Spot the lucky one: can you tell skill from luck in two years of returns?

Why is the real edge so hard to spot?

Because two years of daily ups and downs are louder than a modest edge. Each investor here takes the market's moves and swings of their own of about 14% a year. An edge of 8% a year adds up to about 16% over two years, while luck alone moves a no-skill investor up or down by about 20% over the same time. So the one with the edge often does not finish first, and a lucky one often does.

How were these investors made?

From a made-up market, with fixed seeds so every visitor plays the same rounds. All four investors in a round move with the market one for one (a beta of 1) and have the same size of day-to-day noise; one also gets a real edge, between 5% and 25% a year. After a pick, the page shows each investor's alpha (the yearly return left over once the market's moves are taken out) and t-stat, and, for the one with the edge, its p-value and whether the edge showed. All of them come from the same skill check Luck Filter runs on a real record, measured against that market. Notice the alphas of the investors with no edge: luck alone makes some of them look ahead.

What does this say about a real record?

That a good stretch on its own proves little, and a modest real edge can take many years to stand out. That is why the skill check asks how easily luck could have produced a result, and how much more data it would take to tell. It describes the past; it never predicts what comes next.