# When Is the Best Time to Buy a Stock? Two Honest Answers

> We tested timing with 86 years of data: what predicts returns at one year vs ten, what perfect timing is worth, and the 17 years Microsoft made buyers wait.

- Author: Barebone Research, Barebone AI
- Published: 2026-04-21
- Canonical: https://barebone.ai/resources/best-time-to-buy-a-stock-two-answers
- Publisher: Barebone AI (https://barebone.ai)

---

## The Question With Two Right Answers

Sooner or later, every investor types five words into a search bar: *best time to buy stocks*.

The results split into two camps. One talks charts — trends, levels, sentiment. The other insists timing is a fool's errand: nobody rings a bell at the bottom, just buy.

Almost nobody says the true part out loud: both camps are right. They're answering two different questions that happen to share a sentence — and the expensive mistakes come from mixing them up.

So we tested it properly. We used Barebone to pull every price series in this piece and re-run the classic timing research: an 86-year predictability study, two *Journal of Finance* anomalies, a five-investor experiment, and the most instructive "great company, wrong price" purchase of the modern era.

The receipts up front: **the best one-year predictor Vanguard could find explains 12% of returns. Twenty years of perfect timing is worth $15,522. And Microsoft once made buyers wait 16 years and 10 months to get back to even.**

Each number answers a different version of the question. That's the whole lesson.

## One Market, Two Clocks

Benjamin Graham called the short run a voting machine — every buy and sell a vote, prices measuring mood — and the long run a weighing machine, where prices settle on what a business actually earns. In 2012, Vanguard's research group, probably without meaning to, measured the metaphor.

Its researchers took sixteen signals investors actually use — P/E ratios, dividend yields, GDP growth, past returns, even trailing U.S. rainfall as a sanity check — and measured how much of the variation in the *next year's* return, and the *next decade's*, each explained ahead of time across 1926–2012. The score is an R², from 0 (useless) to 1 (clairvoyant).

The one-year column is a graveyard. Most signals scored close to zero. The best of all sixteen — a blend of dividend yield and trend earnings growth — managed just **0.12**. The paper's own verdict: stock returns "are essentially unpredictable at short horizons."

The ten-year column is a different machine entirely:

<Chart name="LessonSevenPredictabilityChart" />

Two things are worth staring at. Valuation — the price you pay per dollar of earnings — explained **0.43** of ten-year return variance (using P/E10, the smoothed multiple known as the Shiller CAPE; 0.38 for a plain trailing P/E). Nothing else came close. And rainfall (0.06) outscored trend GDP growth (0.05) and trailing one-year returns (0.00): last year's weather predicted the next decade better than last year's market did.

Read honestly: over months, nothing reliably works. Over a decade, the only signal that ever cleared 0.4 is the price you paid — and even that leaves 57% of the outcome unexplained.

So the question splits. On the short clock, the crowd is everything. On the long clock, the price is everything.

## The Trader's Answer: Read the Crowd

On a weeks-to-months horizon, valuation is statistically useless; the crowd is the only game available. The foundational rule — don't fight the tape — is one of the few short-horizon effects that survived peer review. In 1993, Jegadeesh and Titman showed that buying the U.S. stocks that had beaten the market over the prior six months, and shorting the laggards, earned about **1% per month** over the following six, across 1965–1989 — and all sixteen variations they tested came out positive. Momentum is real: recent winners keep winning for a few more months.

The practical toolkit traders have built on top of that fact comes down to three checks.

**Check one: the levels.** Support and resistance are prices where a stock has repeatedly stopped falling or stalled rising — memory written into the chart by previous buyers and sellers. The cleanest test came from New York Fed economist Carol Osler, who took the levels six trading firms published to clients in 1996–98. Prices bounced off the published levels **60.8%** of the time, versus **56.2%** at arbitrary nearby levels, and the effect held for at least five trading days. The technicians are not hallucinating — but notice the size of the edge: 4.6 percentage points better than chance. A lean, not a law.

