BITCOIN LINEAR REGRESSION

What is Bitcoin Linear Regression?

Bitcoin linear regression fits a straight line through a chosen window of closing prices. The line is selected by minimizing the sum of squared distances between each close and its fitted value. It describes the average direction of that window while allowing individual candles to sit above or below the fitted path.Maketo’s default fits 100 candles and displays the fitted endpoint with two surrounding channel lines. A separate slope view reports the fitted dollar change per candle, also over 100 candles. The forecast view uses 14 candles by default and extends its fitted line one position forward. That extension is arithmetic, not an estimate with demonstrated predictive odds.

Read slope as dollars per candle

Positive slope means the fitted line points upward through the selected window; negative slope means downward. Its unit depends on the candle interval: $100 per hour and $100 per day are different rates. Maketo’s direction label follows the slope sign in all three views, not price’s position relative to the channel.

Understand the channel width

For every close, subtract the fitted value at that candle to obtain a residual, its distance from the fitted line. Maketo squares those residuals, averages them over the full window, and takes the square root. The default channel places two times that residual spread above and below the fitted endpoint.

Read a rolling fit rather than one permanent line

Each new candle produces a new fitted window and a new endpoint. The plotted history connects those successive endpoints, so it can curve even though each individual window was fitted with a straight line. It is not one straight trend line drawn across the entire visible chart.

How Linear Regression is calculated

Choose the line that minimizes squared distances from the last 100 closes. Channel boundaries sit two residual standard deviations from the fit. The forecast variant extends a fitted line by one step. The page is published on Regression channel, deviations 2, period 100. A period counts candles at the interval you choose, so the same setting covers a different stretch of time on a one-hour chart than on a daily one. How smoothing and missing candles are handled.

A calculation you can follow

Five invented closes keep the fit reproducible by hand. This demonstration uses period 5, not the live 100-candle default. Take closes of $100,000, $102,000, $101,000, $104,000 and $103,000. Give them positions 0, 1, 2, 3 and 4. Mean position is 2 and mean close is $102,000.Slope is the sum of (position − 2) × (close − $102,000), divided by the sum of (position − 2) squared. Those totals are $8,000 and 10, giving $800 per candle.The intercept is $102,000 − $800 × 2 = $100,400. The fitted endpoint at position 4 is $103,600; extending the same fit to position 5 gives $104,400.Residuals are −$400, +$800, −$1,000, +$1,200 and −$600. Their squared mean is 720,000 square dollars, giving a residual spread of about $848.53. Two spreads around the $103,600 endpoint give approximately $101,902.94 and $105,297.06.The actual final close is $103,000, below the fitted endpoint, while the fitted slope is positive. Price position and the window’s fitted direction are different facts.

Period 100 for fit and slope

A longer fit includes more historical conditions and can obscure a recent change of direction. A shorter fit responds more to individual candles. The least-squares calculation uses equally spaced candle positions, not the number of trades or how much time price spent at a level.

Channel multiplier 2

This changes only the distance of the channel edges from the fitted endpoint. It does not change the fitted line or slope. Two residual spreads are not automatically a 95% forecast interval: that interpretation would require statistical assumptions and a different predictive calculation.

Forecast period 14

The forecast preset is a shorter fit, evaluated one candle beyond its endpoint. To compare fit, slope and extension from the exact same window, set their periods equally. The page does not establish that the next close will reach the extended value.

An unusual close can rotate the fit

Squared distances give larger departures more influence. A large move near either end of the window can meaningfully change the slope and endpoint. Check the actual candles rather than assuming a neatly fitted line means the path was steady.

Residual spread differs from price spread

A strong, nearly straight rise can have a small residual spread despite covering a large dollar range. The channel measures departures from the fitted path, not distance from one horizontal average price.

A gap is not a hidden extra observation

Missing candles interrupt the fit. Maketo does not fill a missing price with a repeated close or stretch the same five observations across an unknown interval. Enough consecutive candles must exist for the selected period.

Linear Regression compared with Bollinger Bands

Bollinger Bands measure close dispersion around a horizontal moving average. Regression channels measure departures from a sloping fitted path. Their widths can differ greatly during an orderly rise or fall; they answer different questions about the shape of price movement. Open Bitcoin Bollinger Bands.

Why is the displayed regression history curved?

Each point is the endpoint of a newly fitted rolling window. Connecting those changing endpoints creates a path, although the fit inside each window is straight.

Does the forecast come from a trained prediction model?

No. It is the current least-squares line extended by one candle position. It adds no future data, market assumptions, or probability estimate.