your sample, and you will have discovered nothing except that you tested many numbers. Decide the parameter before you look, or accept that your result is a hypothesis rather than a finding.
Hindsight pattern-spotting. Head and shoulders formations are obvious on the left of the chart and ambiguous on the right, which is the only side that pays. If you cannot write a rule that identifies the pattern using only data available at the time, you do not have a signal, you have a story.
What changes from one market to another
Four things differ enough to matter.
Circuit limits truncate moves. Individual stocks carry price bands, and index-level circuit breakers halt trading after large moves. A volatility measure computed across a banded stock understates the move that wanted to happen, because the exchange stopped it. This distorts anything calibrated on standard deviation.
Liquidity falls away quickly outside the large caps. A pattern on a stock outside the NIFTY 100 universe may be a pattern in a series where a handful of trades set the price. The index methodology uses an impact-cost threshold to define tradeable liquidity, which is a reasonable rule of thumb for your own universe too.
Expiry effects are real and periodic. Index option expiry produces recurring flows that have nothing to do with the trend you think you are trading. Any intraday signal tested on Indian index data should be checked for whether its returns are concentrated on a particular weekday.
Corporate actions are frequent. Bonuses and splits are common on Indian equities, and an unadjusted series will show gaps that no trader experienced. Every indicator computed on that series inherits the error.
Where technical analysis fails
It fails at prediction, and it is usually sold as prediction. What it can do is describe the current state of a market and give you consistent rules for acting on that description. That is genuinely valuable, and it is a smaller claim than the one on most course landing pages.
It also fails around information events. A chart has nothing useful to say about an earnings surprise or a regulatory announcement, because the whole method assumes the relevant information is already in the price. When new information arrives, technical levels break without warning, which is one reason position sizing and stops matter more than signal quality. Risk management covers that side properly.
And it fails quietly when a market changes character. A mean-reverting rule tuned on a range-bound period will keep firing during a trend, losing steadily, while looking exactly as it did when it worked.
Where to go from here
Start narrow. One trend tool, one participation check, one timeframe, tested honestly on data you have cleaned yourself. That is more useful than a screen covered in indicators, and you will actually understand why it does what it does.
From here, quantitative analysis formalises the testing, portfolio construction handles combining signals into positions that make sense together, and machine learning for stock selection extends the same thinking to models that infer their own patterns, with all the additional ways those can mislead you.
The charts are worth reading. They are just not worth believing without evidence.