Moving averages: the market’s memory

A moving average answers one modest question: what has the average price been over the last N periods? That is all. It does not know where price is going, it does not exert force, and it has never once held a market up. Yet it remains the most used overlay in technical analysis, because that modest question has a genuinely useful answer — it compresses the recent past into a single line, giving the market a visible memory you can compare against the present.

The trouble starts when traders promote the line from memory to oracle. So before using one, it pays to know exactly what the calculation does, where the standard settings come from, and which of the popular claims about moving averages survive contact with evidence.

SMA versus EMA: how much does the past weigh?

The simple moving average (SMA) averages the last N closes with equal weight — in a 20-period SMA, a close from three weeks ago counts exactly as much as yesterday’s. The exponential moving average (EMA) weights recent prices more heavily, with each older close counting a bit less than the one after it. The practical consequence is responsiveness: the EMA turns faster when price moves, hugging the action, while the SMA lags more but shrugs off single-bar noise. Neither is better. The EMA reacts sooner to real turns and to fake ones alike; the SMA is calmer and later to everything. Pick based on whether false signals or late signals cost your particular style more.

The lengths everyone watches

• 20 (or 21): roughly a month of daily bars; the short-term pulse, popular for tracking momentum phases.
• 50: about a quarter; the conventional dividing line for the intermediate trend.
• 100: the compromise candidate, watched mostly because it is round.
• 200: roughly a trading year; the most cited line in finance, often used as a blunt regime filter — above it conditions are treated as healthy, below it as defensive.

There is nothing physically special about any of these numbers. Their power, such as it is, comes from coordination: enough participants watch the same lines that reactions near them become somewhat more likely. That is a real effect, but it is a social one, and social effects are exactly the kind that work until they suddenly do not.

Crossovers, honestly assessed

A crossover signal fires when a fast average crosses a slow one — the 50 crossing above the 200 is the famous golden cross, the reverse the death cross. Tested over long histories, these signals do roughly what you would expect from their construction: in sustained trends they get you in and keep you in, capturing the meat of multi-month moves, and in ranges they whipsaw you to death, firing buy near the top of the chop and sell near the bottom, each signal arriving well after the move that caused it. The honest summary is that crossovers are a trend-following system with all the usual trend-following economics — a minority of winning signals paying for a majority of small losses — minus any awareness of context. They are also, by definition, late: a golden cross requires months of rising prices to have already happened.

A quick worked example shows the texture. Suppose an asset chops between 90 and 110 for six months. A 10-and-50 crossover system will buy somewhere around 104 as the upswing matures, sell around 96 as the downswing matures, and repeat — clipping 6 to 8% of loss per round trip while price ultimately goes nowhere. Then a real trend arrives, the system buys at 112, rides it to 180, and the scoreboard looks brilliant. Both halves are the same system. Whether you can hold it through the first half determines whether you ever meet the second.

Dynamic support: myth and use

You have seen the chart: price in a strong trend touching the 20 EMA four times and bouncing, as if the line were a rail. The myth is that the average causes the bounce. The reality is mostly selection and arithmetic: a steady trend has pullbacks of fairly consistent depth, and an average of matching length will, by construction, sit near where those pullbacks end. Trending charts are then shared precisely because they look that clean. Add a layer of self-fulfillment — traders who buy at the touch because they expect others to — and you get a real but fragile tendency, not a law.

The defensible use is humbler and better: treat the average as a trend gauge and a pullback region, never as a precise entry trigger. Price above a rising 50, pulling back toward it, tells you the trend is intact and the dip is normal-sized. Where exactly you enter, and where your stop goes, should come from structure — the swing low, the zone — not from the line. When price starts slicing through the average both ways, the message is not that support failed; it is that there is no trend for the average to describe.

Key takeaways

• A moving average is compressed memory of price, nothing more — descriptive, not predictive.
• SMA weights all N periods equally and lags more; EMA front-weights recent prices and reacts faster to truth and noise alike.
• The 20, 50, 100, and 200 owe their importance to crowd attention, not mathematics.
• Crossovers are late by construction and bleed in ranges; they earn their keep only in sustained trends, and only with discipline through the whipsaws.
• Dynamic support is mostly arithmetic plus self-fulfillment — use averages to read the trend and locate pullback zones, and put your stops on structure.