Why Being Right About Copper All Year Proves Nothing

We do not normally write articles like this one. But a friend keeps telling me about someone he knows who has a knack for calling copper's up moves every week. Every big weekly run, apparently, this person saw coming. After hearing it enough times, I went and pulled the data.

The metal went from $4.3785 to $6.6475, a gain of 51.8%. Not one of the 252 trading sessions closed below that starting price. A year rarely moves in one direction this cleanly. That is exactly why anyone who was bullish looks like a genius right now.

Saying "up" every week was right about 60% of the time

Start with the simplest forecast there is: say copper goes higher over the week ahead, without looking at a chart, a headline, or anything at all. Now check how that call would have done starting from every single day of the past year.

Saying "up" was right 59.4% of the time over a calendar week and 61.1% over five trading sessions. So the number to beat is not 50%. It is roughly 60%, and getting there took no skill, no chart, and no opinion.

So when someone tells you they called copper all year, ask one question: what was your hit rate? Under 60% and a broken clock beat them.

A year of weekly calls is only 52 guesses

Saying "up" every week already wins about 60% of the time. So a forecaster is not trying to beat a coin flip at 50%. They are trying to beat 60%, and they have to beat it by enough that luck cannot explain the gap.

The reason is that 52 calls is a small sample. Even someone with no skill at all might get lucky across 52 calls.

So mathematically, even if they are right 63% of the time, there's a 1/3 chance it was just luck.

Only at 73% does it become rare, something that is probably not luck.

Even a good forecaster needs more than a year to prove it wasn't luck

Now turn the question around. Instead of asking whether a good track record of predictions could have been luck, ask the opposite: if someone truly can forecast copper, how long before their track record proves it?

How long this takes depends on how big their edge is.

Think of two runners. If one is much faster, a single race settles it. If one is barely faster, a single race proves nothing, because a bad start or a gust of wind explains the gap just as well. It takes many races before the faster one clearly pulls ahead.

Forecasting works the same way. A small edge needs a huge number of calls before it shows. A big edge shows up much sooner.

Look at the first row. If someone is right about copper going higher 65% of the time, that is a genuine edge, but it takes about nine years to prove it. At 70% it takes two and a half years. Even at 75%, which almost nobody sustains, it takes a full year.

So one good year proves nothing unless they are right 75% of the time.

Where this leaves us

None of this means copper is random, and none of it proves a person cannot forecast copper. It proves that nothing shown so far could tell the difference between a real forecaster and someone who says "up" every week and goes back to bed.