Markets › Energy · WTI crude oil spot (EIA via FRED), daily
Crude Oil Trend Following: 200-Day Average and 12-Month Momentum on 40 Years of WTI
Crude oil is the market best known for long trends. We backtested the two most common trend-following rules on the daily WTI price from 1986 to 2026, long only and long and short, with costs and against the same positions at random times.
Buried · FragileA few great trends paid for everything. Random timing did as well.
Variants tried: 4 · Passed every gate: 0 · Best: 2 of 6 · Holdout opened: yes (fresh)
Short answer: not better than chance in our test. On daily WTI prices from 1986 to 2026, the 200-day moving average and 12-month momentum made money, but the same positions moved to random times did as well (the real rules ranked between the 36th and 53th percentile), and a few long trends made all the profit. All four versions were buried.
00The verdict in three lines
- Does trend following work on crude oil? Both rules made money over 40 years, but no better than the same positions at random times: they ranked between the 36th and 53th percentile of the random runs. All four versions were buried.
- A few trends carried it. The long-only 200-day rule won only 30.8% of its trades; three long trends (1999 to 2000, 2007 to 2008, 2020 to 2021) made most of the profit, and without its 10 best trades every version lost.
- Shorting the collapses did not pay overall. It made money in 2008, 2014 to 2016 and 2020, and gave it back in the years without a trend.
01Trend following on crude oil
Commodity funds have followed trends in crude oil for decades, and oil has some of the biggest trends of any market: from about $11 in 1998 to $145 in 2008, and from $108 in 2014 to $26 in 2016. We test the two simplest and most widely used rules, in their usual form, and change nothing after the run:
- The 200-day moving average: long while oil closes above the average of its last 200 closes; below it, either flat (MA200-L) or short (MA200-LS). It is the same rule as in our 200-day test on SPY.
- 12-month time-series momentum: at each month end, long if oil is above its price of a year earlier, otherwise flat (TSM-L) or short (TSM-LS). Tobias Moskowitz, Yao Hua Ooi and Lasse Pedersen found this effect in crude oil and 57 other futures markets (Time Series Momentum, Journal of Financial Economics, 2012).
This is the first crude oil test on the site. Trend rules have been buried on most markets we tested; we chose oil because it is where they should work best, if anywhere.
02How we tested it
- Data: the daily WTI spot price at Cushing, Oklahoma, from the US Energy Information Administration via FRED (DCOILWTICO), 2 January 1986 to 30 October 2026. Days without a price (about one weekday in nine, in short gaps) are skipped, never filled. The negative price of 20 April 2020 (−$36.98) is kept; each trade's result uses only its entry and exit prices.
- Fills: the file has one close a day, so a signal at a close is traded at the next day's close. No stop and no target.
- A spot price, not a product. Futures and oil ETFs also earn or pay the roll between contracts, which is not included here; for a long position in contango that roll is a cost.
- Costs: 0.05% of the price per round trip, about a tick of slippage on each side and a commission on a crude oil future; 0.15% shown too. Unit: % per trade of the entry price.
- Periods: 1986 to 1999, 2000 to 2012, and a holdout from 2013 that no test on the site had used (fresh). First-period bar t ≥ 2.5, because trend rules are the most tested family on the site.
- Control: each version's real positions, shifted as a whole by a random amount of time (whole months for momentum), 500 times. It keeps the number of trades, their lengths and the long and short mix, so oil's rise is in the control too. Engine check: 28 checks, an independent re-implementation that matched every trade, and 3 deliberate rule changes the test had to catch (wti_selftest_result.txt).
03Results: does trend following work on crude oil?
| Variant | Rule | Trades | Win rate | Net per trade | Sum of trades | Profit factor | Time in the market | Verdict |
|---|---|---|---|---|---|---|---|---|
| MA200-L | 200-day average, long only | 159 | 30.8% | +1.40% | +222.8% | 1.58 | 55.1% | Buried |
| MA200-LS | 200-day average, long and short | 317 | 27.8% | +0.22% | +71.1% | 1.09 | 100.0% | Buried |
| TSM-L | 12-month momentum, long only | 34 | 41.2% | +4.28% | +145.5% | 2.37 | 54.8% | Buried |
| TSM-LS | 12-month momentum, long and short | 68 | 39.7% | +0.46% | +31.5% | 1.10 | 100.0% | Buried |
| Variant | 1986 to 1999 (trades) | 2000 to 2012 (trades) | Holdout 2013 to Oct 2026 (trades) | t, first period | PF, second period | PF, holdout |
|---|---|---|---|---|---|---|
| MA200-L | +129.3% (40) | +88.3% (42) | +5.1% (77) | 0.98 | 1.88 | 1.03 |
| MA200-LS | +129.0% (79) | +18.7% (85) | -76.6% (153) | 0.94 | 1.08 | 0.81 |
| TSM-L | +46.6% (13) | +72.0% (11) | +26.9% (10) | 0.91 | 2.58 | 1.84 |
| TSM-LS | +39.4% (26) | -11.2% (23) | +3.3% (19) | 0.69 | 0.92 | 1.03 |
Summed net % per trade, by month, open positions marked at each close. The holdout years had not been used by any test on the site.
