Hull vs Weighted Moving Average
The weighted moving average scales each bar in its window by age, so recent price dominates. The hull moving average is built from weighted averages — two of them differenced, then smoothed again — specifically to cut the lag a single weighted line still carries.
The hull is not an alternative to the weighted average in the way two indicators usually are alternatives. It is assembled from weighted averages. So the question is narrower and more answerable than most comparisons: does the extra construction buy enough to be worth what it costs?
What each one is
The weighted moving average scales each bar by its age. The newest close carries the most weight and the oldest the least, in a straight line. The weighted moving average covers it.
The hull takes weighted averages and differences them. A short one, a longer one, a subtraction that overshoots deliberately, and a final smoothing over the result. The hull moving average covers the construction step by step.
So one contains the other. Whereas most indicator comparisons put two different measurements side by side, this one puts a measurement next to a thing built out of it, and the differences are all consequences of the assembly.
Where they differ
How much lag each carries. The weighted line is faster than a simple average and still late. The hull is faster again, and on a clean move it turns near the start rather than after it.
Whether the line can leave the price range. This is the real difference. Subtracting one average from another can produce a value above every close in the window, so the hull can point somewhere price has not been. A weighted average is a weighted mean of actual closes and mathematically cannot do that.
How many turns each produces in noise. The differencing that removes lag also amplifies whatever is in the recent bars, so the hull turns more often. In a trend those extra turns are mostly harmless; in a range they are the whole output.
How easy each is to explain. You can describe a weighted average in one sentence to somebody who has never used one. The hull needs three steps and a reason for each, and an indicator you cannot explain is one you cannot reason about when it disagrees with you.
Where they agree
Both are built from the same closes, over the same kind of window, and neither leads price in any sense that survives inspection.
Both fail in a range, differently. Direction runs on this site’s shared series average 2.01 bars with a longest of 11. The weighted line goes flat; the hull whips.
Both cost a round trip per signal acted on — 0.0098 here, about 2% of the median bar range of 0.493. The hull produces more signals, so it pays that more often.
And neither is a stop. The ninetieth percentile bar range on this series is 1.101 and the largest bar spanned 2.338, so a stop resting on either line is inside a single ordinary bar’s reach.
Which one to use
Run the weighted average when you want speed you can still describe. It fixes the worst artefact of equal weighting, it never leaves the price range, and it is the default answer unless you can name the specific entries the remaining lag is costing you.
Run the hull when you have counted what the lag costs and it is more than the extra turns. That is a countable thing rather than a preference: how many bars late is the weighted line at your reversals, and how many additional turns does the hull produce on the same chart?
Run the hull when your instrument trends cleanly and rarely ranges. The overshoot is a problem in chop and close to irrelevant in a sustained move.
And when neither answer is obvious, take the weighted one. The simpler tool is the correct default, whereas reaching for the more elaborate one because it looked better on a chart you already knew the outcome of is how most people end up here.
Why the overshoot is structural
Because a difference of two averages is not an average. The moment you subtract, the result stops being a weighted mean of the closes and stops being bounded by them. That is precisely the mechanism that removes the lag, so you cannot keep the speed and discard the overshoot.
And it gets worse where the data is worst. In thin or gappy conditions the recent bars are less reliable, and the construction weights exactly those bars hardest.
The original data
Of the 24,971 unique videos in the search corpus, exactly one title pairs these two — and it has 38 views. Separately the hull appears in 83 titles at a median of 1,923 views across 67 channels, and the weighted average in 5 titles at a median of 4,406 across 5.
83 videos at a 1,923 median is a lot of teaching for a small audience. The hull is recommended far more often than it is watched, and the weighted average — five videos — is barely taught at all despite being the thing the hull is made from.
On the chart above, the line being outside the price range is the warning. A hull reading no close in the window supports is the overshoot, not an early call, and acting on it is acting on the artefact.
When it fails
The characteristic failure is running both and calling it confirmation. The hull is made out of weighted averages, so when the two agree you have learned nothing you did not already have. Worse, the moments they disagree are exactly the moments the hull has overshot, which means the pair is quiet precisely when you needed a warning and vocal precisely when the fast line is least trustworthy. Two tools that share their entire input cannot check each other, whereas a volume measure or a structure read genuinely can.
A second failure is adopting the hull from a backtest of a trending period. The overshoot costs nothing in a sustained move, so any sample without chop in it flatters the hull enormously.
A third is placing stops at the hull line. It can sit outside the price range entirely.
A fourth is tuning the length to remove the overshoot. The overshoot comes from the differencing, not the period, so a longer hull is simply a slower hull with the same artefact.
And a fifth is switching between the two after losses, which turns one record into two half-samples and answers nothing.
Related
The hull moving average covers the construction and the overshoot. The weighted moving average covers linear weighting on its own. And moving average covers the family and what the weighting choice sets.
The thing nobody says about the hull is that it can go somewhere price never went. That is not a bug in an implementation, it falls straight out of subtracting one average from another, and it is the honest cost of the lag reduction rather than a flaw you can tune away.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.