What Technical Analysis Is, and What It Is Not
Technical analysis studies price and trading volume to describe the balance of supply and demand. It describes conditions; it does not forecast outcomes.
MadStockAlerts Research · Updated August 28, 2026
What to take away
- It analyses the price record rather than the business behind it.
- Every indicator is a transformation of price and volume, so none contains new information.
- Indicators lag by construction, because they are computed from data already in.
- It describes what has happened and what is happening, not what will happen.
- Its weakest ground is thin securities and scheduled events.
MAD Academy Training Video · 0:44
What a Chart Can and Cannot Tell You
The honest version: an indicator is a restatement of price, not a source of new information about the company.
This lesson is part of a Stock Alerts + Tools plan.
What it studies
Fundamental analysis asks what a business is worth. Technical analysis asks what has been happening to the price of its shares: where buying has been persistent, where selling has appeared, whether participation is rising or fading.
The two are not competitors. They answer different questions, and a great many participants use one to decide what to look at and the other to decide when. A company can be undervalued for three years, and nothing in a balance sheet says which of those years it stops being so.
The underlying claim is modest and defensible: prices are set by supply and demand, supply and demand leave a record, and that record is worth reading. Everything contentious in the field is an extension of that claim rather than the claim itself.
Everything is derived from the same inputs
There are five raw numbers per bar: the open, high, low and close, plus volume. Every indicator ever devised is an arithmetic transformation of those five. None of them adds information that was not already in the series.
This is why stacking eight indicators on one chart usually adds confusion rather than confidence. Most of them are computed from the same closes and therefore agree with each other by construction, which feels like confirmation and is arithmetic.
A simple moving average, an exponential moving average and MACD are all functions of the same closing prices. When all three turn up together that is not three independent signals; it is one signal reported three times.
The one genuinely separate input is volume, which is why volume confirmation recurs throughout the pattern literature. It is the only readily available measure that is not another view of price.
- 1Open, high, low, close, volumeEverything a chart records. There is nothing else in it
- 2Averaging, differencing, ratiosSmooth it, subtract a lagged copy, divide by a range
- 3Every indicator on every platformIncluding the proprietary ones
The lag is structural
A 50-day simple moving average is an average of data that has already happened. It cannot turn before the prices that constitute it turn. That lag is not a flaw to be engineered away; it is what an average is.
Attempts to remove it, by shortening periods or weighting recent data more heavily, trade lag for noise. Shorter and faster means earlier and more often wrong. This trade-off has no solution, only positions along it, and every parameter choice in this pillar is a position on that line.
| Setting | Responds | Cost |
|---|---|---|
| Short lookback | Quickly | Frequent false turns in a range |
| Long lookback | Slowly | Confirms a change well after it began |
| Heavier recent weighting | Sooner than a simple average | More sensitive to a single outlier bar |
The honest framing
An indicator reading is a description of a condition. RSI at 78 says recent gains have substantially outweighed recent losses over the lookback. It does not say the stock will fall, and readings can stay above 70 for months in a strong trend.
The word signal does a lot of unearned work in this field. Treat every reading as a description of the present, and any inference about the future as an assumption laid on top of it that may not hold.
There is a practical test for whether a statement about an indicator is descriptive or predictive: can it be falsified by the next bar? Overbought describes the last fourteen bars and cannot be wrong about them. Due for a pullback is a forecast, and it is wrong far more often than the confidence with which it is usually said.
Where it is weakest
- Thin securities, where the price series reflects a handful of trades rather than continuous interest.
- Around scheduled events, where a filing or a rate decision overwhelms any prior condition.
- After corporate actions, if the data has not been properly adjusted for splits and dividends.
- In extremely low-volume periods, where the same shape forms from far less participation.
- In the first minutes of a session, where the order book is still forming and every reading is unstable.
The common thread is that technical analysis assumes a continuous, liquid, correctly adjusted price series produced by many participants. Wherever that assumption fails, the arithmetic still produces a number and the number stops meaning anything.
The efficient market objection, stated fairly
The weak form of the efficient market hypothesis states that past prices carry no information about future prices, because any pattern that did would be traded away by the people who found it. If it holds, technical analysis cannot work in principle, and this is the strongest objection to the entire field.
The honest position is that the evidence is mixed rather than settled either way. Momentum, the tendency of securities that have risen to continue rising over intermediate horizons, is one of the most replicated anomalies in finance and is a price-based effect. At the same time, most specific chart patterns have failed to survive careful testing, and the studies that support them are frequently unable to rule out data mining.
| Claim | State of the evidence |
|---|---|
| Momentum over 3 to 12 months | Strong, replicated across markets and decades |
| Short-term reversal | Reasonably supported, and difficult to trade after costs |
| Volume confirming price | Mixed, and sensitive to how it is measured |
| Specific chart shapes | Weak. Results rarely survive out-of-sample testing |
| Round numbers acting as levels | Some support, and effects are small |
The defensible position for a reader is neither dismissal nor acceptance. Chart reading is a way of describing what has happened to a security and where the decision points sit, and any claim that it predicts what happens next needs the same evidence any other such claim would.
Why an indicator that works stops working
There is a structural reason why published indicators degrade, and it is not that markets are mysterious. An effect that is real and widely known becomes traded, and being traded changes the thing it was measuring.
- 1An effect is discoveredSomeone identifies a pattern that has held historically, in data that was available to everyone.
- 2It is published or leaksWhether in research, a book, or a platform's default indicator set.
- 3Participants trade itWhich moves prices earlier, since everyone is acting on the same trigger.
- 4The effect shrinks or movesThe edge is competed away, or it shifts to a different horizon where fewer people are looking.
Studies of published anomalies have found that returns tend to decline materially after publication, which is what this process predicts. It is also why an indicator that appears in every default chart package is the least likely place for an undiscovered effect to be sitting.
There is a counterweight worth stating: an effect grounded in a durable behavioural or structural cause can persist even once known, because the cause does not go away. Reluctance to realise losses is not removed by being documented.