On Sept. 23, Goldman Sachs lowered its year-end target on the S&P 500 for the fourth time in 2022. Goldman’s new estimate is 3600, a number that’s 29% below the 5100 mark it was forecasting as late as mid-February.

In one sense, you can’t blame Goldman for failing to foresee how forces would suddenly join hands to drag equities downward. It was making a short-term forecast, and unsustainably balmy conditions like those reigning in late 2021 can last for a couple of years. Indeed, the trade winds and currents could have flowed as the mega-brokerage contemplated for a while. The rub: When Goldman unveiled its 5100 number in November of 2021, big caps sported valuations only equaled in the 1998 to 2000 bubble; profit margins hovered gigantically above their long-term norms; and real interest rates were negative, a rare circumstance that kept price/earnings ratios (P/Es) extremely rich on top of outsize earnings.

In positing 5100, Goldman was projecting that companies would squeeze even more profits from each dollar of sales this year, and that Treasury yields rise but remain below the course of inflation. The confluence of those highly unusual forces would keep the S&P just as richly priced at the close of 2022 as at its all-time highs when Goldman laid its marker.

The original Goldman forecast suggested that in late 2021 and early 2022, the S&P 500 was a good deal. But the basics that determine long-term value showed that big caps were the opposite of a bargain. Goldman was auguring the happiest of happy outcomes across the board to get to 5100. A more reasonable forecast was anticipating what’s actually happened, and was bound to happen later even if the party raged through 2022: a reversion to the mean in margins and multiples that also brought valuations from seldom-seen highs closer to historical averages.

Let’s examine the reasons why stocks were hugely overpriced when Goldman saw them going substantially higher, and why getting to 5100 would keep equities way above the fundamental value where they’re destined to return, via the inevitable downshift that’s now underway. So far, that reckoning has pounded the S&P by 24% from the time Goldman guided that the index would end 2022 nearly double digits higher.

Goldman, like many of Wall Street, wasn’t bothered by last year’s immense valuations

I don’t mean to single out Goldman. Many big banks set year-end 2022 expectations at 5000 or higher late last year, including Wells Fargo, RBC, Credit Suisse, and Citigroup. Despite valuations at levels so towering that in the past they’ve proved temporary at such extremes, Goldman and the other banks sighted still more gains in 2022.

A highly respected valuation metric was already flashing red when Goldman made its 5100 call. It’s the cyclically adjusted price-to-earnings or CAPE ratio developed by Yale professor and Nobel laureate Robert Shiller. The CAPE is so valuable because it uses as its denominator not current earnings, but a 10-year average of GAAP trailing, four-quarter net profits, adjusted to current dollars. When earnings hit an ephemeral spike, the regular P/E looks artificially low, suggesting equities are fairly priced when they’re really superexpensive.

That was the picture in November 2021 when Goldman billboarded 5100 for this year. Then the CAPE exposed that big caps were shockingly overvalued. I make an adjustment to official CAPE earnings by increasing the 10-year average by 10%; that’s because the measure raises past earnings per share (EPS) only for inflation, and doesn’t include potential “real” gains that typically add another couple of points a year. At the time of the 5100 call, the S&P was trading near an all-time summit at 4700. Using my earnings metric that adds 10%, the CAPE was reading between 33 and 34 (or almost 39 by the official yardstick). Two factors explained the dangerously high ratio. First, earnings were so inflated that the balloon was poised to pop. Second, the price being paid for each dollar of earnings when applied to a “normalized” EPS number—or what companies were really going to deliver once profits returned to normal—was too hot not to cool down.

The CAPE had only hit 33 to 34 in a single period over a century and a half, stretching from January 1999 to September 2000 during the tech bubble. From the peak of that span, stocks fell 43% by early 2003. The CAPE’s starting position, in fact, exhibits a strong record in gauging where stocks will go over the following decade. Buying at a big CAPE usually means tough going ahead.

Had the S&P ended 2022 at Goldman’s 5100, the CAPE then would still have been clocking around 33 by my adjusted CAPE measure, showing the index was poised for a kind of Appointment in Samarra, a deep slide that hadn’t started yet, but was inescapable.

Still another key metric was signaling a downward shift, but apparently didn’t cloud Wall Street’s sunny outlook. One of Warren Buffett’s favorite measures of whether stocks are cheap or dear, nicknamed “the Buffett Indicator,” is the ratio of the market value of all stocks to the GDP. The idea is that the total worth of equities can’t grow to supersize proportions versus the national income that fuels earnings. Since 1998, that benchmark has averaged 110%. It hit 127% at the end of 2019, when the S&P and earnings touched all-time highs. But at Goldman’s prognostication moment in November of last year, the Buffett Indicator stood at 180%, two-thirds above its 24-year average. By Fortune’s calculations, if the 5100 had happened by the end of this year, the measure would be just a tad lower at 172%. In other words, cruisin’ for a bruisin’.

The high-flying valuations of late last year ordained extremely poor returns in the years ahead. That was the view from Research Affiliates, a firm that oversees investment strategies for around $180 billion in mutual funds and ETFs, and employs many of the best economic minds in the business. Late last year and in early 2022, RA reckoned that investors were buying at P/Es so juiced that an S&P 500 portfolio would simply match inflation over the following 10 years, and generate a real return of zero. Exerting the downward pressure, RA determined, would be major compression in severely stretched multiples.