**Check two: the catalysts.** A level only holds in the absence of new information, and scheduled events are new information with a date attached. Meta closed at $323.00 on February 2, 2022, reported earnings that evening, and closed the next day at $237.76 — **-26.4%** in one session, straight through every support level drawn that year. Earnings, launches, policy decisions: the moments the chart's memory gets overwritten. No floor survives a bad enough number landing on it.

**Check three: the sentiment.** CNN's Fear & Greed Index compresses seven market internals — the S&P 500 versus its 125-day average, net new 52-week highs, up-versus-down volume, put/call ratios, junk bond spreads, the VIX, and the stock-bond return gap — into one 0–100 fear-to-greed score. For a trader it's a momentum gauge. But fear cuts two ways depending on its species: when we tested every VIX panic spike of the past 30 years, buying *containable* panics won 93–100% of the time within six months, while buying systemic crises was a coin flip. The dial tells you what the crowd feels, not whether it's wrong.

When trend, calendar, and sentiment align, a trader calls it a setup. Even then it's a weighted coin — and the weighting expires. In 1985, De Bondt and Thaler ran the momentum logic out three years and found it inverts: the market's biggest past losers beat it by **19.6%** over the following 36 months while past winners lagged by 5%. Momentum is a sprint. The moment a trade quietly becomes a "position," the evidence that justified it hasn't just expired — it has flipped sides.

## The Investor's Answer: Weigh the Price

If your horizon is years, timing collapses into valuation: is the price below what the business is worth? The discipline is to demand the answer three ways.

**Cheap against its own history.** The P/E ratio is what you pay per dollar of annual profit — a stock at $100 earning $5 a share costs 20 times earnings. Compare today's multiple to the company's own five-year average: a business that normally commands 30x and now trades at 18x is on sale *if the business is intact* — the clause that is the entire game, because multiples also compress when something is genuinely breaking.

**Cheap against its peers.** The same multiple, set against direct competitors. A discount to the sector can mean an overlooked bargain — or the market correctly diagnosing the weakest business in the group. A relative discount is a reason to investigate, never a reason to buy.

**Cheap against its own cash.** A DCF — discounted cash flow — is the formal version: project the cash the business will generate, shrink each future year back to today's dollars, and sum it. If the market price sits below that sum, the stock is undervalued. The footnote is arithmetic: in a simple perpetuity model, lowering your discount rate from 9% to 8% with 3% growth raises "fair value" by 20%. A DCF is a range wearing the costume of a number — one vote of three, not an oracle.

When all three say cheap, the timing is right — whatever the chart looks like that week. Warren Buffett compressed thirty years of mistakes into the standard in his 1989 letter to Berkshire shareholders:

> "It's far better to buy a wonderful company at a fair price than a fair company at a wonderful price."

The corollary investors keep relearning the expensive way: a wonderful company at the wrong price is still a bad investment. The proof has a name.

## The 84x Lesson

On December 27, 1999, Microsoft hit its dot-com record close: **$59.56**, adjusted for its later stock split. Its fiscal-1999 earnings were $0.71 a share on the same basis — so buyers that day paid roughly **84 times** annual profit. Not because the business was doubted, but because it wasn't: the price prepaid decades of assumed success.

What makes Microsoft the perfect specimen is that the assumed success *happened*. Revenue grew from $19.7 billion in fiscal 1999 to $85.3 billion in fiscal 2016; earnings per share roughly tripled; the antitrust wars were survived and the enterprise and cloud franchises built. The weighing machine confirmed nearly everything the 1999 bulls believed.

And the stock:

<Chart name="LessonSevenMicrosoftChart" />

| | Dec 1999 / FY1999 | Oct 2016 / FY2016 | Change |
|---|---|---|---|
| Revenue | $19.7B | $85.3B | +332% |
| Diluted EPS (split-adjusted) | $0.71 | $2.10 | +196% |
| Share price | $59.56 | $59.66 | **+0.2%** |

Microsoft did not close above its 1999 peak until **October 21, 2016** — sixteen years and ten months, including a 74.6% drawdown to $15.15 in March 2009. Reinvested dividends soften the sentence to about fourteen and a half years. That is the price of buying one of the best businesses ever built at 84 times earnings.

