Moving averages
A moving average smooths the price path by averaging over recent periods. It shows direction more clearly, but follows it with a lag.
A twenty-day average takes the closing prices of the last twenty days and computes the mean. Tomorrow, the oldest value drops out and a new one comes in. Hence, moving.
The result is a smooth line that makes direction more visible than the jagged price path. The longer the period, the smoother the line and the more sluggish its reaction.
A common use is as a signal: a short average crossing above a long one is read as positive. That's a clearly defined rule, which makes it testable, unlike much else discussed about charts.
The price of smoothing is delay. A 200-day average only reacts once a move has already run its course substantially. It doesn't signal turning points, it confirms them after the fact.
The simple moving average weights every included period equally. The exponential version weights recent values more heavily via a smoothing factor, which makes it react faster but more prone to false signals. Both are linear filters, whose delay can be roughly stated as half the period length.
Crossover systems can be precisely formalized and are therefore testable. Studies over long periods and multiple markets find partly positive results for trend-following approaches of this kind, particularly in reducing maximum drawdown compared to a plain buy-and-hold position. Results, however, depend heavily on parameter choice, trading costs, and the period studied.
Behavior varies characteristically across market phases. In pronounced trends, trend-following approaches produce a few long positions with good results. In sideways phases, they produce a string of small losses through repeated false signals. Expected value therefore depends on the mix of both phases, which explains the high variability of results across different time periods.
Summary
- The delay is roughly half the period length.
- Crossover systems are clearly defined and therefore testable.
- In sideways phases, they produce many small losses.
Did you get it?
How does the exponential average differ from the simple one?
It weights recent values more heavily, reacts faster, and produces more false signals in exchange.
Why are crossover systems more testable than chart patterns?
Because they can be precisely formalized.
How do trend-following systems behave in sideways phases?
They produce a string of small losses through repeated false signals.
Related
- Chart patternsStage 3
- Momentum and trend followingStrategy Types
- Reading a chartStage 3