Backtest · S&P 500 and Nasdaq-100 · 1928 → 2026

Leveraged ETFs in the Year After a Midterm Election: What 75 Years of Data Say

The year after a US midterm election has been the stock market’s best year of the four-year cycle. We asked how to make the most of it with leveraged ETFs — and tested more than 3,000 ways, from plain buy and hold to day trading, scalping, chart indicators, the VIX and the Fear & Greed index.

Published October 14, 2026 · Stable Income Strategy research

The short answer

Since 1950 the S&P 500 has risen in all 19 of the 12-month periods after a US midterm election (typical gain +18.6%). Of more than 3,000 trading rules we tested, the best was the simplest: hold a 3× S&P 500 fund (such as UPRO) from the close the day before the election to July 31, then sit in Treasury bills until the year is over. In each of those 19 years it turned $10,000 into between $11,207 and $25,169 (typical +52.6%), but the account fell as much as 43% along the way, and the same rule lost money before 1950 and in ordinary years.

post-midterm years the S&P 500 rose, since 1950: 19 of 19
typical year of the winning 3× rule: +52.6%
deepest drop inside a year along the way: −43%
rules tested to find it: 3,000+

Key takeaways

The year after a midterm is special: the S&P 500 rose 19 of 19 times since 1950 (typical +18.6%), against +9.1% and 42 of 56 in all other years.
Best rule: 3× S&P 500 from the day before the election to July 31, then Treasury bills — typical year +52.6%, worst year +12%, no losing year since 1950.
The Nasdaq-100 version (a 3× fund such as TQQQ) made more: typical +80%, worst +24%, 10 of 10 years since 1986.
It beat buy and hold: SPY all year made +18.6% a typical year, UPRO all year +39.9% with a −35% year.
Nothing else beat it once picked without seeing the year: not day trading, scalping, averaging down, 89 chart indicators, the weekday, the VIX, the Fear & Greed index or more leverage.
The risk is real: in late 2018 $10,000 fell to $6,189 before ending at $11,888, and the rule lost money in 1930 and 1938 and in ordinary years (worst −63%).

What we tested

Five studies, each one a step further. All of them use real daily prices back to 1928 (1-minute prices since 2016 for day trading), the cost of every trade, and interest on cash.

1Leveraged S&P 500 funds, all years

Tested: SSO (2×), UPRO and SPXL (3×) since 2009, a built 3× fund back to 1993: every day-trading and swing rule of our futures tools, and 1,764 rules from 98 chart-indicator conditions.

Found: No rule made more dollars than simply holding UPRO. The best ones (buy when RSI(3) is under 20; hold only overnight above the 200-day average) earned about half as much with less than half the drop. Short-side rules lost or made little, and none passed the indicator search; opening-range scalps lost.

2TQQQ with the least risk, 40 years

Tested: 72 rules on the Nasdaq-100 since 1986 — trend filters, stops, fixed shares, moving-average crosses, momentum, overnight holds, volatility sizing.

Found: Only sizing by volatility held up on years it was not tuned on: 23% a year with a −32% worst drop since 2016, against TQQQ’s −82%.

3Averaging down

Tested: Buy a slice each 1–10% lower, sell everything at +1% to +10% on cost, start again — on SPY, SSO, UPRO, SPXL and a 3× fund since 1993.

Found: Every cycle closed at a profit, even through 2000–2009, but it needed up to $315,000 of cash per $1,000 slice, and per dollar it earned less than holding at the same risk.

4The year after midterms

Tested: 46 rules on the S&P 500, Nasdaq Composite, Nasdaq-100 and Russell 2000; 256 leverage schedules across four parts of the year; leverage up to 6×; switching between indexes; nine buy and sell dates.

Found: 3× until spring or summer, then out, beat every other rule — and was chosen 19 times out of 19 when each year was picked blind.

5The year after midterms, part two

Tested: 13 day-trading and scalping rules on 1-minute bars; 1,405 rules from 89 chart indicators, the weekday, the turn of the month, the VIX and the Fear & Greed index; averaging down inside the year.

Found: Some beat the calendar rule on paper — as luck guarantees when 1,400 rules meet 16 years — but none once picked without seeing the year.

The rule that won: 3× until July 31, then Treasury bills

Three trades every four years. Nothing to watch in between.

1
At the close, the day before the election
Put the account in a 3× S&P 500 fund (UPRO or SPXL) — or a 3× Nasdaq-100 fund (TQQQ).
2
At the close on July 31 (or the last trading day before it)
Sell, and put the money in Treasury bills (a T-bill fund such as SGOV or BIL).
3
One year after the election
The year is over. Nothing until the next midterm, four years later.

This cycle’s dates

Midterm election: November 3, 2026. The rule buys at the close on November 2, 2026, moves to Treasury bills at the close on July 30, 2027 (July 31 is a Saturday), and the year ends on November 3, 2027.

Why it works in the data: the 3× part sits in November to July, the strongest stretch of the strongest year of the cycle; it is out for August to October, when the 1987 crash and the 2011 and 2015 sell-offs fell.

