Technical Analysis Foundations

Technical Analysis Foundations

Technical analysis is the practice of inferring probable future price behaviour from the historical record of price and volume alone. Done well, it is humble and statistical. Done badly, it becomes pareidolia at $50 a chart. This lesson lays out the assumptions, the core tools, and — just as importantly — the boundaries of what TA can and cannot tell you.

1.8.1What technical analysis claims

Technical analysis (TA) makes one core empirical claim: that price and volume contain enough information about supply and demand to give a small but real edge in forecasting near-term behaviour. It does not claim to know fair value, future cash flows, or what management will do next quarter. Those are fundamental questions. TA studies the footprints, not the animal.

This is a defensible claim, but a narrow one. Decades of academic work — Lo and MacKinlay (1988), Jegadeesh and Titman (1993), Moskowitz, Ooi and Pedersen (2012) — have demonstrated that simple price-based signals (momentum, mean reversion at extremes) deliver abnormal returns net of plausible transaction costs across asset classes and decades. That is the technical analyst’s empirical floor. It is also a long way from “the head and shoulders pattern works.”

The honest framing: TA gives you a probability distribution conditioned on the recent past. A rising 50-day moving average with breadth above 60% and volume 1.4x average tilts the next two-week SPY return slightly toward positive. +0.6% versus the unconditional +0.2%, say. Multiply that small edge by hundreds of independent trades and you have a business. Mistake it for prophecy and you have a problem.

Charts do not predict the future. They describe the present in a way that occasionally rhymes with the past.

1.8.2The three Dow assumptions

Modern TA inherits its philosophical foundation from the 1900-era writings of Charles Dow. Three assumptions are baked into nearly every technique and remain useful as testable claims:

1.8.2.1Price discounts everything

Public information about earnings, sentiment, geopolitics and flows is reflected in price relatively quickly. The technical analyst therefore studies the synthesis (price) rather than each input separately. This is the weak-form efficient market claim, with which TA is largely compatible — weak-form efficiency rules out trivial historical-price patterns, not all of them.

1.8.2.2Prices move in trends

Once a tendency establishes, it persists more often than chance until counter-evidence accumulates. The empirical version: 12-month price momentum on US equities has produced positive risk-adjusted returns in roughly 8 of every 10 rolling decades since 1927. That is robust enough to bet on, fragile enough to drawdown.

1.8.2.3History rhymes via human behaviour

The third assumption is the softest: that recurring market psychology produces recurring price structures. Fear, greed and herding are stable enough that bases, breakouts and capitulations look similar across eras. This is partly true and partly self-fulfilling — enough traders watching the same patterns make the patterns somewhat real.

1.8.3Reading a chart: timeframe, scale, candle

Before any indicator is added, three choices shape what the chart says.

1.8.3.1Timeframe matters more than indicators

The same NVDA price series tells four different stories at 1-minute, 15-minute, daily, and weekly resolution. A daytrader and a swing trader looking at the same name should expect to disagree about “the trend,” because they are measuring different objects. Best practice: identify your operating timeframe and one timeframe up. The higher chart sets context; the operating chart sets entries.

1.8.3.2Linear vs log scale

Linear charts make equal price moves equal pixels. Log charts make equal percentage moves equal pixels. For multi-year charts of names that have moved an order of magnitude (NVDA from $13 in 2018 to $118 in 2026), log scale is mandatory — a 10% pullback in 2018 looked enormous on linear and invisible today, but psychologically it was the same move.

1.8.3.3The candle as a one-bar story

A candle compresses four prices: open, high, low, close. The body shows open-to-close; the wicks show extremes. A long lower wick on a daily SPY candle that closes near the high tells a story — sellers tried, buyers reasserted by the bell. That story is more informative than the close alone but only when paired with context (where, on what volume, in what trend).

1.8.4Trend, momentum, mean-reversion

Almost every TA strategy is some combination of three primitives. Knowing which one you are pressing is most of the discipline.

