Study · run 2026-11-05 · registered before the run · the Rule v1 study

Supertrend vs a moving average on crypto: a pre-registered comparison

Supertrend(10, 3) and a simple moving average were put through the same test on the 20 most-traded crypto assets, 2019-01-01 to 2026-10-09: the same data, the same basket rebuilt each quarter, the same costs of 0.30% per side, both long or cash. The rules, the lengths and the three sentences one of which would be published were fixed before any moving average had been run on this data. The result is below, with the numbers of both rules beside it.

Pre-registered outcome · ladder N

In past data Supertrend(10, 3) and the closest-matched moving average were not clearly different in how far they fell. Supertrend's return was higher.

Worst fall -46.1% for Supertrend against -73.5% for the 250-day moving average. Compound return 42.5% a year against 13.8%. Holding the same basket: -85.1% and 38.5%. This describes the past. It is not a forecast and not advice.

Registered caveat: matched at the edge of the grid. No moving average on the grid traded as rarely as Supertrend. The closest, 250 days, still made 871 round trips against 682.

Wording amended (deviation D2, 2026-11-05). As registered, the sentence read: “In past data Supertrend(10, 3) and a moving average that traded as often were not clearly different in how far they fell.” With the match on the edge of the grid, that phrase claimed a match the data does not show. The replacement claims less. The outcome, the conditions and every number are unchanged.

Headline numbers

Supertrend(10, 3)SMA 250Hold the basket
Compound return per year42.5%13.8%38.5%
Worst fall from a peak-46.1%-73.5%-85.1%
Ulcer index (pain of drawdowns)26.6957.1359.01
Longest time under water922 days1,980 days1,980 days
Sharpe1.040.520.82

Growth of 1 (log scale)

Supertrend(10, 3)SMA 250Hold the basket

Drawdown from the last peak

Supertrend(10, 3)SMA 250Hold the basket

Why "not clearly different", with falls that far apart

The two worst falls are 27.3 points apart, and the sentence still says the difference is not clear. That is the test working as registered. A worst fall is a single event in 7.78 years, and one event can be luck. So the registration set four conditions, all of which had to hold before the difference could be called clear:

  • the worst falls at least 5 points apart: yes
  • a resampling interval for the difference that excludes zero: no
  • the same direction in both halves of the period: yes
  • the same direction at double the costs: yes

The resampling is where it fell short. The period was redrawn 2,000 times in blocks of 10 to 20 days and both rules were measured again each time. The gap between their worst falls came out anywhere from 6.5 points in the moving average's favour to 36.7 points in Supertrend's, 15.1 points on average. A range that includes zero does not rule out chance, so the registered sentence is the cautious one.

For the return, the same resampling gave a gap of 1.6% to 59.2% a year in Supertrend's favour, a range that excludes zero. That is why the second sentence names it.

How the moving average was chosen

A fast rule and a slow rule differ mostly in how often they trade. To compare where the line is drawn rather than how often it is crossed, the headline comparator was fixed in advance as the moving average whose number of round trips came closest to Supertrend's, on a grid of lengths from 10 to 250 days. Only trade counts entered the choice; no return or fall did. The rule itself is the plainest one: long when the daily close is above its simple average of the last N closes, otherwise cash.

Round trips by length of the average · Supertrend made 682
10 d · 5,018
20 d · 3,217
30 d · 2,531
40 d · 2,078
50 d · 1,833
60 d · 1,609
70 d · 1,506
80 d · 1,428
90 d · 1,359
100 d · 1,343
110 d · 1,275
120 d · 1,197
130 d · 1,181
140 d · 1,114
150 d · 1,100
160 d · 1,025
170 d · 1,002
180 d · 983
190 d · 947
200 d · 926
210 d · 956
220 d · 968
230 d · 953
240 d · 914
250 d · 871

In this data price crossed every moving average on the grid more often than it crossed the Supertrend line. Even the 250-day average made 871 round trips, and 68% of them were lost within ten days, against 12% for Supertrend: when price sits near its average it steps over and back, and each step is a trade.

What each was like to follow

Supertrend(10, 3)SMA 250
Round trips682871
Share that lost money58%89%
Average win49%109%
Average loss-14%-7%
Lost within ten days12%68%
Longest run of losing trips1764
Average days held3727
Profit factor2.462.01
Days with at least one position95%95%

The two rules held the same position on 67% of 56,819 asset-days in the basket: both in on 29%, both out on 37%. The differences above come from the remaining 33%.

The big falls

Bitcoin peak to troughBasket fellFall avoided: SupertrendSMA 250Rebound missed: SupertrendSMA 250
2019-06-26 → 2020-03-12-54%95%-5%72%80%
2021-04-13 → 2021-07-20-42%41%-5%49%75%
2021-11-08 → 2022-11-21-81%49%63%46%72%
2025-10-06 → 2026-06-30-50%69%50%38%61%

The study's four episodes: each Bitcoin drawdown of at least 30%. "Fall avoided" is the share of the basket's fall the rule did not take; a negative figure means it fell a little more than the basket. "Rebound missed" is the share of the recovery it did not take.

