DRAWUP

backtested · commodities · 1,020 daily bars

natural gas, backtested

five classic strategies run on 1,020 days of natural gas, with fees, split so nothing was chosen with hindsight. one of them made money in the first half, one in the second.

best rule, second half
31.5%
holding, second half
-10.3%
rules that beat holding
1 of 4
made money, second half
1 of 5
the DRAWUP bot holding a glowing coin

natural gas right now

2.6300

as of 2026-11-03, from federal reserve (fred) public data. the backtest below runs to 2026-11-03.

what this is

what we did to natural gas, and what came out

Five strategies, 1,020 days of natural gas, 2022-10-03 to 2026-11-03, fees paid on every trade. We split the history at 2025-08-12 and judged all five on the half after it, which none of them were tuned on. How the test is built.

rsi(14) mean reversion led both halves on natural gas, at +123.8% then +31.5%. That is unusual, and it is the only version of this result worth much.

Holding natural gas without trading returned -10.3% over that window. one of the four active rules beat it. one ended up ahead in absolute terms. The gap between the best and worst rule was 99.8%, on the same prices over the same days.

the whole table

every strategy, not the winner

natural gas daily, 2022-10-03 to 2026-11-03. fees 5 bps a side. in sample to 2025-08-12.
strategyin sample out of sampleequity, out of sample
returnmax ddsharpe returnmax ddsharpe
rsi(14) mean reversion 123.8%-45.2%0.88 31.5%-35.9%0.82
50/200 moving average -45.5%-95.0%0.65 -45.3%-91.0%0.65
20 day breakout -87.7%-90.0%0.38 -52.7%-91.7%0.57
12 month momentum -81.4%-82.4%0.06 -68.3%-94.7%0.46
buy and hold -45.9%-90.8%0.80 -10.3%-92.1%0.88

the in sample winner was rsi(14) mean reversion at 123.8%. picking it on that evidence then returned 31.5% in the window that followed. that is the finding, and it is why this page prints all five.

the five

what each strategy tries to do, and what it did to natural gas

buy and hold — buy it on day one and never touch it again. -45.9% first half, -10.3% second, and kept losing. Worst fall 92.1%, one trade.

50/200 moving average — hold only while the recent average price sits above the long-run one. -45.5% first half, -45.3% second, and kept losing. Worst fall 91.0%, 2 trades.

12 month momentum — hold only what has already risen over the past year. -81.4% first half, -68.3% second, and kept losing. Worst fall 94.7%, 38 trades.

rsi(14) mean reversion — buy after a sharp fall, betting it bounces. +123.8% first half, +31.5% second, and held up. Worst fall 35.9%, 10 trades.

20 day breakout — buy when the price breaks above its highest point in twenty days. -87.7% first half, -52.7% second, and kept losing. Worst fall 91.7%, 10 trades.

the catch

where each one went wrong on natural gas

buy and hold — You take the whole fall with it. It lost 10.3% out of sample, and -45.9% was the best it ever showed. It made one decision in that window. A result built on that few calls is a coin toss with a long wait attached, not an edge.

50/200 moving average — Averages only turn after the move has happened. It lost 45.3% out of sample, and -45.5% was the best it ever showed. It made 2 decisions in that window. A result built on that few calls is a coin toss with a long wait attached, not an edge.

12 month momentum — It is buying what is already expensive. It lost 68.3% out of sample, and -81.4% was the best it ever showed. It traded 38 times to get there and paid the fee on every one, which is how a rule that is roughly right still finishes behind.

rsi(14) mean reversion — A thing that fell hard can keep falling. It made +31.5% in the half it was never allowed to see, which on this page nothing else managed. It held natural gas only 28.2% of the time, and that is why it lost less than holding did. Being out of the market is not the same as being right about it.

