Indicator Lag And Whipsaws

The dial you are really turning

Shortening an indicator's period feels like making it sharper. It is not. It is moving a slider between two failure modes, and both are always present.

• Short settings. Less delay, more signals, and a much higher proportion of them reverse before going anywhere.
• Long settings. Fewer signals, most of the range-bound noise filtered out, and confirmation arriving after a meaningful part of the move has gone.
• There is no setting in the middle that avoids both. Optimising a period on past data moves you along the slider, it does not lift you off it.
• Different instruments sit differently. A period that is sensible on a slow-moving pair is jumpy on a volatile one.

Any claim that a particular setting eliminates whipsaws should be read as a claim that the setting has been fitted to a sample where whipsaws happened to be rare.

Checking the regime before trusting the line

Because lag is tolerable in a trend and expensive in a range, the highest-value thing you can do is establish which one you are in before acting on a lagging signal at all.

Trend strength measures exist for this, and so does looking at the chart: a market making higher highs and higher lows is trending, and a market oscillating between two levels is not. Either way, a written condition beats a judgement made mid-candle. One filter, checked before the session, removes the majority of signals in exactly the conditions that generate them.

A second, cheaper filter is time. A large share of false crossovers occur during the thin hours, when a handful of orders can push price across a line that nobody is defending. Restricting signals to the hours when the market you trade is actually busy removes a meaningful proportion of them at no cost beyond patience.

What to do with the whipsaws you still take

Some will get through, because no filter is complete and a range does not announce itself at the start. Three things limit the damage.

First, a stop that exists before the entry, so a false signal has a bounded cost rather than an open-ended one. Second, size derived from that stop rather than from conviction, so a run of small false signals stays a run of small losses. Third, a limit on how many times you will re-enter the same idea in a day, because the whipsaw pattern punishes persistence specifically. The arithmetic behind all three is in the risk management guide, and the short version is that a signal being wrong is survivable while a signal being wrong at full size is often not.