Backtest · 13 futures markets · 1993 → 2026

Do Classic Futures Trading Rules Beat Buy and Hold? 13 Markets, Up to 34 Years

Books and papers have published dozens of rules for trading futures: follow the trend, buy the dip, trade the calendar, hold only overnight. We ran the classic ones on the Nasdaq-100, S&P 500, Dow, Russell 2000, gold, the 10-year Treasury note and seven other futures markets, on three separate price histories, to see which held up — and whether any beat simply holding.

Published October 14, 2026 · Stable Income Strategy research

The short answer

We tested the classic futures trading rules — trend above the 200-day average, 12-month momentum, RSI(2) and IBS dips, losing streaks, the turn of the month, holidays, overnight holds, moving-average crosses, volatility breakouts, weekdays and Halloween — on 13 markets and three price histories reaching back up to 34 years. The ones that held up in both halves of the history were long-only rules, mostly on the stock indices and gold (the steadiest: RSI(2) dips on the S&P 500, 80% of 256 trades won); breakouts, moving-average crosses, weekdays and shorting the indices with trend rules never did. Run together, the robust rules made $11.8k a year since 1993 against $10.5k for holding the same markets always long at the same risk, with a slightly lower Sharpe ratio (0.67 vs 0.71) — so over the long histories they did not beat buy and hold. Their real value was a worst fall about 30% smaller (−$36.5k vs −$52k) and milder bear years, such as −$10k instead of −$20k in 2008. These are past results, and on one micro contract each rule earned only a few hundred to about two thousand dollars a year.

markets × price histories, every classic rule, up to 34 years: 13 × 3
of 256 RSI(2) dip trades won on the S&P 500 over 34 years: 80%
a year for the robust rules combined — $10.5k always long at the same risk: $11.8k
smaller worst fall than always long (−$36.5k vs −$52k): ~30%

Key takeaways

What held up in both halves of the history was long-side rules, mostly on the stock indices and gold: trend above the 200-day average (Sharpe 0.4–0.6 on the Nasdaq-100, S&P 500, Dow and gold, front-month prices), 12-month momentum, RSI(2) and IBS dips, three lower closes in a row and the turn of the month.
The steadiest single rule: RSI(2) dips on the S&P 500 — 80% of 256 trades won over 34 years, a 5% worst fall, in the market 13% of the time, about $490 a year on one micro contract (MES), roughly half of holding’s return.
Never held up: volatility breakouts (lost on every market), moving-average crosses (too few trades), the day session alone, weekdays, Halloween outside the S&P 500, and shorting stock indices with trend rules.
Overnight holds above the 200-day average were the strongest result on the last two years of real futures (about $3,800 a year on one MNQ, Sharpe 0.78) — but the Nasdaq night lost from 1999 to 2012, and the S&P 500 night was flat over 34 years after financing.
Combined, the robust rules made $11.8k a year since 1993 against $10.5k for always long at the same risk; Sharpe ratio 0.67 vs 0.71. About the same — no extra return.
What the rules bought was calmer holding: a worst fall of −$36.5k vs −$52k, losses of −$6k, −$3k and −$7k in 2000–02 against −$12k, −$15k and −$18k, and −$10k vs −$20k in 2008.
The sizes are small: each rule earned a few hundred to about two thousand dollars a year on one micro contract, while holding one micro S&P 500 contract overnight ties up $2,307 of margin.

What we tested

Five families of published rules, each run at its default settings on one micro contract, with fees, slippage and the cost of financing a futures position. Each card says what the rules do, what the research behind them found, and what held up here.

1Trend following

Rules: Long above the 200-day average (in while the price closes above its 200-day moving average, out below it); 12-month time-series momentum (long while the past year’s return is positive); moving-average crossover (long while a fast average is above a slow one); breakout to a new multi-week high (the Donchian channel).

