Rolling Pearson r
A coefficient from -1 to +1 calculated on log returns, not raw prices.
Binance stream · rolling log-return correlation & lead-lag
Pick at least one leader and one follower, then press Start live. A rolling buffer needs about 20 ticks (~20 seconds) before correlations appear.
r = Pearson correlation of 1-second log returns (green positive, red negative). The lag is found by shifting follower returns from −maxLag to +maxLag ticks and keeping the shift with the strongest |r|. "Lags" means the follower reacts after the leader; "Leads" means it moves first. A peak below |r| = 0.50 is treated as noise and shown as "no clear lag".
Quick answer
Correlation measures how often two assets move together. TradeFind computes Pearson correlation on one-second log returns for Binance USDT pairs, then slides the follower series across the leader to estimate lead/lag. A high r means the direction is similar; a lag estimate means one series’ past values line up better with the other’s present values.
A coefficient from -1 to +1 calculated on log returns, not raw prices.
The shift with the strongest absolute alignment, converted from ticks using the measured cadence.
Dim non-matching cells so the matrix stays stable while you focus on medium or strong relationships.
Read exact coefficients in the table, or scan magnitude and direction in the heatmap.
Use the matrix as a map of current relationships, not a prediction engine. Check the sample count and the selected window first, then compare the coefficient with the lag estimate. A strong pair with no clear lag is still useful context; a weak pair with a dramatic lag is usually noise. Re-check after major volatility because crypto relationships are regime-dependent.
There is no universal threshold. Above |r| = 0.5 is often treated as strong and above 0.3 as moderate, but sample size, volatility and regime changes matter.
No. It describes the best alignment in the sampled window. It can disappear when liquidity, volatility or the set of leaders changes.
Log returns are additive over time and comparable across assets with very different prices, from BTC to a low-priced altcoin.
One-second tick data cannot resolve sub-second timing reliably, so tiny differences collapse to “Sync” instead of being sold as a timing edge.
Read next: How to read the Correlation Radar · Risk disclaimer
Stores leaders, followers, timeframe, filter, alerts and view settings in this browser (localStorage).