How to Use Moving Averages
To use a moving average, pick one length deliberately and read its slope rather than waiting for crossovers. The average lags because it smooths, and those are the same property. It works as a level only because enough people are watching the same one.
A moving average is the average price over a window, drawn late on purpose. Every argument about which type and which length is really an argument about how much lateness to accept.
Before you start
One chart with a single average on it, not three. Multiple averages make the chart look sophisticated and make the decision harder to attribute to anything.
A length chosen deliberately rather than left at a default. You should be able to say why 20 or 50 or 200, even if the reason is that other participants watch it.
A record of what the average did the last twenty times price reached it. Held, broke, or was irrelevant. This is the only evidence you will have about whether it functions as a level here.
The steps
1. Understand that lag is the feature
An average smooths by including old prices. Those old prices are what make it late. Smoothing and lag are one property described two ways.
2. Choose the length and write down why
A short window follows price closely and changes direction often. A long one is steadier and later. There is no correct answer, only the trade-off stated plainly.
3. Pick the type once and stop revisiting it
Exponential weights recent prices more heavily, so it turns earlier and also turns falsely more often. Simple is steadier and later. The difference is smaller than the argument about it.
4. Read the slope before anything else
Rising, falling, or flat. The direction of the line answers most of what people use crossovers for and answers it earlier.
5. Treat a crossover as confirmation, not as a signal
Two lagging lines agreeing is later than either one alone. If you use it, use it to confirm something you already had a reason for.
6. Use it as a level only where it has behaved as one
Check your record of the last twenty touches. If price ignored it most times, it is not a level on this instrument, whatever it does on somebody else’s chart.
7. Check participation at the touch
The calculation uses price only. Volume tells you whether a bounce off the line had anyone behind it or was drift.
8. Count the cost of every cross you act on
A round trip on the shared series is 2% of a median bar’s range. A crossover system in a range fires repeatedly and pays that toll each time.
How to tell it worked
Audit the last 20 signals your average produced, over at least 60 days.
Count how many were followed by a move larger than your round-trip cost. If fewer than half cleared 2% of a median bar’s range, the signal is generating activity rather than opportunity, and that is a conclusion about the tool rather than about your execution.
Count how many of those 20 you would have taken without the average. If most had a level or a structure reason behind them anyway, the average is decorating decisions rather than making them — which is fine, but it should be known.
Then check the 20 touches you recorded in the prerequisites. Fewer than 10 holds means it does not act as a level on this instrument, and step six should not be used here.
Why the length matters more than the type
Length changes the tool’s character; type changes its manners. Moving from 20 to 200 changes what the line is measuring entirely. Moving from simple to exponential at the same length changes when it turns by a bar or two.
The measured texture of price explains why. On the shared series the efficiency ratio has a median of 0.34, with only 30% of readings above 0.5 — most movement is back-and-forth. A short average tracks that noise faithfully; a long one ignores it. That is the decision.
And no length removes the lag. After a gap the average is stranded behind price and takes as many bars to catch up as its window is long, regardless of which weighting you chose.
The original data
Of the 24,971 unique videos in research/search-study-corpus.jsonl, 109 have an instruction-shaped
title mentioning the exponential moving average, at a median of 18,862 views across 85 channels, with
a maximum of 1,653,330. The general term appears in 105 at a median of just 1,681. The counts come
from site/rank_howto.py.
109 videos on the exponential version at 18,862, against 105 on moving averages generally at 1,681. Near-identical supply, and an eleven-fold difference in median views. The specific named term is what people search; the generic category is what creators title their videos, and the gap between those two habits is one of the clearest patterns in this corpus.
The answer to the question on that chart is that the cross is the latest possible version of the information. The slope turned first, and the level it crossed at was either somewhere you had marked or it was not. A crossover with no level behind it is two averages agreeing about a move that has already been paid for.
When it fails
A range is where averages do their worst work and produce their most signals. Price oscillates around the line, the average flattens, and crossovers fire in both directions every few bars. Each one costs a round trip and none of them leads anywhere. It is the classic failure and it is not fixable by changing the length — a shorter one crosses more often and a longer one crosses later, and both lose in a market that is not going anywhere.
The second failure is treating the crossover as the primary signal. It is the most delayed thing the tool produces.
A third is using it as a level without evidence. On many instruments price passes straight through.
A fourth is stacking three averages. It multiplies signals without adding information.
A fifth is arguing about simple versus exponential. The length matters far more.
And a sixth is expecting it to help after a gap. It is furthest from price exactly when you most want guidance.
Related
Exponential moving average explains the weighting and where the earlier response comes from. Moving average covers the general tool and its variants. And trend following is the approach these are usually a component of rather than the whole of.
The thing that changed my use of these was reading the slope instead of waiting for the cross. A crossover is two averages agreeing about something that already happened, which puts it very late. The slope of a single average turns earlier and says most of the same thing, and dropping the second line removed a signal I had been treating as more meaningful than it was.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.