1.8.4.1Trend

A trend is a sequence of higher highs and higher lows (uptrend) or lower highs and lower lows (downtrend). A simple operational test: is price above its rising 50-day moving average and is the 50 above the 200? If yes, you’re in a primary uptrend on the daily. SPY at $723 with the 50-day at $711 and the 200-day at $682 qualifies.

1.8.4.2Momentum

Momentum is the rate of change of price. The cleanest measure is N-period return: 252-day momentum is the standard academic version. Momentum strategies buy strength and sell weakness. Their long-run edge is real but punctuated by sharp “momentum crashes” — February 2009, April 2020 — when the prior winners rotate violently to losers.

1.8.4.3Mean reversion

The opposite bet: extreme moves over short windows tend to retrace. A stock down −5% on a benign day with no news has a measurable bounce probability over the next 3–5 sessions. Mean-reversion edges work best on liquid names, on short timeframes, in non-trending tape.

1.8.5Indicators: a calibrated toolkit

An indicator is a transformation of price (and sometimes volume) into a number meant to make a property visible. Most retail traders bolt too many onto a chart; the result is decoration, not information. A working professional’s chart usually has two to four indicators, chosen to be statistically distinct.

1.8.5.1Moving averages: simple but powerful

The 50– and 200-day SMAs are the most-watched lines in finance. Their crossing — the “golden cross” up, “death cross” down — is meaningful not because the math is clever but because so many funds price-action mandates reference them.

1.8.5.2RSI and the overbought trap

The 14-period RSI is bounded [0, 100]. Conventionally, >70 is “overbought,” <30 “oversold.” The trap: in strong trends, RSI camps above 70 for weeks. Selling NVDA simply because RSI hit 78 in a primary uptrend is a recipe for chasing it back later. RSI works as confirmation, not as a standalone signal.

1.8.5.3Volatility-aware sizing

ATR and Bollinger Bands convert price into a volatility-adjusted view. NVDA with a 14-day ATR of $4.30 moves about 3.6% per day on average; SPY with an ATR of $5.20 moves 0.7%. Position size, stop distance and target should all scale to ATR rather than fixed dollar values.

1.8.6What TA cannot do

A serious foundations lesson must include the boundaries.

1.8.6.1It cannot price a company

If NVDA is structurally overpriced at $118 on cash-flow grounds, no chart pattern will tell you. TA times entries and exits within whatever fundamental regime the market is pricing.

1.8.6.2It is poor at unique events

By construction TA assumes recurrence. The COVID crash of March 2020, the August 2024 JPY-carry unwind, and the 2008 Lehman week were regime breaks where most pattern-based systems took outsized losses. Risk management, not chart patterns, saves you in those weeks.

1.8.6.3It is not a replacement for risk management

TA can find favourable entries. It cannot tell you how big to size, when to stop, or when the regime has changed. Those decisions belong to the trading plan (Lesson 8.6).

1.8.7Common mistakes and how to fix them

1.8.7.1Indicator stacking

Symptom: a chart with 7+ indicators, multiple of which measure the same thing (RSI + Stochastic + Williams %R are all the same family). Fix: pick one momentum, one trend, one volatility, one volume gauge. Anything more is decoration.

1.8.7.2Single timeframe blindness

Symptom: trading off the 5-minute chart while ignoring that the daily is breaking down. Fix: always read context one timeframe up. Don’t fight the higher trend.

1.8.7.3Pattern hunting without setup definition

Symptom: “I see a head and shoulders” — said about a chart that, on objective measurement, has neither a clear neckline nor symmetric shoulders. Fix: write the rule for each pattern with quantitative thresholds (e.g. shoulders within 3%, neckline slope under 5°) before classifying.

1.8.7.4Confirmation lookback

Symptom: deciding the trend bias after seeing the breakout, then finding the indicators that agree. Fix: read context first, decide bias, then look for entry. The order matters.

1.8.7.5Treating TA as oracle

Symptom: pyramid up a position because “the chart says” the move is going to $140. Fix: the chart says probabilities. Position sizing assumes worst-case, not target.