Dynamic (auto increment) input length (count) on a linear regression channel drawing, starting from given bar_index/datetime in Pine Script language - plot

Question

My question is about the Linear Regression drawing.
The example in the documentation uses a fixed length (100), and is therefore :
shifting to the right on each new bar
of constant width (here 100 bars)
I'm trying to make it start from a custom point in time (x bars from now or bar_index or datetime...), so that :
it keeps extending on each new bar
but the starting point remains on the same location (until we change it in the settings).
That means that the length (input) would be dynamic and increase on each new bar.
I am getting the following error : Pine cannot determine the referencing length of a series. Try using max_bars_back in the study or strategy function.
Is it possible to do ?
Here is the code
//#version=4
study("Linear Regression", shorttitle="LinReg", overlay=true)
upperMult = input(title="Upper Deviation", defval=2)
lowerMult = input(title="Lower Deviation", defval=-2)
useUpperDev = input(title="Use Upper Deviation", defval=true)
useLowerDev = input(title="Use Lower Deviation", defval=true)
showPearson = input(title="Show Pearson's R", defval=true)
extendLines = input(title="Extend Lines", defval=false)
// ====================================================================
// ====================================================================
// Original parameter (the one that should increments)
// len = input(title="Count", defval=100)
// Unsuccessful attempt : "Starting from given bar_index"
barIndexOfStartingBar = 6392 - 80 // 6392 : Current bar_index, 80 : Offset to the starting bar
len = bar_index - barIndexOfStartingBar
len := nz(len[1]) + 1
// Unsuccessful attempt : "x bars from current bar"
startingPointFromCurrentBar = input(title="Count", defval=80)
len = (bar_index + startingPointFromCurrentBar) - bar_index
len := nz(len[1]) + 1
// ====================================================================
// ====================================================================
src = input(title="Source", defval=close)
extend = extendLines ? extend.right : extend.none
calcSlope(src, len) =>
if not barstate.islast or len <= 1
[float(na), float(na), float(na)]
else
sumX = 0.0
sumY = 0.0
sumXSqr = 0.0
sumXY = 0.0
for i = 0 to len - 1
val = src[i]
per = i + 1.0
sumX := sumX + per
sumY := sumY + val
sumXSqr := sumXSqr + per * per
sumXY := sumXY + val * per
slope = (len * sumXY - sumX * sumY) / (len * sumXSqr - sumX * sumX)
average = sumY / len
intercept = average - slope * sumX / len + slope
[slope, average, intercept]
[s, a, i] = calcSlope(src, len)
startPrice = i + s * (len - 1)
endPrice = i
var line baseLine = na
if na(baseLine) and not na(startPrice)
baseLine := line.new(bar_index - len + 1, startPrice, bar_index, endPrice, width=1, extend=extend, color=color.red)
else
line.set_xy1(baseLine, bar_index - len + 1, startPrice)
line.set_xy2(baseLine, bar_index, endPrice)
na
calcDev(src, len, slope, average, intercept) =>
upDev = 0.0
dnDev = 0.0
stdDevAcc = 0.0
dsxx = 0.0
dsyy = 0.0
dsxy = 0.0
periods = len - 1
daY = intercept + (slope * periods) / 2
val = intercept
for i = 0 to periods
price = high[i] - val
if (price > upDev)
upDev := price
price := val - low[i]
if (price > dnDev)
dnDev := price
price := src[i]
dxt = price - average
dyt = val - daY
price := price - val
stdDevAcc := stdDevAcc + price * price
dsxx := dsxx + dxt * dxt
dsyy := dsyy + dyt * dyt
dsxy := dsxy + dxt * dyt
val := val + slope
stdDev = sqrt(stdDevAcc / (periods == 0 ? 1 : periods))
pearsonR = dsxx == 0 or dsyy == 0 ? 0 : dsxy / sqrt(dsxx * dsyy)
[stdDev, pearsonR, upDev, dnDev]
[stdDev, pearsonR, upDev, dnDev] = calcDev(src, len, s, a, i)
upperStartPrice = startPrice + (useUpperDev ? upperMult * stdDev : upDev)
upperEndPrice = endPrice + (useUpperDev ? upperMult * stdDev : upDev)
var line upper = na
lowerStartPrice = startPrice + (useLowerDev ? lowerMult * stdDev : -dnDev)
lowerEndPrice = endPrice + (useLowerDev ? lowerMult * stdDev : -dnDev)
var line lower = na
if na(upper) and not na(upperStartPrice)
upper := line.new(bar_index - len + 1, upperStartPrice, bar_index, upperEndPrice, width=1, extend=extend, color=#0000ff)
else
line.set_xy1(upper, bar_index - len + 1, upperStartPrice)
line.set_xy2(upper, bar_index, upperEndPrice)
na
if na(lower) and not na(lowerStartPrice)
lower := line.new(bar_index - len + 1, lowerStartPrice, bar_index, lowerEndPrice, width=1, extend=extend, color=#0000ff)
else
line.set_xy1(lower, bar_index - len + 1, lowerStartPrice)
line.set_xy2(lower, bar_index, lowerEndPrice)
na
// Pearson's R
var label r = na
transparent = color.new(color.white, 100)
label.delete(r[1])
if showPearson and not na(pearsonR)
r := label.new(bar_index - len + 1, lowerStartPrice, tostring(pearsonR, "#.################"), color=transparent, textcolor=#0000ff, size=size.normal, style=label.style_labelup)

Answer 1

With these changes you should be able to go back a few thousand bars:
study("Linear Regression", shorttitle="LinReg", overlay=true, max_bars_back = 4999)
and either one of these:
// #1
offsetToStart = input(100, minval = 1, step = 100)
len = max(1, offsetToStart)
// #2
startingBar = input(10000, minval = 1, step = 100)
len = max(1, bar_index - startingBar)

Answer 2

What you are looking for is called anchoring. You can do this with a trick. What I usually do is the following:
p = input("D", title="Period", type=input.resolution)
offsetToStart = input(0, title="Begin Offset") // remove the minval
newbar(p) => change(time(p)) == 0?0:1
nb = newbar(p)
bars = barssince(nb) + offsetToStart
var line l1 = na
l1 := line.new(bar_index[bars], open, bar_index[bars], open + syminfo.mintick, extend=extend.both)
line.delete(l1[1])
Remember to remove the minval. Chose the period and you can now anchor beginning of your indicator. Playaround with p