Where Price Goes Is Not How to Profit: Testing "Liquidity Magnet" Price Levels in Bitcoin Perpetual Futures

Traders often treat prior highs and lows as "magnets" that attract price. Using 2.1 million one-minute Bybit BTCUSDT candles (2021-2024), this paper tests the idea as a destination forecast and as a source of trades. The nearer of two candidate levels was reached first in 71.9% of 7,085 events, against 70.5% predicted by a driftless first-passage model using distance alone. An exact-distance mirror placebo on the same price paths left a 2.2 percentage-point residual, larger at previous-day highs and lows, but the effect was unstable across years, absent in trending conditions, and did not improve walk-forward out-of-sample forecasts over distance alone. No entry design converted the forecast into profit: the best earned 8.3 basis points per trade before costs, below a round-trip cost of about 13 basis points, and simple non-magnet confirmation placebos performed as well. Independent reconstructions and a machine-learning extension to ETH and XRP reached the same conclusion.