Is Price Just Random? What 502,988 Backtests Say About Market Predictability

The two wrong answers

'Price is 100% random, nobody can predict it' and 'the market is a code you can crack' are both wrong, and our data shows why. Across 741 assets and 502,988 backtests, 57% of assets had some indicator that beat buy-and-hold out-of-sample — so it's not pure noise. But any single indicator beat holding on only a small minority of assets, and only 26% of indicator combos beat it at all.

Faint structure, hard to exploit

That's the honest middle ground: markets are not a random walk in the strict sense — there's faint, real structure — but it's weak, asset-specific, and easily eaten by trading costs and curve-fitting. The median 'best' strategy we found on an asset earned a Sharpe of about 0.61 — a modest reward for a lot of work, not a money printer.

Statistical tests people cite (ADF, Hurst, variance-ratio) tend to agree: most price series are close to a random walk with a small, unstable amount of structure. 'Close to random' is not 'exactly random' — but it's far from 'predictable.'

What it means for you

If price had real, easy structure, our 358 indicators would beat buy-and-hold far more than they do. They don't. So treat any claim of a reliable, universal signal with suspicion, expect edges to be small and specific, and remember that on many assets — especially trending indices — the honest answer really is 'just hold.'