20 day breakout — Most breakouts fail and hand the move straight back. It lost 52.7% out of sample, and -87.7% was the best it ever showed. 1178.4% of everything it made came from five days. Miss those and the answer changes completely.

year by year

which years each rule actually won on natural gas

calendar year returns, natural gas, each rule run on the same days.
yearbuy and hold50/200 moving average12 month momentumrsi(14) mean reversion20 day breakout
2022+5.7%+0.0%+5.7%+11.8%-0.4%
2023-0.7%-43.5%-41.1%-6.7%-42.5%
2024-3.8%-4.2%-29.7%+24.8%-6.3%
2025-6.7%+0.0%-22.5%-1.1%-5.5%

so what works

the hard part is not the trading

One rule survived the half they had not seen on natural gas. Running a rule without flinching is the easy part for a machine; knowing which rule is worth running is not, which is why we test this many and publish the failures too.

reading it

what this says

natural gas fell 10.3% through the out of sample window, and exactly one rule came out ahead: rsi(14) mean reversion, at 31.5%. the honest reading is not that these five are bad, it is that a backtest chosen on past performance tells you little about the next window. the federal reserve publishes this series on a schedule rather than continuously, so the window ends on its last published day, 2026-11-03.

the one rule that came out ahead, rsi(14) mean reversion, was in the market 28% of the window. it earned its 31.5% by being out of it for the rest, which is a timing claim that one window cannot settle. its five best days alone compound to 125.3%, so most of the result is a handful of bars rather than a steady edge.

method

how it was run

data
natural gas daily closes, federal reserve (fred) public data, 1,020 bars, 2022-10-03 to 2026-11-03, read 2026-11-16. this series is published as one value a day with no intraday range, so the 20 day breakout is measured on closing prices.
fees
5 basis points a side on every change of position. carry is not modelled. a commodity position earns or pays the roll between contracts, which a spot index does not carry, so the omission does not run one way.
no hindsight
every decision is taken on one close and applied to the next day's move.
the split
first 70% sets expectations, last 30% tests them. nothing was tuned on it.
missing
one asset, one timeframe, long only, no sizing, no stops.

this market

what stands out about natural gas

Held and not traded, natural gas returned -10.3% out of sample, 43rd of 144 markets measured the same way.

The fall hurt more than most: holding it meant sitting through a 92.1% drawdown, the 3rd deepest in the set.

1 of the four rules beat holding. The best, rsi(14) mean reversion, returned +31.5% out of sample against -10.3% for doing nothing.

1681.1% of the whole gain from holding natural gas arrived on its five best days. Miss those five and the rest of the period is close to flat.

rsi(14) mean reversion was only in the market 28.2% of the time and traded 10 times across 1,020 days, so most of that return was earned while holding nothing.

Run inside a funded account with a 10% trailing drawdown, the best rule would have breached on day 96 of 306. The return was real; the account would have been closed before it arrived.

Year by year, holding natural gas gave 2022 +5.7%, 2023 -0.7%, 2024 -3.8%, 2025 -6.7%. That is 1 up years and 3 down.

The four rules, best to worst out of sample on natural gas: rsi(14) mean reversion +31.5%, in the market 28.2% of the time, worst fall 35.9%; 50/200 moving average -45.3%, in the market 28.2% of the time, worst fall 91.0%; 20 day breakout -52.7%, in the market 57.0% of the time, worst fall 91.7%; 12 month momentum -68.3%, in the market 52.5% of the time, worst fall 94.7%.

questions

what people ask

what is the best strategy for natural gas?

rsi(14) mean reversion came out best out of sample at +31.5%, against -10.3% for holding and not trading. 1 of the four rules beat holding.

would natural gas have passed a funded account challenge?

not on these rules. run with a 10% trailing drawdown, the best rule breaches on day 96 of 306. the return was real and the account would have been closed before it arrived.

how much of natural gas's gain came from a handful of days?

1681.1% of it arrived on the five best days. miss those five and the rest of the period is close to flat, which is the argument against being out of the market at the wrong moment.

how bad was the worst fall in natural gas?

92.1% from peak to trough while simply holding, the 3rd deepest of the 144 markets we have run this way.

how often was the best rule actually holding natural gas?

28.2% of the time, across 10 trades in 1,020 days. most of the period it held nothing, which is why its worst fall is shallower than the market's.

bots we have actually tested, with the results published

more

other backtests