04Trend, or random timing?
| Variant | Trades | Real net per trade | Shifted, median | Shifted, 5th to 95th | Percentile (95 needed) |
|---|---|---|---|---|---|
| MA200-L | 159 | +1.40% | +1.61% | +0.16% to +2.81% | 40.0 |
| MA200-LS | 317 | +0.22% | +0.17% | -1.16% to +1.25% | 53.2 |
| TSM-L | 34 | +4.28% | +6.03% | -0.21% to +14.20% | 36.4 |
| TSM-LS | 68 | +0.46% | +0.44% | -6.36% to +7.31% | 50.4 |
The same long and short periods, moved to other times, made about as much: for the long-only versions the median random run made more per trade than the real rule. Oil rose from about $26 to $108 a barrel over these years, so being long half of the time made money whenever it was. The rules did not choose those times better than chance.
05A few trends paid for everything
| Variant | Sum of trades | 10 best trades | Without the 10 best | Without the 20 best | Win rate | Median days held | Longest |
|---|---|---|---|---|---|---|---|
| MA200-L | +222.8% | +457.4% | -234.6% | -339.4% | 30.8% | 4 | 328 |
| MA200-LS | +71.1% | +512.6% | -441.5% | -618.2% | 27.8% | 4 | 415 |
| TSM-L | +145.5% | +236.1% | -90.6% | -100.1% | 41.2% | 105 | 523 |
| TSM-LS | +31.5% | +288.2% | -256.7% | -315.3% | 39.7% | 84 | 535 |
The long-only 200-day rule's three best trades were the long trends of 1999 to 2000 (+119.4%), 2007 to 2008 (+68.4%) and 2020 to 2021 (+67.9%). Most of its other trades were short whipsaws: the median trade lasted 4 days with a price. That is what trend following looks like, but here the big trends were not enough to beat random timing.
06Long, short and the great collapses
| Variant | Long trades | Long, summed | Short trades | Short, summed |
|---|---|---|---|---|
| MA200-L | 159 | +222.8% | – | – |
| MA200-LS | 159 | +222.8% | 158 | -151.6% |
| TSM-L | 34 | +145.5% | – | – |
| TSM-LS | 34 | +145.5% | 34 | -114.0% |
| Episode | Oil spot | MA200-L | MA200-LS | TSM-L | TSM-LS |
|---|---|---|---|---|---|
| 1990-91 Gulf War spike and fall | +14% | +11 pts | +0 pts | -1 pts | -16 pts |
| 2008 boom and bust | -55% | +14 pts | +74 pts | -47 pts | -16 pts |
| 2014-16 collapse | -76% | -10 pts | +63 pts | -10 pts | +64 pts |
| 2020 crash, negative price | -80% | -13 pts | +64 pts | -24 pts | +52 pts |
| 2022 spike | +42% | +44 pts | +44 pts | +59 pts | +59 pts |
Going short below the 200-day average worked in the collapses: +74 pts in 2008, +63 pts from 2014 to 2016 and +64 pts in 2020, when oil briefly traded below zero. It lost it in years without a trend: -54.6% in 2003, -68.7% in 2019 and -35.1% in 2024. Over 40 years its short trades summed -151.6%.
07How solid is it?
| Variant | Net per trade | At 0.15% cost | Filled at the signal close | In dollars per barrel, summed | Worst open loss at a close (open positions included) |
|---|---|---|---|---|---|
| MA200-L | +1.40% | +1.30% | +1.43% | +$55.34 | -27.1% |
| MA200-LS | +0.22% | +0.12% | +0.26% | +$18.80 | -27.8% |
| TSM-L | +4.28% | +4.18% | +4.81% | +$37.40 | -31.7% |
| TSM-LS | +0.46% | +0.36% | +1.00% | -$14.15 | -42.4% |
Costs and the fill day change little; these rules trade rarely. The dollar column weighs the expensive years more, which is why it differs from the percentages.
