Variance Ratio (Lo-MacKinlay)
Lo-MacKinlay variance ratio.
What it measures
The Lo-MacKinlay (1988) variance ratio: Var(k-period returns) / (k · Var(1-period returns)). Under a random walk the ratio is 1; values above 1 indicate positive autocorrelation (trending, momentum-friendly tape) and values below 1 indicate mean reversion. It is a direct statistical test of which strategy class the current tape favors.
References: Lo-MacKinlay 1988.
Point-in-time, leak-free
Like every QUANT_API feature, regime.variance_ratio is computed point-in-time: each value uses only data that had actually arrived at the timestamp you query — live or historical. No restatements, no backfills that quietly rewrite the past, no look-ahead. The value your backtest sees at a given stamp is the value the live API would have returned at that stamp. How we enforce this is documented on the methodology page.
Windows & transforms
The signal is computed over rolling windows; each window can be served raw or through a transform (z-score, percentile rank, delta…). Which windows and transforms you can query depends on your plan — the signal itself supports:
Plan & access
regime.variance_ratio unlocks on the Quant plan ($2,500/mo) and every plan above it. Every new account starts with a 14-day free trial of the Signal plan — no card required. The trial covers Signal-plan features, so you can evaluate the API end-to-end before upgrading to Quant.
Example call
Resolve the latest value for BTC (5m window, zscore transform — both available on the Quant plan):
curl -G https://api.quant-api.dev/v1/features/live \
-H "Authorization: Bearer fk_live_<your_key>" \
--data-urlencode "asset=BTC" \
--data-urlencode "features=regime.variance_ratio@5m:zscore"Same key works on /v1/features/historical for point-in-time backtesting — see the API docs.