Research says: Time-series momentum earned a Sharpe ratio near 1 across 58 futures markets from 1985 to 2009 (Moskowitz, Ooi and Pedersen 2012) — but on a single market the Sharpe ratio is about 0.15 (Nilsson), and the smooth result needs 40–60 markets (Graham Capital 2017). On stock indices the 200-day rule earned about what holding did with half the worst fall (Faber 2007).

We found: Long above the 200-day average held up in both halves of 26 years on the Nasdaq-100, S&P 500, Dow and gold (Sharpe 0.4–0.6 on front-month prices), keeping most of holding’s return with a worst fall of 8–16% of capital against holding’s 14–26%. Twelve-month momentum did the same with more trades. Crossovers and breakouts traded too rarely to prove anything; shorting lost on every stock index.

2Mean reversion (buying dips)

Rules: RSI(2) — a 0-to-100 gauge of the last two days; buy when a sharp two-day dip in an uptrend pushes it very low. IBS — where the close sits in the day’s range; buy when it closes near the low. Three lower closes in a row. A close outside the Bollinger band. Fading an unusually big day.

Research says: Stock indices tend to bounce after short, sharp dips (Connors and Alvarez 2008 on RSI(2); Pagonidis 2013 on IBS). Such rules win 65–80% of their trades and sit in the market a tenth to a fifth of the time, so their falls are small — but their total return is well under holding’s, and in 2008 and 2022 the dips kept dipping.

We found: The steadiest results of the whole test. RSI(2) on the S&P 500: 80% of 256 trades won over 34 years, both halves positive, a 5% worst fall, about half of holding’s return. IBS and three lower closes held up on the stock indices, and IBS again on two years of real futures (66–75% of trades won). The Bollinger and big-day fades were weaker. On commodities and currencies, which trend day to day, the same rules lose.

3Overnight and day session

Rules: Overnight — buy at the 4:00 pm close and sell at the 9:30 am open, only while the market is above its 200-day average. Day session — the reverse: in at 9:30 am, out at 4:00 pm.

Research says: Most of the US stock market’s rise since the 1990s came between the close and the next open (Cooper, Cliff and Gulen 2008; Lou, Polk and Skouras 2019; Liu and Tse 2017 on index futures). The strongest stretch, 2 to 3 am New York time as Europe opens, has faded since 2021 (Boyarchenko, Larsen and Whelan, New York Fed 2026).

We found: The strongest result on the last two years of real futures: positive in both halves on seven markets, 54–60% of nights won, about $3,800 a year on one micro Nasdaq-100 contract (MNQ). The longer history is shakier: the Nasdaq night lost from 1999 to 2012 and paid after, and the S&P 500 night was flat over 34 years once financing was paid. The day session alone lost over 25 years.

4Calendar rules

Rules: Turn of the month — hold from the last session of a month through the first few of the next. Before a holiday — hold the session before a market holiday. Days of the week — hold only on chosen weekdays. Halloween — hold November to April, stay out May to October.

Research says: The turn of the month has carried most of the S&P 500’s return since 1980 (Ariel 1987; Lakonishok and Smidt 1988; McConnell and Xu 2008), though a Fed working paper found it gone in S&P futures after 1990. November–April has beaten May–October in most markets for a century (Bouman and Jacobsen 2002). The pre-holiday effect (Ariel 1990) and weekday effects (Grebe and Schiereck 2025) have mostly faded.

We found: The turn of the month held up in both halves on the Nasdaq-100, S&P 500, Dow and gold, in the market a fifth of the time — $300–600 a year on a micro. The session before a holiday paid on most markets, but only $100–300 a year. Halloween kept holding’s return only on the S&P 500; weekdays found nothing.

5Volatility breakout

Rules: Buy when the price rises above the open plus a share of the previous day’s range (sell when it falls that far below) — Larry Williams’ volatility breakout.