08By year
| Year | MA200-L (trades) | MA200-LS (trades) | TSM-L (trades) | TSM-LS (trades) |
|---|---|---|---|---|
| 1986 | +21.9% (1) | +21.9% (1) | – | – |
| 1987 | – | +4.0% (1) | +2.1% (1) | +8.4% (2) |
| 1988 | +12.0% (3) | +24.4% (5) | – | +2.1% (1) |
| 1989 | +4.3% (6) | -3.8% (12) | +7.9% (1) | +7.9% (1) |
| 1990 | +34.5% (4) | +34.1% (8) | -1.3% (1) | -16.4% (2) |
| 1991 | -34.5% (3) | -47.7% (7) | +4.6% (1) | +5.7% (3) |
| 1992 | -2.3% (3) | +10.0% (6) | -1.6% (1) | +10.7% (2) |
| 1993 | – | – | +8.1% (1) | +13.2% (2) |
| 1994 | -13.2% (5) | -25.2% (9) | -17.7% (2) | -24.6% (3) |
| 1995 | -17.1% (8) | -38.0% (16) | -6.7% (2) | -15.4% (4) |
| 1996 | +14.0% (3) | +4.4% (6) | +5.5% (1) | -3.7% (2) |
| 1997 | -5.6% (1) | +20.6% (3) | -1.2% (1) | +4.5% (3) |
| 1998 | -4.0% (2) | +5.0% (4) | – | – |
| 1999 | +119.4% (1) | +119.4% (1) | +47.0% (1) | +47.0% (1) |
| 2000 | – | -6.6% (1) | – | – |
| 2001 | -24.2% (4) | -21.5% (8) | -0.9% (1) | +1.7% (3) |
| 2002 | +15.3% (5) | +9.8% (9) | -2.8% (1) | -2.8% (1) |
| 2003 | -5.8% (10) | -54.6% (20) | +13.1% (2) | -11.4% (4) |
| 2004 | +7.5% (2) | +1.1% (4) | +77.7% (1) | +84.0% (2) |
| 2005 | +19.2% (1) | +14.6% (2) | – | – |
| 2006 | +4.8% (2) | -2.9% (5) | -9.6% (1) | -13.7% (2) |
| 2007 | +58.8% (5) | +50.5% (9) | -10.0% (2) | -34.0% (4) |
| 2008 | -1.0% (1) | +39.4% (3) | – | -22.2% (1) |
| 2009 | +18.1% (1) | +18.1% (1) | -1.0% (1) | -1.0% (1) |
| 2010 | -4.6% (6) | -22.4% (12) | – | – |
| 2011 | +2.1% (3) | -4.7% (6) | +14.1% (1) | -5.1% (2) |
| 2012 | -2.0% (2) | -2.3% (5) | -8.7% (1) | -6.7% (3) |
| 2013 | -6.9% (8) | -16.3% (15) | +3.2% (1) | +3.2% (1) |
| 2014 | -9.1% (7) | +35.9% (15) | -7.2% (1) | +33.7% (3) |
| 2016 | +2.2% (4) | -10.9% (7) | -1.0% (1) | -1.0% (1) |
| 2017 | +21.1% (9) | +4.9% (18) | +7.7% (1) | +5.8% (2) |
| 2018 | -1.3% (1) | +3.6% (3) | – | -5.7% (1) |
| 2019 | -39.5% (11) | -68.7% (21) | -10.6% (1) | -10.6% (1) |
| 2020 | +34.7% (6) | +30.1% (12) | – | -7.0% (1) |
| 2021 | +23.4% (2) | +10.2% (4) | +42.5% (1) | +42.5% (1) |
| 2022 | -7.4% (3) | -2.2% (7) | – | – |
| 2023 | -14.5% (3) | -28.0% (6) | -8.8% (1) | -25.0% (3) |
| 2024 | -16.4% (12) | -35.1% (24) | -4.0% (1) | +1.1% (2) |
| 2025 | -20.8% (5) | -21.8% (10) | -0.4% (1) | -39.1% (2) |
| 2026 (to Oct) | +39.4% (6) | +21.8% (11) | +5.5% (1) | +5.5% (1) |
09Verdict by variant
| Variant | Cause | Gates failed | Verdict |
|---|---|---|---|
| MA200-L | Fragile | 2, 3, 4, 5 | Buried |
| MA200-LS | Fragile | 1, 3, 4, 5, 6 | Buried |
| TSM-L | Fragile | 2, 3, 4, 5 | Buried |
| TSM-LS | Fragile | 1, 2, 3, 4, 5 | Buried |
| Variant | Trades | 1. PF ≥ 1.2 in validation | 2. ≥ 50 trades in each period | 3. t ≥ 2.5 on development | 4. Above the 95th percentile of the shifted holdings | 5. Positive without the 10 best trades | 6. Holdout net positive (fresh) |
|---|---|---|---|---|---|---|---|
| MA200-L | 159 | pass | fail | fail | fail | fail | pass |
| MA200-LS | 317 | fail | pass | fail | fail | fail | fail |
| TSM-L | 34 | pass | fail | fail | fail | fail | pass |
| TSM-LS | 68 | fail | fail | fail | fail | fail | pass |
Fragile: positive overall, but negative without its 10 best trades. The momentum versions also had too few trades for our rules (34 and 68 in 40 years), and the long-only 200-day rule too few in its first two periods.
