For the Geeks

For the Geeks

This page is optional. You can use the indicator effectively without reading it. But if you want to understand why the oscillator behaves differently from a standard RSI, how your smoothing choices affect the pipeline, and what is actually happening when the blended line moves — this is where those mechanics are laid out.

Nothing here is required for daily use. Everything here helps you make better configuration decisions and interpret the pane with more confidence.

Why the oscillator uses −100 to +100

Standard RSI runs from 0 to 100, with 50 as the neutral midpoint. That scale works fine for a single-instance RSI indicator. But this oscillator needs to stack multiple RSI readings, show regime state relative to a center line, and blend values from different timeframes and tickers into a single composite. For all of those jobs, a symmetric scale centered on zero is a more natural fit.

The conversion is a one-to-one rescaling:

Bipolar = (Standard RSI − 50) × 2

And the inverse:

Standard RSI = (Bipolar / 2) + 50

The values are clamped to −100 and +100 to handle any minor overshoot from smoothing that pushes the output marginally outside the theoretical range.

No information is created or destroyed in this conversion. A bipolar value of +40 is exactly a standard RSI of 70. A bipolar value of 0 is exactly a standard RSI of 50. The transformation is lossless — it changes the visual grammar, not the data.

Why this matters for you: Every threshold and level you have internalized for standard RSI must be mentally translated. Standard RSI 70 (a common overbought level) lives at bipolar +40, not +70. The oscillator's default OB/OS thresholds of ±70 correspond to standard RSI 85 and 15 — genuinely extreme readings that many traders never encounter in normal conditions. If the oscillator seems to "never reach overbought," it is because the default threshold is more extreme than what you are used to, not because the oscillator is misbehaving.

The dual-smoothed regime model

A standard RSI indicator shows you a single line that oscillates between 0 and 100 (or −100 to +100 on the bipolar scale). You interpret it by comparing the value to fixed thresholds — is it above 70? Below 30? Those thresholds define your zones.

This oscillator does something different. Each slot computes two lines:

1. RSI (K): The raw RSI is calculated with ta.rsi(), then smoothed with a moving average of your chosen type and length. This is the slot's primary momentum reading.

1. Signal (D): The smoothed RSI (K) is smoothed again with a second moving average — potentially a different type and length. This is K's own moving average.

Regime is defined by the relationship between K and D:

• K > D → Bullish regime. RSI is above its own recent average. Momentum is increasing relative to where it has been.

• K < D → Bearish regime. RSI is below its own recent average. Momentum is fading.

This is structurally identical to how MACD defines signal crossovers. In MACD, you have a fast EMA minus a slow EMA (the MACD line) and a signal EMA of that line. When the MACD line crosses above the signal, it means short-term price momentum is accelerating relative to its recent average. The same logic applies here, except the underlying measurement is RSI instead of price-based exponential moving averages.

Why this matters for you: The regime model tells you about the direction of change in RSI, not its absolute level. A slot can be in bullish regime at a bipolar value of −20. That means RSI is below the midpoint but recovering — K has crossed above D from below. Conversely, a slot can be in bearish regime at +40. RSI is well above the midpoint but losing momentum — K has crossed below D.

This is a fundamentally different reading from threshold-based RSI. Threshold-based interpretation asks "how high or low is RSI?" Regime-based interpretation asks "is RSI getting stronger or weaker relative to itself?" Both are valid readings. They answer different questions. When you see a slot's line color change on this oscillator, you are seeing a shift in the direction of momentum change — not a crossing of a fixed level.

If you have been trading with standard RSI for years, the habit you need to unlearn is watching where the line sits relative to 70 and 30 (or, on this scale, ±40). That habit is not wrong — level still matters, and the bipolar position tells you something real about absolute momentum. But regime color is giving you a different and often earlier piece of information: whether momentum is accelerating or decelerating relative to its own recent average. A slot at +45 that just turned dim (bearish regime) is telling you that RSI is still well above neutral but has started fading. A slot at −10 that just turned bright (bullish regime) is telling you that RSI is slightly below neutral but climbing. The combination of position and regime is richer than either reading alone, and learning to hold both in mind simultaneously is what separates casual use from informed use of this oscillator.