Against buy and hold

$10,000 in each of the 19 post-midterm years since 1950 (the Nasdaq-100 row: 10 years since 1986).

Each way of holding, across the post-midterm years
Way of holding	Typical year	Worst year	Losing years	Deepest drop inside a year
Buy and hold the S&P 500 (SPY), all year	+18.6%	+4%	0 of 19	−33%
Buy and hold 3× (UPRO), all year	+39.9%	−35%	1 of 19	−78%
3× to April 30, then 1×	+47.1%	+11%	0 of 19	−43%
3× to July 31, then Treasury bills	+52.6%	+12%	0 of 19	−43%
Same, on the Nasdaq-100 (TQQQ), since 1986	+80.0%	+24%	0 of 10	−47%

“Typical year” is the geometric average: $10,000 grows to $10,000 × (1 + typical) in an average year.

Every post-midterm year since 1950

Each row is the year that starts the day before that midterm election. The bar is where $10,000 ended with the rule; “lowest” is the worst point along the way.

$10,000 in each post-midterm year, the rule against holding
Midterm	S&P 500 held	3× held all year	$10,000 with the rule	Lowest
1950	+24%	+75%	$17,092	$9,473
1954	+39%	+136%	$25,169	$10,000
1958	+14%	+34%	$16,364	$9,674
1962	+29%	+94%	$16,750	$9,981
1966	+16%	+39%	$16,303	$9,585
1970	+17%	+41%	$14,996	$9,723
1974	+27%	+65%	$16,542	$6,919
1978	+10%	+5%	$12,699	$9,146
1982	+27%	+59%	$15,879	$9,313
1986	+4%	−35%	$20,179	$8,897
1990	+28%	+74%	$17,544	$9,185
1994	+31%	+94%	$16,982	$8,840
1998	+23%	+52%	$15,218	$9,974
2002	+18%	+42%	$11,617	$6,558
2006	+9%	+9%	$11,236	$9,587
2010	+6%	+2%	$12,662	$9,831
2014	+6%	+10%	$11,207	$9,356
2018	+14%	+31%	$11,888	$6,189
2022	+17%	+33%	$15,918	$9,527

Before 1950 the same rule went: 1930 → $4,600, 1934 → $16,804, 1938 → $6,827, 1942 → $18,796, 1946 → $11,201. Two of five years lost money, one of them more than half the account.

Is it luck? How we checked

Test enough rules on 19 years and one will look perfect by chance. These checks are what separate a pattern from a lucky pick.

Picked blind

Each year was predicted by whichever of 256 leverage schedules did best on the other 18 years. It chose “3× to July 31, then out” all 19 times — and earned the same +52.6% typical year.

Early years → later years

Ranked on 1950–1986 alone, the leverage-until-spring rules came out on top; run on 1990–2022, which they had never seen, “3× to April 30, then 1×” went 9 for 9 (+35% typical).

Ordinary years

The same rule in the 56 years that did not follow a midterm made only +9.8% a typical year, with a −63% worst year. The edge is the midterm year, not the rule.

Before 1950

In 1930–1946 it was up in 3 of 5 midterm years: 1930 turned $10,000 into $4,600, 1938 into $6,827.

Real funds

The 3× funds built from the index tracked the real UPRO, SSO, TQQQ and QLD within 1–4 percentage points a year in the post-midterm years they existed.

Nearby dates

Switching any time from April 30 to July 31 gives a similar typical year (+44% to +53%). Holding 3× into August or September brings back losing years of −8% to −27%.

What did not beat it

Ideas tested against the rule
Idea	What happened
More leverage (4× to 6×, possible only with futures)	Bigger on paper (6× to April 30: +90% a typical year), but the account fell 53–69% inside a year, and 5× held all year was wiped out by the 1987 crash.
Day trading (13 intraday rules, 1-minute bars)	The best made $1,799 and $2,936 on $10,000 in 2018–19 and 2022–23; the calendar rule made $2,185 and $5,959. They need 160–230 day trades a year — an account under $25,000 is allowed 3 in 5 days.
Scalping (opening-range scalps)	Lost money on UPRO in both post-midterm years; on SPY at most $131 on $10,000 in a year.
89 chart indicators, the weekday, the turn of the month (1,189 rules)	65 beat the rule on paper; picked without seeing the year they made +34.4% a typical year, against the rule’s +47.1%.
The VIX (97 rules, since 1990)	Picked blind: +34.4% a typical year, against the rule’s +37.3% on the same years.
The Fear & Greed index (121 rules, since 2011)	Only three post-midterm years to test. Picked blind: +21.2%, against the rule’s +30.4%.
Averaging down inside the year	At best +25.6% a typical year on 3×, with a −50% year; 5–10% a year on 1×.
Switching between the S&P 500, Nasdaq-100 and Russell 2000	No better than staying in the Nasdaq-100 on the same calendar. The Russell 2000 was the weakest index (a −20% year).
A month-by-month map of the good months	Built from the other years, it made less and brought back −36% years.

