Volume Eleven · The Crystal Ball, Graded

Forecasting by Horizon

Whether stock prices can be predicted, honestly answered: nearly impossible tomorrow, hard next quarter, and surprisingly arithmetic over a decade. One method per clock, with the achievable accuracy of each.

In this series: 1 Foundations · 2 What Moves Prices · 3 Practice Room · 4 Reading a Company · 5 The Dangerous Aisle · 6 Your Portfolio · 7 Case Study · 8 The Toolkit · 9 Crowd Psychology · 10 Crashes · 11 Forecasting · Cheat Sheet

01 · The Honest Premise

Forecasting works, but only on the right clock

Can a stock’s price be forecast? The truthful answer is: it depends entirely on the horizon. Tomorrow is very nearly a coin flip. Next quarter is a mood. But ten years out, returns are surprisingly forecastable, because over decades the noise cancels and only the arithmetic of the business remains.

Most forecasting failures are clock mismatches: using long-term tools to predict Tuesday, or chart squiggles to plan retirement. This volume sorts every tool in the series onto its correct clock, gives you one honest method per horizon, and is blunt about the achievable accuracy of each.

02 · The Predictability Ladder

Noise cancels; fundamentals compound

Every price move is a mix of two ingredients: noise (flows, moods, randomness) and fundamentals (earnings, dividends, valuation gravity). Their mix flips completely as the horizon stretches.

Slide the horizon and watch what actually drives the outcome

The paradox worth memorizing. The market is hardest to predict where most people try (this week) and easiest where almost nobody bothers (this decade). The crowd competes ferociously over coin flips and leaves the forecastable horizon nearly uncontested. That mispriced attention is a genuine edge for the patient.
03 · Short Term: Days to Weeks

Mostly noise, slightly leaning

Over days, prices are driven by order flow, news reactions, positioning, and randomness. No math reliably predicts them, but a few forces lean the odds slightly, and short-term trading is entirely about those leans plus ruthless risk control.

Momentum: follow, never predict

The one robust short-term pattern: things in motion tend to stay briefly in motion. Trend tools (the moving-average crossover from Volume 1) do not forecast; they confirm and follow, always late by design.

Levels: memory, not magic

Support and resistance mark prices where the crowd previously acted, and may again. They are probability zones for placing entries and stops, never predictions.

The reaction test

Volume 9’s master gauge is the best short-term tell that exists: strength that rallies on bad news, weakness that sinks on good. It reads what is already priced in.

The honest odds

Costs, spreads, and randomness eat tiny edges; study after study finds the large majority of short-term traders lose money over time. If you play here, the position-size math of Volume 1 is your entire survival, because the forecast never will be.

Rule of thumb. Short-term, you never forecast the price; you forecast your exit: where you are wrong (stop) and where you are paid (target), before entry. The market decides which. That is the whole craft.
04 · Mid Term: Months to Two Years

The hardest horizon, tamed by scenarios

Over months, both engines from Volume 8 run at once: earnings move and the multiple moods around. Getting both right is why this horizon humbles professionals. The tools that help: the direction of earnings revisions (are analysts raising or cutting?), the cycle location (Volumes 2 and 9), the Fed’s direction, and the multiple’s room to re-rate or compress.

Because point forecasts fail here, professionals think in scenarios: a bear case, a base case, and a bull case, each with a price and a probability. The payoff is an expected value and, more usefully, a risk/reward shape. Build one:

The stock trades at $100 today. Sketch your one-year scenarios.

$70
$112
$150
25%
25%
Practical notes. The discipline is in the bear case: make it genuinely ugly, not politely mild. A trade is attractive when the expected value clears today’s price and the upside distance is at least twice the downside (Volume 1’s 2:1 rule, reborn at a longer horizon). Write the three scenarios in the journal, then grade yourself when the year ends; that grading, repeated, is how forecasting skill is actually built.
05 · Long Term: Five Years and Beyond

The only truly forecastable horizon

Over a decade, a stock’s return stops being a mystery and becomes arithmetic, a decomposition John Bogle used for fifty years:

return ≈ dividend yield + earnings growth ± multiple change

Two engines are the business (yield and growth). The third is the crowd (what P/E people will pay at the end versus now). Over one year the crowd engine dominates and ruins every forecast. Over ten, it gets divided by ten while the business engines compound. Set the dials:

Forecast a decade the Bogle way

1.5%
6%
24
17
10

Where the inputs come from

Yield is printed on any quote page. Growth: start from history and the analyst range, then haircut it, because growth fades. Ending P/E: gravity points toward long-run norms (roughly 15 to 20 for the US market historically), higher for durable quality, lower for cyclicals.