Research says: It made Williams’ 1987 trading-contest record; he writes it has worked less well in recent decades (Williams 1999). A newer cousin, the opening-range breakout on the Nasdaq-100 fund (Zarattini and Aziz 2023), reported an annualized alpha of 33% from 2016 to 2023, but an independent replication found the edge survives only at very low costs.

We found: Lost on every market and every price history after costs: the entry stop fills on quiet days that give the move back by the close.

The rules that held up, market by market

A rule is listed where its Sharpe ratio was positive in both halves of that price history. Market codes: NQ Nasdaq-100, ES S&P 500, YM Dow, RTY Russell 2000, GC gold, SI silver, HG copper, ZN 10-year Treasury note. All long only.

Rules that held up in both halves, by price history and market
Rule	Fund prices + financing, from 1993	Front-month futures, from 2000	Real futures, from 2024
Long above the 200-day average	ES, RTY, GC	NQ, ES, YM, GC (Sharpe 0.4–0.6)	—
12-month momentum	ES, YM, RTY, GC, ZN	Held up (indices)	—
RSI(2) dip	ES: 80% of 256 trades won, Sharpe 0.62, 5% worst fall, about $490 a year on 1 MES	Held up (indices)	—
IBS: close near the day’s low	NQ, ES, YM	Held up (indices)	NQ, ES, YM, RTY (66–75% of trades won)
Three lower closes in a row	ES, NQ, RTY, ZN	Held up (indices)	—
Turn of the month	NQ, ES, YM, GC	NQ, ES, YM, GC	—
Session before a holiday	Several markets, only $100–300 a year	—	—
Overnight, above the 200-day average	NQ lost 1999–2012, paid after; ES flat over 34 years	—	NQ, ES, YM, RTY, GC, SI, HG (NQ about $3,800 a year on 1 MNQ, Sharpe 0.78)

“Held up (indices)”: on the front-month history these rules held up on the stock-index futures, where the old quotes are reliable; the test did not break them out market by market. A dash: not among the rules that held up on that history — the real-futures history is only two years, too short for most slow rules. Dollars are per year on one micro contract. Oil, natural gas, the euro and the yen gave no rule that held up on reliable prices.

The rules that never held up

These lost money, or never passed the both-halves test on any reliable price history.

Classic rules that did not hold up
Rule	What happened
Volatility breakout (buy above the open plus a share of yesterday’s range)	Lost on every market and every price history after costs.
Moving-average crossover	About a dozen trades in 25 years on one market — too few to prove anything.
Breakout to a new high (Donchian channel)	Also about a dozen trades in 25 years, and it lost after costs on most markets.
Day session only (9:30 am to 4:00 pm)	Lost over 25 years, and earned nothing on the last two years of real futures.
Days of the week	No weekday choice held up in both halves on the stock indices.
Halloween (November to April only)	Kept holding’s return on the S&P 500 by skipping its 2008 and 2022 summers; nothing on the other markets.
Shorting stock indices with trend rules	Lost on every stock index.
Bollinger-band and big-day fades	Weaker than RSI(2), IBS and losing streaks; not among the rules that held up.
Dip-buying on commodities and currencies	These markets trend day to day, and the same rules lose. They look spectacular only on unreliable old quotes (copper, gas, the euro).

All the robust rules together vs simply holding

The robust rules run as one portfolio — long above the 200-day average on the Nasdaq-100, S&P 500 and gold, plus RSI(2) and IBS dips on the S&P 500 — against a yardstick that holds the same markets always long, re-sized every day to the same risk.

