Most people will tell you to wait for a crash before you invest at a record high. I went back through 427 months since 1872 when the S&P 500 sat at or near a record, and tracked what waiting would have done.

The question isn't abstract this week. The S&P 500 set its latest record on October 6 at 7,818.93 and closed at 7,690.73 on October 19, 1.6% below it. A 20% fall from that record would put it near 6,255.

Below: how the test works, why it lands near a coin flip, what waiting cost when the crash never came, and the one input that moved the odds. Then three questions to ask before you decide.

Two investors start in the same month, with the S&P 500 within 3% of its record high. One invests everything at once. The other holds cash until the index falls 20% below that starting high, then invests. If the fall never comes, the second investor stays in cash. After ten years I compare what each one has. Dividends are reinvested, and the waiting cash earns interest.

The rule only uses what was known that month. That matters, because it's easy to make any rule look good by picking the perfect entry afterward. The data is Robert Shiller's monthly S&P 500 history. It gives 427 start months between January 1872 and June 2013, each with a full ten years after it.

I also ran 10% and 15% falls. I ran two versions of cash too: one that earns nothing, and one that earns 84% of the long-term rate. That 84% is today's 3-month Treasury bill (4.26%) divided by the 10-year (5.09%), as of October 16.

Waiting beat investing at once in 40% to 62% of the 427 starts. The range depends on how big a fall you waited for and what the cash paid. At a 20% fall, the one most people mean by a crash, it won 47% of the time with cash paid like today and 40% with cash paid nothing.

There's a simple reason it lands near the middle. A 20% fall from the starting high showed up within ten years in 195 of the 427 windows, or 46%. With cash paid like today, waiting won all 195, because the waiter bought at least 18% cheaper than the investor did. In the other 232 windows the fall never came, and waiting won only 7 of them. Those waiters spent ten years in cash.

So waiting wins when the fall comes and costs you when it doesn't. That turns “should you wait for a crash?” into two smaller questions: how likely is the fall, and what do you give up if it never comes?

The two outcomes aren't the same size. When the fall came and waiting won, the waiter finished a median 1.28 times ahead of the investor. When the fall didn't come and investing won, the investor finished a median 1.86 times ahead, or 2.71 times if the cash earned nothing. Across all 427 starts, the investor ended with a median 2.09 times their money after ten years. The waiter, with cash paid like today, ended with 1.68 times.

The gap is uneven because stocks have usually earned more than cash, so every year the waiter sits out costs the difference. The fall has to come early enough and cut deep enough to pay it back. When it did come, it took a median of 29 months to arrive.

Buying the fall also wasn't buying the bottom. In the 195 windows where the 20% fall arrived, the index dropped a further 20% from that trigger level in 99 of them, or 51%. A waiter who put everything in at the 20% mark was early about half the time.

Eight of the 427 starts, picked because you'll recognize the dates, show both sides. Waiting won four and lost four. They aren't a sample.

Look at December 1996. The dot-com bust was still ahead, and all it did was take the index back to about where it stood that month. A waiter held out for a 20% fall that never came, and sat in cash for ten years.

In October 2008, Warren Buffett wrote in the New York Times that investors who “cling now to cash are betting they can efficiently time their move away from it later.” It gets quoted against waiting. The part people leave out is the date. By then the S&P 500 was already 37% below its October 2007 level, on monthly averages. He was buying after a large fall, not at a record. His rule is that price is what you pay and value is what you get.

The test points the same way. I split the 404 starts that have a CAPE reading by how expensive the market was. CAPE is the cyclically adjusted price-to-earnings ratio, which divides the index by ten years of inflation-adjusted profits. When it was under 15, waiting beat investing at once a third of the time. At 30 or above, it won three times out of four.

That fits the mechanism. The more you pay for each dollar of profit, the less room for error you have if prices drift back toward their long-run average. In 22 of the 36 starts with a CAPE of 30 or more, investing at once was still underwater after ten years, even with dividends. Across all 427 starts, that happened 35 times, or 8%.