Why this works when nothing else does

The crowd’s engine, the multiple, is bounded and mean-reverting, so time dilutes it. The business engines are not bounded; they compound. Ten years of compounding beats ten years of mood, almost every time.

Its limits, honestly

It forecasts diversified indexes and stable franchises far better than young or story stocks, whose growth input is a guess. And it gives a center of gravity, not a promise: actual decades land in a band around it, sometimes a wide one.

The case-study echo

Run Volume 7’s Microsoft through it: ~1% yield, your growth estimate, P/E 23 drifting wherever you believe. The whole bull-bear argument compresses into which dials you trust. That is what a real forecast disagreement looks like.

06 · Which Tool Belongs to Which Clock

The matching table

Days to weeks

Candles, support/resistance, volume, the reaction test, momentum following. From Volumes 1 and 9. Accuracy: a slight lean at best; risk control does the real work.

Weeks to months

Moving averages and trend structure, relative strength versus the index, crowd-state diagnosis. Volumes 1 and 9. Accuracy: modest; trends persist until they do not.

Months to two years

Earnings revisions, cycle location, Fed direction, multiple room, scenario trees. Volumes 2, 8, 9, and this page. Accuracy: genuinely hard; ranges only.

Five-plus years

The Bogle decomposition, valuation gravity, the five business questions. Volumes 4, 8, and this page. Accuracy: the best available in finance, for diversified and durable assets.

Clock-mismatch errors

Using P/E to trade a Tuesday. Using a chart pattern to plan retirement. Selling a decade-long holding over a monthly candle. Every one is a category error, and each is committed daily by millions.

The stop-clock rule

Decide the horizon before the trade, then only consult that clock’s tools. Checking daily charts on a ten-year position is how good forecasts get abandoned at the worst moment.

07 · Forecasting Hygiene

How good forecasters actually operate

Research on elite forecasters (Tetlock’s “superforecasters”) found their edge was not intelligence or information. It was habits, and every habit is copyable.

1

Ranges, never points

“Between $90 and $130, centered near $110” is a forecast. “$117” is a lottery ticket wearing a suit.

2

Probabilities, not certainties

Think 60/40, not yes/no. The words “always” and “cannot” are how forecasters go broke.

3

Update without shame

New evidence moves the estimate a notch, not to zero or to certainty. Stubbornness and capitulation are the same error at different speeds.

4

Keep score in the journal

Write the forecast, the reasoning, and the probability. Grade it later. Volume 3’s journal is a forecasting gym; ungraded predictions teach nothing.

5

Start from the base rate

Before any clever analysis, ask what usually happens: most years are up, most breakouts fail, most growth fades, most forecasts of doom and of miracles both miss. Adjust from there, modestly.

6

Beware anyone selling certainty

Confident point-forecasts about next month are entertainment products. The louder the certainty, the better the television and the worse the track record.

08 · Rules of Thumb

The forecasting survival kit

1

Match the tool to the clock

Charts for weeks, cycles for quarters, arithmetic for decades. Never cross the streams.

2

Short term: forecast your exits, not the price

Stop and target written first; the market grades the paper.

3

Mid term: three scenarios beat one conviction

And the bear case must be honestly ugly, or the exercise is decoration.

4

Long term: yield + growth ± multiple

Three dials, one center of gravity. The crowd engine shrinks as years divide it.

5

High starting valuations forecast low future returns

Not next year, but reliably across decades. The price you pay is the forecast.

6

Growth fades; haircut every growth input

Trees do not reach the sky, and 20% growers become 8% growers on schedule.

7

The market forecasts you back

Whatever you predict is partly priced in already (Volume 2). Your real forecast is always “versus expectations,” never versus zero.

8

When your horizon changes, your forecast expires

A ten-year thesis checked hourly becomes a one-day trade with extra steps. Guard the clock you chose.

09 · Practical Notes and Further Reading

Where to go deeper

Forecasting is where humility and arithmetic meet, and its best books are about exactly that pairing.

Superforecasting · Philip Tetlock

The landmark study of who predicts well and why. Every hygiene habit above comes from its findings.

Common Sense on Mutual Funds · John Bogle

The home of the return decomposition, with fifty years of receipts showing it works on decade horizons.

Mastering the Market Cycle · Howard Marks

The mid-term horizon’s best guide: positioning by where you probably are, without pretending to know dates.

The Little Book of Valuation · Damodaran

For turning the long-term dials, growth, fade rates, and reasonable multiples, with actual craft.

Free data for the dials. Shiller’s Yale dataset gives a century of yields, earnings, and P/Es. Damodaran’s NYU pages publish growth and return estimates yearly. Your broker’s research tab (Volume 7) supplies the per-stock inputs. Everything this volume needs is public; the discipline is the only scarce ingredient.