The robust rules combined against always long at the same risk
Measure	Robust rules	Always long, same risk
Average a year, 1993–2026	$11.8k	$10.5k
Sharpe ratio, 1993–2026	0.67	0.71
Worst fall, 1993–2026	−$36.5k	−$52k
2000	−$6k	−$12k
2001	−$3k	−$15k
2002	−$7k	−$18k
2008	−$10k	−$20k
Sharpe ratio, front-month prices from 2000 (no financing)	0.83	0.88
Sharpe ratio, real futures (about 14 months)	1.09	0.88

Fund prices with financing unless stated. The rules were up in 25 of 34 years; their worst year was 2022 (−$19k). Peak margin was $27k, so $11.8k a year is about 19% a year on peak margin plus the worst fall. On real futures they made $21k a year, but only about 14 months were left after the 200-day warm-up. Adding overnight holds on NQ, ES and gold lowered the Sharpe ratio to 0.53: overnight on the S&P 500 since 1993 lost $1.8k a year after $101k of fees, and overnight added nothing before 2010.

Is it luck? How we checked

Run enough rules on enough markets and some will look good by chance. These checks are what separate a pattern from a lucky pick.

Both halves

Each history was split in two, and a rule passed only with a positive Sharpe ratio in each half. The Nasdaq overnight rule shows why: over 27 years of fund prices it earned 60% of holding’s return with a third of the worst fall — but it lost from 1999 to 2012 and made its money after.

Three price histories

Fund prices go back to 1993 and are charged financing; the front-month quote starts in 2000 and leaves financing out, so it is optimistic; real futures are exact but start in October 2024. Most rules that held up did so on more than one.

Thin markets left out

On the euro, yen, 10-year note, silver, copper and natural gas, the old front-month quotes did not match real trading: near each roll they belong to the expiring contract. One-day rules look falsely great there — IBS on copper showed a Sharpe ratio of 2.1, fading big days on the euro a profit factor of 71. No edge is claimed from those prices.

Fund prices flatter the Nasdaq

In an earlier comparison over the two years of real futures, intraday Nasdaq rules looked better on the QQQ fund than on real Nasdaq-100 futures bars of the same days. A result seen only on fund prices deserves less trust than one that repeats on futures.

Financing counts

A futures position carries the cost of financing it — roughly the Treasury-bill rate. Charged every session on the fund history, it left the S&P 500 night flat over 34 years and, with the dot-com bear and financing at 3–5% a year, turned the Nasdaq night negative from 1999 to 2012.

Default settings, related markets

Every rule ran at its default settings, not settings tuned to the data. Most rules that held up did so on several related markets at once — two or more stock indices, not one lucky market.

What it takes: margin and room for falls

Futures need margin — a deposit per contract — not the full value. The swing rules here hold overnight, so they need the overnight margin, plus room for the falls: a fall of 20% of the contract’s value ($10,000 on one MNQ) is an ordinary year.

Margin per micro contract, autumn 2026
Contract	Day-trading margin	Overnight margin
Micro Nasdaq-100 (MNQ)	$100	$3,542
Micro S&P 500 (MES)	$50	$2,307

One broker’s margin table, autumn 2026; margins change. Holding one MES all the time earned about $900 a year after financing, with a worst fall of 14% of the test’s capital. The rules that made money earned a few hundred to about two thousand dollars a year each on one micro — the same order as the margin they tie up. Nothing here turns a small account into a living.

The risks

At the same risk the rules earned about what holding did (Sharpe ratio 0.67 vs 0.71 since 1993). What they offer is smaller falls, not more money.
Dips can keep dipping: in 2008 and 2022 the dip-buying rules bought into falls that kept going, and the combined rules lost $19k in 2022, their worst year.
The real-futures history is only two years — about 14 months after the 200-day warm-up. Results there, like the overnight rule’s $3,800 a year on NQ, rest on a short stretch.
Trend following has had long dry spells since 2010: the SG Trend index of trend-following funds lost 20% from May 2024 to May 2025.
Leverage: an ordinary 20% fall moves one MNQ about $10,000, against $3,542 of overnight margin.
The retail record is poor: of Brazilian day traders who kept at index futures for 300 days, 97% lost money (Chague, De-Losso and Giovannetti 2020).
Published edges fade: the 2–3 am overnight stretch since 2021, the pre-holiday effect for large US stocks since the 1980s, and a Fed working paper found the turn of the month gone in S&P futures after 1990.