I'd weigh that top bucket carefully. Its 36 starts are really two episodes: August and September 1929, and the stretch from June 1997 to September 2000. Two episodes isn't a sample. Waiting didn't win every time even there. Nine starts between June 1997 and February 1998 never saw a 20% fall within ten years.

Today's CAPE is about 41.6, per Multpl's version of Shiller's series on October 8. In the data I used, which runs through September 2023, CAPE topped that level only between April 1999 and September 2000. One comparable stretch describes the past. It doesn't forecast anything.

I read the 18.6-year land cycle as a stage rather than a date. Smart X Terminal currently places this cycle at Peak, year 14 of 18.6, as of October 20.

So I checked how waiting did when a start fell in the late stage of earlier cycles. I dated four crises my framework flags at October 1929, January 1973, mid-1990 and October 2008. In the three years before them, waiting won in 82% of 66 starts, against 41% of the other 361. With two-year and four-year windows it was 77% and 74%.

It's the pattern my framework leads me to expect, which is a reason to distrust it. It rests on four episodes, and I drew the windows knowing when the crashes happened. Half the starts, 34 of 66, come from the 1929 run-up alone. The 1990 run-up was the weak one: waiting won in only 2 of its 14 starts.

What it can't tell you is whether anyone can spot such a window in real time. I can't show that the clock can, so I treat it as one input next to price. This is pattern recognition, not prediction.

If you're deciding whether to wait, the test points to three questions, in this order.

This test is aimed at me as much as at you. Last week I wrote that I don't plan to spend my reserves in one go. I'll stage it, a slice at a time. Staging is a way of waiting, and the numbers say waiting needs a better reason than “the market is high.” My reason is the price, the one input that moved the odds. If I'm wrong, the cost is the one in the second figure: finishing well behind.

I also ran the middle path, half in at the record and half when the fall comes. Its median finish was 1.99 times, against 2.09 for investing at once and 1.68 for waiting. I didn't test staging by calendar, so that result is about a different plan than mine. And that's my own plan. Yours depends on your situation.

I could be wrong in ways this test can't show. It covers one country's stock market, and the US is the one that kept winning. Japan's Nikkei 225 peaked in December 1989 and didn't close above that level until February 2024, which would have made waiting look brilliant there. A US-only record flatters whoever invested at once, and I don't know by how much.

Two design choices matter too. Monthly averages smooth out the sharpest drops. In February and March 2020 the S&P 500 fell 34% on daily closes but only 19% on monthly averages. So the test undercounts how often a 20% fall arrived, and 47% is a little low. I also measured the fall from the starting high. If the trigger instead rises with every new high, waiting won between 47% and 51% across the three fall sizes.

Quick note: the readings and figures here are as of October 20, 2026, when I wrote this — markets move, so for the live picture I keep Smart X Terminal updated in real time.

To see the cycle-position reading next to your own answers to the three questions, open Smart X Terminal.

So the next time someone tells you to wait for a crash, ask them how big a fall, and by when. Next Tuesday is Election Day in the US. I'll go through what the record shows for the S&P 500 in midterm years, and whether the pattern people quote holds up.

If you find this article interesting, please let us know your thoughts down below here. We also have have other similar articles on our website too.

Sharing with you the tools I’m using at the moment.

📚 Books I read: https://smartxcapital.com/books

I’m building something for investors who want to move smarter — not faster.

A place where we’ll track these cycles together, share real-time insights, and learn how to invest with the cycle — not against it. I’ll be offering workshops, tutorials, and in-depth guides to help you build a timeless investing system that grows through every boom and bust.

📚 The Smart X Capital Platform is coming soon — a place to learn, connect, and stay ahead of every major market cycle using data, history, discipline and our community.

👉 Join the Smart X Capital Platform Waitlist

Because when every major cycle converges — the prepared don’t panic. They profit.

Talk soon,

Ace — Smart X Capital’s Founder