Every check in the previous section exists to block that purchase: against its own history, against any peer, against any defensible cash flow projection, Microsoft in December 1999 failed all three — loudly. The weighing machine vindicated the company and still buried the buyer for half a career.

On the long clock, the price you pay *is* the timing.

## What Perfect Timing Is Actually Worth

One version of the timing question quietly costs the most: not *which* moment to buy, but whether to wait for a better one.

The Schwab Center for Financial Research ran the clean experiment: five investors, $2,000 each at the start of every year for twenty years (2005–2024), all destined for the S&P 500. One buys the exact low of every year. One invests the moment the money arrives. One splits the cash into twelve monthly pieces. One buys the exact peak every year. And one waits in Treasury bills for a better opportunity that never feels like it arrives.

<Chart name="LessonSevenTimingChart" />

Twenty consecutive years of *perfect* timing was worth **$15,522** over just investing immediately — about $700 a year for omniscience. The unluckiest buyer on earth, hitting the annual top twenty years straight, still finished with **$151,343** — more than three times the cash-sitter's $47,357. Across all 80 rolling 20-year windows since 1926, the finishing order was identical in 70; investing immediately never once came last.

The mechanism is boring: the market finished a random 12-month stretch higher 75.6% of the time. Waiting for a better entry means repeatedly declining a coin flip weighted three-to-one in your favor.

On the long clock, the expensive mistake isn't buying at the wrong time. It's waiting for the right one.

## Where Both Answers Fail

Every framework owes you its failure modes, so here are this one's.

**The trader's edge is thin, and it decays.** Published support levels beat chance by 4.6 points, for days, not weeks. Momentum's 1% a month is measured before trading costs and taxes, which for a fast-turnover strategy are not a footnote. Short-clock advantages are small, perishable, and re-earned on every trade — which is why casual users tend to lose to industrial ones.

**The investor's tools have no clock at all.** Underpriced stocks can keep falling for years, and overpriced ones can keep rising — Microsoft looked indefensibly expensive long before December 1999. And the cheapness checks can be honestly passed by a dying business: a P/E below its average is a bargain only if the earnings survive. Cheap and getting cheaper is the most expensive pattern in value investing.

**The deadliest failure is using one clock's tools on the other clock's game.** Running a DCF to justify a two-week trade. Holding a broken momentum entry for three years — the trade that gets demoted to an "investment" the day it drops. The tools above are not wrong; in most blown-up portfolios they are simply cross-threaded.

## What This Means

The question "when should I buy?" has two honest answers, and the work is knowing which one you're asking.

**Decide your clock before your entry.** A months-long game trades the vote: trend, calendar, sentiment — an edge measured in single percentage points that expires in days. A decade-long game trades the scale: the entry *date* barely matters; the entry *price* compounds forever.

**Respect what the short clock actually offers.** The documented edges are tilts — 60.8% versus 56.2%, 1% a month before costs — not certainties. The question is whether your sizing assumes that, or assumes you're right.

**Respect what the long clock actually punishes.** Not bad dates — bad prices. Sixteen years and ten months is what 84 times earnings cost a buyer of one of history's great businesses. The three cheapness checks exist to stop you prepaying decades of success.

**And notice what the data answers loudest.** It isn't "when." Perfect timing was worth $700 a year; sitting out cost $103,986. The market quotes one price and runs two clocks — but on either of them, the position that reliably loses is the one on the sidelines, waiting for a certainty that 86 years of evidence says will never come.

---

*Data: Barebone | Sources: Vanguard Research, Forecasting Stock Returns (2012), Jegadeesh & Titman, Journal of Finance (1993), De Bondt & Thaler, Journal of Finance (1985), Osler, FRBNY Economic Policy Review (2000), Schwab Center for Financial Research (2025), Microsoft FY1999 and FY2016 Form 10-K (SEC EDGAR), Berkshire Hathaway 1989 shareholder letter | Data as of April 21, 2026*