What Transfers and What Does Not

What an indicator actually is

A transformation of past price into a line. That is the whole definition. Nothing in the
arithmetic reaches into the future, and any predictive power comes from elsewhere: enough
participants watching the same line and acting on it that their collective behaviour makes the
level meaningful.

This is not a dismissal. A level that many people trade is a real feature of a market, and
knowing where those levels sit tells you where behaviour is likely to cluster. It is simply a
different claim from the one usually made, and it fails in a predictable way: when the crowd
stops watching, the level stops working.

Four assumptions crypto breaks

Each of these has a concrete effect. Patterns defined by an opening gap lose their meaning
without an open. Anything relying on a floor under a decline has no floor. Anything calibrated
on volume inherits whatever a venue chose to report.

What still transfers

Support and resistance, in the weak form: prices where a lot of trading happened previously
are prices where a lot of participants have a reason to act. Trend, as a description of what
has been happening. And volatility measures, which describe the distribution of recent moves
and are among the few things a chart genuinely measures rather than infers.

What transfers badly is anything requiring a stable relationship between volume and price,
anything assuming a mean the market reverts to, and anything with parameters tuned on a
different asset class. Those are not slightly less reliable here; they rest on conditions that
do not exist.

Whose chart are you actually reading

A chart of a traditional equity is drawn from one venue’s trades. A crypto chart is drawn from
whichever venue or aggregate the platform chose, and the choice is rarely stated on the chart
itself. Two platforms can show different wicks, different highs and different volumes for the
same asset over the same hour.

This matters more than it sounds, because a great deal of technical analysis keys on precise
levels. A support line touched three times on one data source may have been touched twice or
four times on another. A liquidation cascade on one venue prints a wick that never existed
anywhere else, and any level derived from it is an artefact of that venue’s order book rather
than a property of the market.

The practical handle is to find out which source your chart uses and stay on it. Consistency
matters more than picking the correct one, since a level is only meaningful if the same series
generated it and will test it. Switching platforms mid-analysis quietly changes the data under
a conclusion that was drawn from something else.

Backtests, and why most of them are worthless

A strategy tested on the data used to design it will look excellent, because the parameters
were chosen to make it so. This is not fraud in most cases, it is a subtle error that requires
deliberate effort to avoid, and the effort consists of testing on a period the design never
saw.

Crypto makes it worse than usual. The available history is short, dominated by a few enormous
moves, and a strategy fitted to those moves is fitted to a handful of events rather than to a
regime. A result quoted without the period, the fees and the position sizing is not a result.

Fees and slippage eat most edges

A strategy showing a small positive expectancy before costs is usually negative after them.
Every trade pays a fee, crosses a spread and suffers some slippage, and a system that trades
frequently pays all three repeatedly. This is where most apparently profitable rules go to die,
and it is routinely omitted from the presentation.

The practical consequence is that trading less is a genuine improvement to almost any system.
That is an unglamorous conclusion, which is why it appears so rarely in material selling an
approach.