How people fool themselves with charts

After this lesson you can: spot the four ways charts mislead.

The dashboard gives you dozens of charts and the ability to overlay anything on anything. That is power, and power to mislead yourself. Five traps.

1. Linear versus log

On a linear scale, bitcoin's history looks like a vertical cliff and every past cycle looks flat. On a log scale, each cycle's percentage moves are visible and comparable. Neither is "right"; know which you are looking at and why it feels the way it does. The dashboard lets you toggle.

The same illustrative growth series on a log scale, where steady growth is a straight line. — The same series on a log scale: equal percentage moves take equal space.

The same illustrative growth series on a linear scale, where only the recent years look like anything. — Illustrative series on a linear scale: the early years look flat, the recent ones look like a cliff.

2. Overfitting

Any curve can be fit to any history with enough parameters. Rainbow charts, power-law corridors and stock-to-flow all fit the past. Fitting is not forecasting. The test of a model is its out-of-sample predictions, and most bitcoin models have failed theirs.

3. Survivorship

You hear about the indicator that called the last bottom. You do not hear about the hundred that did not. Indicators that worked three times are not proven; they are undefeated with a small sample.

4. Narrative after the fact

Every morning, the news explains yesterday's move: "bitcoin fell on ETF outflows", "bitcoin rose on rate-cut hopes". These are written after the price moved and would have been written the other way had it moved the other way.

5. Correlation on the overlay

![Two unrelated illustrative series that both rise over time and therefore look correlated.](mk-fooling-yourself-f3.svg "Two things that both go up will "correlate". Illustrative series, no relationship between them.")

Overlaying M2 or the S&P on bitcoin will often look convincing. Two things that both go up over time will correlate. The dashboard's overlay tool is for asking questions ("do these move together in this period?"), not for proving causes.

The habit

Use the data to understand what happened and where things stand. Be suspicious of anything that tells you what happens next, including your own instinct after you have looked at a chart for an hour.

Course checklist: what you can now do

Tick each one you have actually done. The course is finished when they are all ticked, not when the quiz is.

01Write down the largest drawdown you would sit through before you hold any
02Find one spot ETF's holdings page and one on-chain dashboard
03Spot one log-scale and one survivorship-bias trick in a chart this week

MINI CHALLENGELog scale

On a log chart, a move from $10 to $100 takes the same vertical space as a move from $100 to what?

dollars

CHECK YOURSELF

1/3 Why does the log scale matter for bitcoin charts?It hides crashes so the trend looks smootherIt makes prices look lower and less frighteningIt shows percentage moves, so cycles can be compared

2/3 What is the test of a price model?How many people share and discuss itHow its predictions perform on data it has not seenHow well it fits the data it was built on

3/3 What is survivorship bias in this context?Older charts are more accurate than newer onesYou hear about the indicator that called the last bottom, not the many that failedOnly indicators that survive a bear market are reliable

pick an answer for each question