TWhat this shows, and what to take into your own tests
What the results show
- The 200-day moving average and 12-month momentum both made money on crude oil from 1986 to 2026 (the long-only 200-day rule summed +222.8% over 159 trades), but the same positions moved to random times did as well: the real rules ranked between the 36th and 53th percentile, against the 95th we require.
- A handful of long trends paid for everything: without its 10 best trades every version lost. Only about 31% of the long-only 200-day trades won.
- Going short in the downtrend helped in the big collapses (2008, 2014 to 2016, 2020) but lost more in the years without a trend: the shorts of the 200-day rule summed -151.6%.
What remains unknown
- Crude oil futures or oil ETFs, which also earn or pay the roll between contracts. We used the spot price, which leaves that out.
- Other trend rules (breakouts, faster averages, volatility-scaled positions as in the momentum paper), or a basket of commodities, where trend following is usually applied.
- Brent, natural gas and the other energy markets.
Why they were buried
Every version was positive over 40 years, and every one failed the control: the same holdings, moved to random times, did about as well. The profit came from oil's rise, from about $26 to $108 a barrel, and from a few huge trends, not from good timing. By our rule the cause is Fragile: positive, but negative without its 10 best trades.
What you can apply to your own tests
- Compare a trend rule with the same holdings at random times, not with zero: in a market that rose fourfold, random long periods made money too.
- Count how much of the result comes from the best few trades before trusting a trend system.
- Look at the years without a trend, not only at the famous crashes: that is where the short side of a trend rule loses.
A note on the numbers: drawdown figures are measured on closed trades unless a table says it includes open losses. Open losses inside a trade are not counted in a closed-trade drawdown, so real account risk is higher, especially for systems with no stop. Backtests are not live trading.
FAQCommon questions
What is the best crude oil trading strategy?
None we tested is proven. The two classic trend rules, the 200-day moving average and 12-month momentum, made money on 40 years of WTI but did no better than the same positions at random times.
Does the 200-day moving average work on crude oil?
Long only, it summed +222.8% over 159 trades from 1986 to 2026, but only 31% of trades won, the result depended on three long trends, and random timing did about as well. Adding shorts below the average lost money on the short side (-151.6% summed).
What is time series momentum?
A trend rule from Moskowitz, Ooi and Pedersen (2012): be long a market whose price is above its level of a year earlier, short (or out) when it is below. On crude oil alone it traded only 34 times in 40 years in our test, too few to judge.
Why does trend following lose in sideways oil markets?
Each false break of the average costs a small loss, and years without a trend have many of them. In our test the long-and-short 200-day rule lost most in 2003, 2019 and 2024.
How was the crude oil trend strategy tested?
On the daily WTI spot price from 1986 to 2026 (EIA via FRED), 0.05% of costs per round trip, a control that moves the same positions to random times, and rules written and dated before the data was run, costs included on every trade, results split into development, validation and holdout periods, a random-entry control as the benchmark, and every variant reported. Read the protocol.
10Check our arithmetic
wti_spec.md
wti_lib.js · wti_ref.js · wti_selftest.js · wti_run.js
control_audit_lib.js · em_lib.js · gh_lib.js · hourly_lib.js · ma_lib.js · rb_lib.js · rsi2_protocol.js
wti_selftest_result.txt · wti_results.txt · wti_summary.json
Code provenance: our own code, written from the specification.
11Data notes
Source. The EIA's daily WTI spot price at Cushing, Oklahoma, published by FRED. Gaps. About one weekday in nine has no price, most in 1999 and 2003; those days are skipped. Close only: no open, high or low, so trades fill at the next close. Spot, not futures: no roll yield. One market, two rules in their usual form.
12Corrections
None since publication.