Huge profit margins were much more likely to fall than increase

Here’s a summary of the Goldman analysis behind the 5100 forecast. Chief U.S. equity strategist David Kostin saw operating earnings per share rising by 8% over year-end 2021 to $226, and reasoned that the P/E-based on net earnings would remain constant at between 20 and 21. In TV interviews from late 2019 and early 2022, Kostin argued that the key lever pushing higher prices was the durability of already ultrawide margins that he expected to expand. “It’s not a valuation-led story but a profits-led story,” he declared. “This year you had supply-chain disruptions, hard times for companies to find employees, the Delta variant, and [rising] commodity prices, and yet companies were still able to raise margins to record levels. I’m forecasting that this year margins will also be increasing.” The 8% earnings increase, coupled with a rise in margins, would send the S&P 9% higher by the close of 2022, to hit the 5100 target.

But at the end of 2021, S&P profits already looked overripe. Operating earnings reached $208 in Q4, a jump of 32% from the pre-COVID peak in the final quarter of 2019, which was already half-again the 2016 number. The source was an explosion in profitability, not rampaging sales. For 2021, operating margins ran at 13.3%. That’s 30% above the 10-year average of 10.3%. To arrive at 5100, Goldman spyglassed an increase in that stupendous, never-before-seen number to around 13.7%. Once again, it could have happened. These out-of-the-ordinary trends can go on a long time. But a more sober reading, and one that offered higher odds, would have foreseen a shift back toward a more normal portion of revenues going to profits.

That’s what occurred. First, instead of rising 8% as Goldman thought, total operating profit will end the year flat at $209 according to the analysts’ forecasts, and their numbers are coming down, meaning the final tally could be a lot lower. Second, operating margins are falling fast, cratering to 10.9% in Q2, an 18% decline from the 2021 average. That pattern constitutes a powerful pullback toward the mean, and probably indicates that the 13%-plus margins posted at the market’s peak were a one-off phenomenon.

The risk of falling P/Es was substantial

In the TV interviews in late 2021 and early 2022, Kostin correctly stated that by all valuation metrics, stocks were extremely expensive at P/Es of 20 to 21 times net earnings. Although 20 to 21 times doesn’t sound overly high, keep in mind that investors were awarding those still excellent valuations to outsize profits. Once again, the CAPE showed that investors were paying 33 times a reasonable EPS number. That’s the multiple that counts and foreshadows what’s going to happen. Kostin also identified the reason shares looked so pricey: incredibly low interest rates. In November of 2021, the 10-year Treasury yield treaded at a lowly 1.6%. The “real” rate after subtracting expected inflation over the next decade, was minus 1%. Kostin said that he expected the long bond yield to rise to 2% by the end of 2022, but would still slightly trail inflation, keeping the real rate in negative territory, though not as deeply negative as in 2021.

Offsetting the rise in real rates would be a decline in the bonus investors expect from stocks over safe government bonds, known as the equity risk premium or ERP. Kostin thought that a continuation of good times characterized by rising wages, increasing consumer confidence, and falling unemployment would boost the ERP. As a result, Kostin reasoned, the P/E in 2022 would remain at the 2021 level of around 21, still elevated considering the jaw-dropping EPS numbers.

Why his outlook now appears overly optimistic: Negative real interest rates, like 13%-plus profit margins, are a passing phase. The rare times inflation-adjusted yields fall below zero generally occur when the Fed and banks flood the market with cheap credit, so that the supply of dollars available for loans outstrips demand—the situation that caused the episode of negative real rates from the beginning of the COVID crisis. Of course, neither Goldman nor virtually any other bank predicted the size of the inflationary outbreak or the severity of the Fed’s tightening to rein in raging prices. It was the latter that sent Treasury yields soaring starting in the spring. The real rate went positive in April, and as of early October sits at around 1.5 points. That’s a gigantic swing of 2.5 points from this time last year.

In its reports, Goldman correctly points to how declines in real rates lift multiples, and how increases in inflation-adjusted borrowing costs pressure P/Es. Had it seen real rates going sharply positive instead of staying negative, Goldman would have predicted a big drop in stock prices. But the bank believed the history-defying trend enabling a 33 Shiller multiple would continue.

Getting back to normal meant lower multiples, just as it meant much reduced margins. That’s what today’s big selloff is all about. Instead of the $226 in operating earnings Goldman was forecasting into February, the number was almost 10% lower at $205 in Q2, and appears to be on track to fall further. Net earnings, the number used by Shiller, are pretty much flat at $192. But the Shiller measure shows EPS are still overcooked. The CAPE suggests that based on fundamentals, EPS will retreat to around $150. At 3624 as of midmorning, Oct. 10, the S&P was trading at over 24 times $150. I agree with Kostin that a reasonable multiple is between 20 and 21. So let’s take the midpoint at 20.5. A P/E of 20.5 times $150 in earnings puts the S&P fairly valued at around 3100. That’s a long, long way from the Goldman wrong-way call of 11 months ago, echoed by most of Wall Street.

This article is part of Fortune’s quarterly investment guide for Q4 2022.