The research behind the rules

The papers behind each family of rules. Their numbers are theirs; everything else on this page is our own backtest.

Moskowitz, Ooi and Pedersen (2012), “Time series momentum”, Journal of Financial Economics — a Sharpe ratio near 1 across 58 futures markets, 1985–2009.
Faber (2007), “A quantitative approach to tactical asset allocation” — the 200-day moving-average rule.
Nilsson, “An analysis of expected returns of trend-following strategies” — a Sharpe ratio of about 0.15 per single market.
Graham Capital (2017), “Market diversification” — 40–60 markets for the diversified trend result.
Connors and Alvarez (2008), “Short term trading strategies that work” — the RSI(2) pullback.
Pagonidis (2013), “The IBS effect: mean reversion in equity ETFs”.
Cooper, Cliff and Gulen (2008), “Return differences between trading and non-trading hours”; Lou, Polk and Skouras (2019), “A tug of war: overnight versus intraday expected returns”; Liu and Tse (2017) on index futures.
Boyarchenko, Larsen and Whelan (2026), “The overnight drift”, Federal Reserve Bank of New York.
Ariel (1987); Lakonishok and Smidt (1988); McConnell and Xu (2008), “Equity returns at the turn of the month”, Financial Analysts Journal.
Bouman and Jacobsen (2002), “The Halloween indicator”, American Economic Review.
Ariel (1990), “High stock returns before holidays”; Grebe and Schiereck (2025), day-of-the-week effects, Eurasian Economic Review.
Williams (1999), “Long-term secrets to short-term trading”; Zarattini and Aziz (2023), “Can day trading really be profitable?”.
Chague, De-Losso and Giovannetti (2020), “Day trading for a living?” — the record of persistent retail day traders.
For intraday rules, not part of this daily test: Gao, Han, Li and Zhou (2018), “Market intraday momentum”; Lucca and Moench (2015), “The pre-FOMC announcement drift”.

Questions

Do classic trading rules beat buy and hold on futures?

Not on return at the same risk. The rules that held up — long above the 200-day average on the Nasdaq-100, S&P 500 and gold, plus RSI(2) and IBS dips on the S&P 500 — made $11.8k a year together since 1993, against $10.5k for holding the same markets always long at the same risk, with a Sharpe ratio of 0.67 vs 0.71 (0.83 vs 0.88 on front-month prices since 2000). Their gain was smaller falls: a worst fall of −$36.5k vs −$52k, and −$10k vs −$20k in 2008.

Does trend following work on futures?

Across many markets, it has: time-series momentum earned a Sharpe ratio near 1 across 58 futures markets from 1985 to 2009 (Moskowitz, Ooi and Pedersen 2012), but on a single market the Sharpe ratio is about 0.15, and the smooth result needs 40–60 markets. In our test, long only above the 200-day average held up in both halves on the Nasdaq-100, S&P 500, Dow and gold (Sharpe 0.4–0.6 on front-month prices since 2000), with a worst fall of 8–16% of capital against holding’s 14–26%. Shorting with trend rules lost on every stock index, and the SG Trend index of trend-following funds lost 20% from May 2024 to May 2025.

Does the RSI(2) strategy still work?

On the S&P 500, it held up in our test: buying sharp two-day dips (Connors and Alvarez 2008) won 80% of 256 trades over 34 years, was positive in both halves, had a 5% worst fall and was in the market 13% of the time — about $490 a year on one micro S&P 500 contract (MES), roughly half of holding’s return. In 2008 and 2022 the dips kept dipping, and on commodities and currencies the same kind of rule loses.

What is the IBS strategy, and does it work on futures?

IBS (internal bar strength) is where the close sits in the day’s range, from 0 at the low to 1 at the high; the rule buys a close near the low (Pagonidis 2013). In our test it held up in both halves on the Nasdaq-100, S&P 500 and Dow over the long fund-price history, and again on the last two years of real Nasdaq-100, S&P 500, Dow and Russell 2000 futures, where 66–75% of its trades won.