The conventional lengths, in full

The lengths people usually mean by "the moving average" were fixed in advance and are all shown. This is a disclosure, not a search: no row is offered as the length to use, and picking a row after seeing the results is what a registration exists to prevent.

RuleRound tripsReturn per yearWorst fallUlcerTrips that lost
Supertrend(10, 3)68242.5%-46.1%26.6958%
SMA 203,21732.5%-59.4%30.5976%
SMA 501,83337.6%-48.8%28.6179%
SMA 1001,34320.7%-78.0%57.1987%
SMA 20092613.9%-76.2%59.5588%
SMA 50/200 crossover18014.4%-72.4%56.3772%

SMA N: long when the close is above its N-day simple average. The crossover: long when the 50-day average is above the 200-day average, which is the golden cross and death cross; that row is read on its own in the guide to the golden cross.

Does it hold up?

  • Double the costs (0.60% per side): worst fall -48.0% against -77.9%; return 38.7% against 10.0%.
  • Acting a day later: worst fall -49.1% against -66.9%; return 41.5% against 20.0%.
  • The first half (from 2019-01-01): worst fall -43.6% against -64.7%; return 69.2% against 28.6%; holding the basket -84.6% and 49.1%.
  • The second half (from 2022-11-20): worst fall -45.9% against -53.6%; return 20.1% against 0.8%; holding the basket -58.0% and 28.7%.
  • The neighbouring length, SMA 240: 914 round trips, worst fall -75.2%, return 13.0%. The matched length is not an isolated result.

What this does and does not show

It shows what two rules did on one market over one period, under one definition of a fair comparison. It does not show which will fall less in future, and it says nothing about moving averages or Supertrend in general: exponential and other averages were not tested, other markets were not tested, and matching by trade count is one choice among several. Crypto history is short with few full cycles. The assets move together, so the effective sample is far smaller than the asset count. Assets that delisted are not in the data. Real execution, fees and discipline differ from a simulation.

How it was registered, and the files

The protocol and the code were committed before the run; the run recorded that commit (f71db9e) and a clean working tree. Before producing anything, the harness had to reproduce the published Rule v1 study on its frozen data exactly, and did: 2019-01-01 to 2026-10-09, 2,839 days, 42.5% a year, a worst fall of -46.1%, 682 round trips. Two deviations are logged. The registration called the code repository public, and it is private. And the outcome sentence was amended to claim less, as shown under the result. The registration is published here as it was registered.

The pre-registration · Deviations and notes · Full results (JSON) · Technical report (Markdown) · snapshot fd1d67415b8f

Questions people ask

Is Supertrend better than a moving average?
This comparison was registered so that it could not answer in those words, and one market over one period cannot support them. What it found on the 20 most-traded crypto assets, 2019 to 2026: Supertrend(10, 3) and the closest-matched moving average, 250 days, were not clearly different in how far they fell, and Supertrend's return was higher.
How did Supertrend and the moving average compare in numbers?
Supertrend(10, 3) had a worst fall of 46% and returned 42.5% a year. The 250-day moving average had a worst fall of 73% and returned 13.8% a year. Holding the same basket had a worst fall of 85% and returned 38.5% a year. Same data, same basket, same costs, past data only.
Why is the result "not clearly different" when the worst falls were so far apart?
Because a worst fall is one event, and one event can be luck. When the period was resampled 2,000 times, the gap between the two worst falls ranged from 6.5 points in the moving average's favour to 36.7 points in Supertrend's. The test fixed in advance required that range to exclude zero, and it does not.
Which moving average was used, and why that one?
The simple moving average whose number of trades came closest to Supertrend's, chosen by trade count alone on a grid of 10 to 250 days. No length traded as rarely as Supertrend: the closest, 250 days, made 871 round trips against 682. The rule is long when the daily close is above the average, otherwise cash.
How did the 50-day and 200-day moving averages do?
Both are shown as a disclosure, not as lengths to use. In this data the 50-day average made 1,833 round trips, returned 37.6% a year and had a worst fall of 49%. The 200-day average made 926 round trips, returned 13.9% a year and had a worst fall of 76%.
Was the comparison decided before the results were known?
Yes. The rules, the lengths, the way the comparator would be chosen and the three sentences one of which would be published were written down and committed before any moving average was run on the data. The harness then had to reproduce the published Supertrend study exactly before it produced anything. The registration and the full output are published with the page.

TideTrade is decision support, not financial advice. Cryptoassets are high-risk; trend rules lose money in sideways markets and past results do not predict future prices. See the disclaimer. Every figure on this page is read from the comparison's published output.