Moving averages

A simple moving average is the mean closing price over a fixed number of recent sessions, recalculated each day as the newest close enters and the oldest drops out. Its purpose is to remove daily noise so the direction underneath becomes visible.

It lags by construction, and the lag is not a defect to be tuned away. An average of the last two hundred days describes the last two hundred days, so it turns after the price has turned, and roughly half the window has to change direction before the line does. Every advantage and every complaint attached to moving averages follows from that one property.

An exponential moving average weights recent prices more heavily, so it responds faster at the same length. It lags less and whipsaws more, which is the same trade in the other direction.

The charts here offer the 50-day and 200-day simple moving averages. The 50 covers roughly a quarter of trading and responds within weeks; the 200 covers about ten months and moves slowly enough to read as a trend rather than as a signal.

Neither length is special. They are conventions that became self-reinforcing because enough participants watch them, which is a real effect on behaviour but a different thing from the averages carrying predictive content of their own. A price crossing an average states where it sits relative to its own recent history.

The 200-day is the more commonly cited of the two, and its main use is as a coarse regime marker: a price persistently above it describes one kind of period and persistently below it another. Neither is a forecast.

A golden cross is the 50-day average rising through the 200-day; the reverse is called a death cross. Both are widely reported and the names carry more confidence than the evidence behind them.

Both signals are derived from averages that lag by definition, so the crossing confirms a move that has already happened, typically weeks after it began. In sideways markets the two averages cross repeatedly, producing signals in both directions within weeks and no trend in either.

They are worth knowing about mainly because other people watch them, which is an argument about flows rather than about the company.

The volume-weighted average price is the average price paid across a session, weighted by the volume traded at each level, rather than an average of closing prices. A session where most of the volume traded at $50 and a thin tail traded at $55 has a VWAP near $50.

That makes it a benchmark for execution: a desk filling a large order is measured against whether it beat the day's VWAP. It resets each session, so it describes one day rather than a trend, and reading it as a slow moving average misapplies it.

Bollinger Bands draw a 20-period simple moving average with a line two standard deviations above it and another two below. The charts here use those settings, which are the ones John Bollinger published and the default almost everywhere.

The width of the band is the useful part. It widens when prices have been swinging and narrows when they have been calm, so it is a measure of recent volatility drawn in the same units as the price. A narrow band says the last month has been quiet, not that a large move is due.

A close outside a band is less unusual than it sounds. Two standard deviations would contain about 95% of observations if returns were normally distributed, and they are not: prices trend, and a strong trend can walk along one band for weeks. Touching the upper band describes a price high relative to its own recent range, which is a different statement from overbought.

MACD subtracts a slower exponential moving average from a faster one, twelve and twenty-six periods on the charts here, and plots a nine-period average of that difference as a signal line. The histogram is the gap between the two lines.

Because it is a difference between two averages, it measures whether momentum is building or fading rather than where the price is. A rising MACD with a falling price means the decline is slowing, not that it has ended.

It inherits every limitation of its inputs. It lags, it whipsaws in range-bound conditions, and it knows nothing about the company. Its more defensible use is divergence, where price makes a new extreme that the indicator does not confirm, rather than the line crossings, which are frequent.

Treating a crossing as a cause. An average is arithmetic performed on prices that have already happened, and nothing about the crossing itself moves a market.

Optimising the period against past data. Any length can be made to look good on a chosen sample, and the result rarely survives the next one.

Comparing signals across intervals. A 50-period average on an hourly chart and on a daily chart describe different lengths of time.

Using a price-derived indicator through an earnings report or an acquisition, when the price is moving on information the indicator cannot see.

Reading VWAP as a trend line, or comparing it across sessions, when it resets every day.