$2,500 a month starts at 6% a year.
Can SPY help decide when to hold semiconductor stocks?
Strategy in short: Hold SMH when SPY beat T-bills in at least 5 of the past 12 completed months. Otherwise hold T-bills.
Retirement answer: SPY-guarded SMH supported more starting spending than Always SMH: 4.4% versus 2.2% a year funded 30 years in at least 95% of both sets of histories. These are historical thresholds, not recommended spending.
Return comparison: SPY-guarded SMH made more money in this backtest. SPY-guarded SMH ended with $427,659 and Always SMH with $194,531.
Growth of $10,000 — logarithmic scale
SPY-guarded SMH had a deepest fall of 61.0%, compared with 82.1% for Always SMH.
See when the broad-market signal protected SMH
The strategy uses the broader SPY market as its warning signal while owning the narrower and more volatile semiconductor fund.
The rule reduced exposure during 29 of the 312 months tested.
SPY-guarded SMH$427,659-61.0% deepest fall
Always SMH$194,531-82.1% deepest fall
July 2000 through June 2026. Distributions are reinvested. The calculation charges 0.10% of each risky asset bought or sold. There are no withdrawals in this graph. Equal percentage moves take equal vertical space. Past results do not predict future results.
Did the SPY guard make SMH easier to spend from?
By May 2006, Always SMH had run out while SPY-guarded SMH still had $106,564.
Avoiding more of the earlier losses left money available to keep paying withdrawals.
Example: You have $500,000 and need $2,500 a month—6% a year.
The same spending then equaled 33.1% a year.
The portfolio had run out of money.
What funded at least 95% of both history tests?
SPY-guarded SMH supported a 4.4% starting annual withdrawal for 30 years: 307 ordered starts and 4,769 of 5,000 mixed histories funded every withdrawal.
These are historical thresholds, not recommended spending amounts or a promise of future success.
Test if SPY-guarded SMH can fund your early retirement with your own numbers
Enter your portfolio, monthly spending, and how long the money needs to last.
$2,500 a month for 30 years starts at 6.0% of the portfolio each year.
In these backtests, sometimes. Neither strategy funded this early retirement in at least 8 out of 10 histories tested.
312 starting months; the fund's record repeats in order when needed
- SPY-guarded SMH
- 295 funded
- Always SMH
- 233 funded
5,000 histories built from 8–12-year pieces
- SPY-guarded SMH
- 4,292 funded
- Always SMH
- 3,509 funded
A path funded the goal only if it paid every inflation-adjusted monthly withdrawal for 30 years. Results use the nearest 0.1 percentage-point starting withdrawal rate. They reuse one 26-year market record and are not a probability about your future.
How the FIRE calculator works
The calculator applies the portfolio value, monthly withdrawal, and number of years you enter. The first withdrawal occurs before the first month's return, and later withdrawals rise with U.S. inflation. A history counts as funded only if it pays every monthly withdrawal for the full period and finishes above zero.
One result begins at each of the 312 months in the fund record and follows that record forward in order, returning to its first month when more months are needed. The other builds 5,000 longer histories from linked 8-, 10-, and 12-year pieces of the same record. The calculation charges 0.10% whenever SMH is bought or sold and excludes personal tax.
At every repeated or rearranged join, market signals use the real observations that preceded the incoming historical month. Portfolio value, withdrawals, inflation and strategy state continue without resetting.
Strategy in detail
Hold SMH when SPY beat T-bills in at least 5 of the past 12 completed months. Otherwise hold T-bills.
Semiconductor cycles can amplify broad economic and market weakness. A broad-market signal may therefore leave the sector before some long declines, while also missing sector rebounds.
Research has found that market direction can persist for months in many asset classes. See the 2012 time-series momentum study and a longer study across 67 markets. Those broad findings make the idea plausible; they do not validate this exact rule.
Can I trust these results?
My evidence-weighted estimate is about 54% that the main effect would persist in a new long market period. That makes the result plausible, not dependable enough to treat as a promise.
The direct test covers July 2000 through June 2026. It includes fund distributions, T-bill returns, inflation and the stated trading costs, but it is still one U.S. market record. The longer retirement histories repeat it or link 8–12-year pieces; they test difficult orders, not new historical evidence.
Treat the result as evidence about how the rule behaved, then compare it with simpler alternatives and your own ability to follow it through a bad period.
Methodology details for the nerds
How I ran the return comparison
The test starts with the amount shown on the graph. Each decision uses only information already complete at that month-end. Distributions are reinvested, T-bills earn the published three-month rate, and the calculation charges 0.10% of each risky asset bought or sold.
How I tested retirement withdrawals
The first withdrawal occurs before the first month's return. Spending then rises with U.S. inflation. One panel checks every starting month in order; the other links 5,000 histories from 8-, 10-, and 12-year pieces. A path counts only if it pays every withdrawal and remains above zero.
Data and important limits
The calculation uses adjusted fund returns plus U.S. government T-bill and inflation data. Personal tax, spreads beyond the stated cost, and investor behavior are not included. A real fund can differ from its index because of fees and tracking.
Keep comparing
See where this strategy ranks.
Use this result
See how much you need to retire—and compare what to invest in.
Enter what you have and what you want to spend each month. Compare three portfolios against the same market history. Your financial figures stay in your browser.
Build my free FIRE plan