Is the turn-of-the-month effect real?

It has been. The last session of a month and the first three of the next carried most of the S&P 500’s return since 1980 (Ariel 1987; Lakonishok and Smidt 1988; McConnell and Xu 2008), though a Fed working paper found it gone in S&P futures after 1990 and others still find it through 2024. In our test it made money in both halves on the Nasdaq-100, S&P 500, Dow and gold, on both front-month and fund prices, while in the market only a fifth of the time — a small, steady $300–600 a year on one micro contract.

Does holding futures only overnight work?

On the last two years of real futures it was the strongest result: holding from the 4:00 pm close to the 9:30 am open while above the 200-day average was positive in both halves on seven markets, won 54–60% of nights and made about $3,800 a year on one micro Nasdaq-100 contract (Sharpe 0.78), while the day session earned nothing. Over longer history it is shakier: the Nasdaq night lost from 1999 to 2012, the S&P 500 night was flat over 34 years after financing, and adding overnight holds to the combined rules lowered their Sharpe ratio from 0.67 to 0.53.

Do moving-average crossovers and breakouts work on futures?

Not in our test. The moving-average crossover and the Donchian breakout traded only about a dozen times in 25 years on one market — too few to prove anything — and the breakout lost after costs on most markets. Larry Williams’ volatility breakout lost on every market and every price history after costs.

How much money do you need to trade these rules with micro futures?

More than the margin. In autumn 2026 one broker asked $3,542 of overnight margin for one micro Nasdaq-100 contract (MNQ) and $2,307 for one micro S&P 500 (MES); day-trading margin was $100 and $50, but swing rules hold overnight. A fall of 20% of the contract’s value — $10,000 on one MNQ — is an ordinary year, and each rule that worked earned only a few hundred to about two thousand dollars a year on one micro.

Is this investment advice?

No. It is a backtest of past prices for education. Futures are leveraged and can lose more than the margin put up, and past results do not promise future ones.

How the test works

Three price histories for each of 13 futures markets: its fund (QQQ, SPY, DIA, IWM, GLD, SLV, CPER, IEF, FXE or FXY) with dividends put back and scaled to today’s futures price, from 1993 where the fund goes back that far (oil and gas have no fund that tracks well enough); the continuous front-month futures quote since 2000, not adjusted for rolls; and real futures sessions since October 2024, using each day’s 9:30 am–4:00 pm open, high, low and close.
Financing: holding futures instead of the fund costs about the Treasury-bill rate. On the fund history it is charged for every session held (the 13-week T-bill rate ÷ 252) on the stock indices, gold, silver and the 10-year note. The front-month quote turned out to track the index itself, so that history leaves the cost out and is optimistic.
Trades: a signal at a session’s close, entry at the next open or at that close; stops and targets measured in average true ranges (14 sessions), checked against the session’s high and low assuming the worse came first; gaps fill at the open.
The test: every one of the 17 daily rules (buy and hold included) at its default settings, on one micro contract at the prices of the day — so early years’ dollars are smaller — with $0.80 a side in fees plus slippage. A rule counts as robust when its Sharpe ratio is positive in both halves of the history.
The portfolio: the robust rules sized at $600 per average true range, $0.80 fees and one tick of slippage, with margin at today’s overnight rate as a share of the contract’s value. The yardstick holds the same markets always long, re-sized every day to the same risk.
Sharpe ratio: average return divided by how much it swings — the fair way to compare a rule that is often out of the market with one that is always in. Worst fall: the deepest drop from a peak.

Not investment advice

This is a study of past prices, published for education. Leveraged funds, options and futures can lose most of their value quickly, and a pattern seen in the past can break. Nothing here is a recommendation to buy or sell anything. Decide with your own judgment, or with a licensed adviser.