The risks

Only 19 post-midterm years since 1950, and 10 for the Nasdaq-100. A pattern seen 19 times can still break.
Deep drops inside the year: $10,000 fell to $6,189 in late 2018, $6,558 in 2002 and $6,919 in 1974 before recovering. Selling in a panic turns those into real losses.
It has failed before: after the 1930 and 1938 midterms, and in ordinary years (worst −63% on the S&P 500, −95% on the Nasdaq-100).
3× funds reset every day: in a choppy year they can lose even when the index ends flat.
Selling before a full year is a short-term gain for taxes. Taxes are not counted here.

Where things stand

As of the close on October 13, 2026

The S&P 500 closed at 7,805: up 14% this year, 7.5% above its 200-day average, with the VIX at 15.3 — no midterm-year slump this time. The seven years since 1950 that also came into the election up more than 10% (1950, 1954, 1958, 1982, 1986, 1998, 2006) all ended up with the rule, from +12% to +152%. Before 1950, 1938 is the counterexample: up 27% into the election, then −32% with the rule.

Questions

Does the stock market go up after midterm elections?

Since 1950, yes every time: the S&P 500 rose in all 19 of the 12-month periods that started the day before a midterm election, by +18.6% in a typical year (smallest gain +4%, in 1986–87). In the other 56 years the same 12 months rose only 42 times, +9.1% typical. Before 1950 it was not reliable: the S&P 500 fell 33% after the 1930 midterm and 1% after 1938.

What is the best leveraged ETF strategy for the year after a midterm?

In our backtest of more than 3,000 rules, the best was to hold a 3× S&P 500 fund (such as UPRO) from the close the day before the election to July 31, then move to Treasury bills until one year after the election. It made +52.6% in a typical year since 1950, with no losing year (worst +12%).

How much would $10,000 have made after each midterm?

With the 3×-to-July-31 rule, $10,000 ended each post-midterm year since 1950 at between $11,207 (2014) and $25,169 (1954), about $15,260 in a typical year. Holding the S&P 500 (SPY) turned it into about $11,860.

Is UPRO or TQQQ better after a midterm?

The Nasdaq-100 version (TQQQ until July 31, then Treasury bills) made more: +80% in a typical year and at least +24% in each of the 10 post-midterm years since 1986. It has fewer years of history and slightly deeper drops (−47% inside a year).

Does day trading or scalping beat holding after a midterm?

No. The best intraday rule made $1,799 and $2,936 on $10,000 in the 2018–19 and 2022–23 post-midterm years, against $2,185 and $5,959 for the calendar rule; opening-range scalps lost. Day trading also needs 160–230 trades a year, and accounts under $25,000 are limited to 3 day trades in 5 business days.

Do RSI, MACD, the VIX or the Fear & Greed index improve the strategy?

Not reliably. We tried 89 chart indicators, the weekday, the turn of the month, 8 VIX levels and 10 Fear & Greed readings, each 12 ways (1,405 rules). Some beat the calendar rule on paper, but when each year was picked without seeing it they did worse: +34.4% vs +47.1% for the indicators, +34.4% vs +37.3% for the VIX, +21.2% vs +30.4% for Fear & Greed.

What is the biggest risk?

Deep drops inside the year and the chance that the pattern breaks. In late 2018 $10,000 fell to $6,189 before ending at $11,888. The same rule lost after the 1930 and 1938 midterms, and in ordinary years its worst year was −63% on the S&P 500.

When does the 2026–2027 post-midterm window start and end?

The midterm election is on November 3, 2026. The rule would buy at the close on November 2, 2026, switch to Treasury bills at the close on July 30, 2027 (July 31 is a Saturday), and the year ends on November 3, 2027.

Is this investment advice?

No. It is a backtest of past prices for education. Leveraged ETFs can lose most of their value quickly, and past results do not promise future ones.

How the test works

Prices: the S&P 500 every day since 1928 with its dividends added back; the Nasdaq Composite since 1971, the Nasdaq-100 since 1985, the Russell 2000 since 1987; the VIX since 1990; the Fear & Greed index since 2011; 1-minute bars of SPY and UPRO since 2016; the 3-month Treasury bill rate for cash.
Leveraged funds are built from the index: 2× or 3× its daily move, less a 0.9% yearly fee and the cost of the borrowed part (the T-bill rate + 0.5%). The real funds’ own prices check them where they exist.
The year: bought at the close the day before the midterm election (the first Tuesday after the first Monday of November), sold at the close one year later.
Trades at the close, 0.03% cost on every trade (a cent a share on the 1-minute bars), Treasury-bill interest on idle cash, no taxes, no commissions.
“Typical year” is the geometric average across the years, so one big year cannot hide a bad one.
Honest tests: each year predicted by the rule that did best on the other years; rules picked on early years and run on later ones; the same rules on ordinary years and before 1950.

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.