EMA vs Weighted Moving Average
The exponential moving average weights every past bar, with the weight decaying as bars get older. The weighted moving average weights only the bars inside its window, in a straight line. In practice the two lines sit almost on top of each other.
Two averages that both give recent bars more weight. They are frequently presented as a meaningful choice. Plot them together on any chart and the honest observation is that they are nearly the same line.
What each one is
The exponential moving average weights every past bar, with the weight falling off as bars get older. Nothing ever drops out of the calculation entirely. The exponential moving average covers it.
The weighted moving average weights only the bars in its window, in a straight line from most to least recent, and everything older is excluded. The weighted moving average covers that version.
Both solve the same complaint. An equally weighted average treats a bar from three weeks ago as though it mattered as much as yesterday’s, and both of these fix that in slightly different ways.
Where they differ
What happens to old bars. One never fully discards them; the other drops them the moment they leave the window. In practice that produces a very small difference in the line.
How the line behaves when a large bar leaves the window. The weighted version can shift slightly as an old extreme drops out. The exponential one has no window to drop out of.
How widely each is supported. The exponential version is on every platform and in every strategy description. The weighted one is available but far less commonly discussed.
What each is used for. In practice, exactly the same jobs — a trend filter, a dynamic level, or one half of a crossover.
Where they agree
They agree almost always. At any length you would sensibly use, the two lines track each other closely enough that no method would produce different trades.
Both lag. Weighting recent bars more heavily reduces lag; it does not remove it, and neither can turn before price does.
Both fail in a range. On this site’s shared series direction runs average 2.01 bars with a longest of 11, and short runs produce false crosses in either.
And both cost a round trip per signal acted on — about 2% of the median bar range of 0.493 here — identically, because they produce nearly identical signals.
Which one to use
Run the exponential average. It is supported everywhere, every description of every strategy assumes it, and the behaviour you give up by not running the other one is not measurable in your results.
Run the weighted one when your platform already uses it and you have no reason to switch. Changing it would not improve anything, and the cost of changing a working setup is real.
Spend the attention on the length instead. Moving from a short line to a long one changes the behaviour enormously; moving between these two formulas changes it barely at all.
And when somebody claims one is materially faster, plot both. That check takes a minute and settles it more convincingly than any argument about the formulas.
Why the difference is so small
Because both put most of their weight in the same place. The recent bars dominate either calculation, and the bars where they disagree carry very little weight in both.
And because the shared input dominates. Two functions of the same closes, both emphasising the recent ones, cannot produce very different answers.
What actually changes the line
The length, by a wide margin. A ten-period line and a fifty-period line are different tools; a ten-period exponential and a ten-period weighted line are the same tool twice.
The timeframe you plot it on. The same length on two chart intervals produces two completely different reference lines, and this is the choice people examine least.
The price the average is taken of. Closes, midpoints or typical price all give different lines, and that setting sits in the same panel as the length on most platforms.
And whether you use it as a filter or a signal. The same line used to decide direction and used to trigger entries produces two entirely different trade counts, which is a bigger effect than anything on this page.
The original data
Of the 24,971 unique videos in research/search-study-corpus.jsonl, no title compares these two
directly — the pair is constructed from two subjects the corpus covers separately. Separately, the
weighted moving average appears in just 5 titles at a median of 4,406 across 5 channels. The counts
come from site/corpus_count.py.
5 videos on the weighted version across 5 channels. Almost nothing, against hundreds on the exponential one — which is itself the practical argument, because a tool nobody discusses is a tool you will be working out alone.
The answer to the question on that chart is no. Switching between two lines that overlap costs a round trip and changes nothing — and the time is better spent on the length, which changes everything.
When it fails
The failure is switching between the two after a losing run, and it is pure motion. A method using one of them has a bad stretch, so the other is tried on the grounds that it weights bars differently. It produces nearly identical signals, so the next stretch is decided by the market rather than by the change. The record now spans two variants and cannot be read as one sample, and nothing was learnt about either.
The second failure is arguing about the formulas. Plot both and the argument ends.
A third is optimising the type instead of the length. The length is the lever.
A fourth is adopting the rarer one without support. You will be alone with it.
A fifth is expecting either to lead price. Both average bars that closed.
And a sixth is running both. They are the same line drawn twice.
Related
The exponential moving average covers the decaying-weight version. The weighted moving average covers the linear-weight one. And moving average covers the family and the choice that actually matters.
This is the comparison where the answer is that it does not matter much, and saying so is more useful than manufacturing a distinction. The length you pick changes the line enormously. Which of these two formulas produced it changes it barely at all.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.