How blending works

When multiple slots are enabled with non-zero weights, the indicator computes a blended RSI and a blended Signal using a weighted average.

The process:

1. For each enabled slot where the weight is greater than zero and the K value is not na (not unavailable due to warmup or error), accumulate weight × K and weight × D.

2. Sum the total weight across all qualifying slots.

3. Divide the accumulated sums by total weight.

The result is:

Blended RSI = Σ(weight_i × K_i) / Σ(weight_i)Blended Signal = Σ(weight_i × D_i) / Σ(weight_i)

Weights auto-normalize. This means only the ratios between weights matter. Weights of 1/1/1 and 33.3/33.3/33.3 and 100/100/100 all produce identical blends. You can use any scale that makes intuitive sense to you.

If the total weight is zero (no qualifying slots), the blend is na and nothing draws.

What blending reveals and what it hides: The blend gives you a single summary value that reflects the weighted average of your slot readings. When slots agree, the blend is a useful shorthand for "the stack reads bullish (or bearish) at about this level." When slots disagree, the blend produces a middle value that can look more decisive than the evidence warrants. A blend of +30 could mean three slots all near +30, or it could mean one slot at +70 and two at +10. The blend cannot distinguish these situations. The individual slot lines can.

Cross-ticker blending: RSI is already percentage-based — it measures proportional up/down price movement, not absolute price levels. This means RSI values from different instruments can be meaningfully averaged on the bipolar scale. A +40 reading from an index and a +20 reading from an individual stock both describe the same thing — proportional momentum intensity — even though the underlying instruments are different. The blend of those readings describes the average momentum state across the instruments you chose, weighted by your configuration. It does not describe the momentum of either instrument individually.

The lag chain

Lag is the delay between a price change and the oscillator's response to that change. In this indicator, lag accumulates through multiple stages:

Stage 1: RSI calculation

ta.rsi() uses a lookback of RSI Length bars. A sudden price move only begins to dominate the RSI calculation once it represents a significant portion of the lookback window. With RSI Length 14, the RSI needs several bars of consistent directional movement before it responds decisively.

Stage 2: RSI Smoothing (K)

The raw RSI is smoothed with a moving average of RSI Smoothing length. If the RSI Smoothing MA is an SMA of length 3, the smoothed K is the average of the last three RSI values. This adds delay, and the exact feel of that delay depends on the MA type you chose.

Stage 3: Signal (D)

K is smoothed again with the Signal MA of Signal Length. That adds another layer of delay before regime crossovers appear.

Stage 4: Blending

Blending is a weighted average across slots and adds no temporal lag. It operates on the current-bar values of K and D from each slot. The blend does not look back — it simply averages what the slots report on this bar.

Stage 5: Master Smoothing (optional)

If Master Smoothing is enabled, the blended RSI and Signal are smoothed one more time with the Master MA of Master Length. Another few bars of delay added to the composite.

How to think about total lag

The total lag from a price event to a regime response is not something the source can reduce to one universal formula. Each stage can add delay, but the exact result depends on your MA types, lengths, timeframe, On Bar Close choice, and how abruptly price moved.

With defaults (RSI Length 14, RSI Smoothing 3, Signal Length 3, no Master Smoothing):

• Raw RSI needs enough history before it settles into a usable reading.

• The smoothed K line lags raw RSI because it is another MA pass.

• The Signal D line lags K because it is yet another MA pass.

• A regime flip happens only when K actually crosses D, so crossover timing can arrive later than either line's first turn.

What matters in practice is direction, not fake precision: heavier smoothing and Master Smoothing make the chart calmer and later. If you want to know how much delay your exact setup introduces, measure it on-chart. Note where price makes the move, then note where K/D or the blended regime actually flips for your chosen configuration.

The practical takeaway: Every smoothing knob you turn adds delay. The chart gets smoother and the regime flips get cleaner, but they arrive further from the price events that caused them. There is no free smoothing — every bar of reduced noise costs you a bar of response time. Your job is to find the balance where regime flips are clean enough to be useful and timely enough to be actionable for your specific decision speed.

You can check your effective lag by watching when regime flips appear relative to the price moves that caused them. If the price made a clear directional move and the regime flip shows up six bars later, you know your pipeline introduces about six bars of delay. Whether that is acceptable depends entirely on how quickly you need to respond.