Limitations and Trust Boundaries

Limitations and Trust Boundaries

Every tool has a territory where it tells you something real and a territory where it starts costing you more than it gives. This page maps both for Axiom RSI Osc Pro. It is not a disclaimer list — it is a working guide to where your trust is warranted, where it needs qualification, and how to tell the difference while the chart is open.

Where the tool earns trust

Per-slot regime detection is mechanically sound. Each slot computes RSI from a clean pipeline — ta.rsi() with your chosen length, smoothed by your chosen MA, with a Signal line smoothed again. The regime model (K > D = bullish, K < D = bearish) is deterministic and verifiable. With On Bar Close enabled, historical values match what was visible in real time. You can verify this yourself using the procedure in MTF and Repainting.

The bipolar scale is a lossless transformation. No information is created or destroyed by the rescaling. A standard RSI of 63 is always a bipolar value of +26. You can convert back and forth with perfect accuracy. The scale changes the visual grammar but not the underlying data.

Alerts are bar-close gated without exception. Even if a slot has On Bar Close disabled and repaints intrabar, alerts wait for bar confirmation. You will not get alert notifications from intrabar noise.

Weight normalization is straightforward and verifiable. Weights auto-normalize, meaning the relative proportions are all that matter. 1/1/1 and 33/33/33 produce identical blends. You can verify this by changing the absolute values and confirming the blend does not move.

The blended line trust gap

This is the single most important limitation to internalize.

The blended RSI/Signal line is a weighted average. When the underlying slot readings agree, the blend accurately summarizes that agreement. When the slot readings disagree, the blend produces a middle value that can look decisive even though the evidence is divided.

The chart looks good, but here is what you might be missing — Scenario A

Three slots are enabled with equal weight. Slots 1 and 2 are at +45 and +40 (bullish regime). Slot 3 is at −20 (bearish regime). The blend reads approximately +22, solidly positive, and the blended line shows green with a bullish fill.

The blend looks bullish. But one of three slots is bearish. If slot 3 is your highest timeframe, that bearish reading may be the more important piece of information — and the blend has averaged it away. The composite reads as moderate bullish confidence when the actual evidence is two bullish lower-timeframe readings and a bearish higher-timeframe reading.

How to catch it: Watch individual slot line colors. If slot 3's line is dim while slots 1 and 2 are bright, the disagreement is visible in the pane. The "All RSI Slots Bullish" alert would not fire in this situation, which is a mechanical confirmation that the stack is not unanimous. Do not rely on the blend alone to tell you the full story.

The chart looks good, but here is what you might be missing — Scenario B

You have five slots enabled. Four have a weight of 10, and one has a weight of 60. The high-weight slot is at +50 (bullish). Two other slots are bearish at −30 and −15. The remaining two are near zero.

The blend reads approximately +22 — it looks bullish. But the reading is almost entirely driven by a single dominant slot. The other four slots range from mildly bearish to neutral. The blend has the visual authority of a multi-voice summary, but it is functionally a lightly diluted reading of one slot.

How to catch it: Check your weight distribution. If one slot carries most of the weight, the blend is not a summary of your stack — it is a smoothed version of that one slot with some noise from the others. If that is intentional, fine. If not, rebalance.

The bipolar threshold mismap

This trips up almost everyone who comes from standard RSI. It deserves its own section because it affects every session where you reference OB/OS levels.

Standard RSI 70 = bipolar +40. Not +70.

If you have internalized "70 means overbought" from years of using standard RSI, that intuition will mislead you on this oscillator. The default OB threshold of +70 corresponds to standard RSI 85 — a reading so extreme that it occurs rarely, even in strong trends. A trader waiting for the oscillator to reach +70 before considering overbought conditions may wait through long periods of strong momentum without ever seeing the threshold hit.

The opposite error is also common. A trader who lowers the OB threshold to +40 (the correct bipolar equivalent of standard RSI 70) may then get frequent alerts, because standard RSI 70 is a moderate extension, not an extreme one. The oscillator's default of ±70 was chosen to flag genuinely extreme conditions. Lowering it is valid, but know what you are choosing: more frequent signals at less extreme levels.

How to handle it: Decide what standard RSI level you consider overbought for your methodology. Convert it: bipolar = (standard RSI − 50) × 2. Set the oscillator's OB and OS levels accordingly. Write it down if you need to. The conversion is simple but the habit of mapping old intuitions to new thresholds takes deliberate practice.

Compounding lag

Lag accumulates through the processing pipeline, but not in a clean bar count the code can promise in advance.

The pipeline per slot is: raw RSI (RSI Length bars of lookback) → RSI Smoothing (more bars) → Signal (more bars). Then across slots: blending (instantaneous, no additional lag). Then optionally: Master Smoothing (more bars again).

With default settings (RSI Length 14, RSI Smoothing 3, Signal Length 3, no Master Smoothing), the reliable statement is simpler: the Signal line and any regime flips will trail raw price more than the raw RSI does, and the effect feels different on every chart timeframe. The exact delay depends on your MA types, lengths, timeframe, On Bar Close choice, and how abruptly price moved.

If you increase RSI Smoothing to 7 and Signal Length to 7, and then layer on Master Smoothing, the pane will usually look calmer and later. That tradeoff is real. The exact number of bars is something you need to measure on-chart with your chosen setup, not something the source can justify as a fixed formula.

How to diagnose it: Watch where regime flips land relative to the price moves that caused them. If you look at a regime flip and the price move it represents is already several bars behind you — if you find yourself thinking "I already knew that" — your smoothing chain is too long for your decision speed. The oscillator is not wrong. It is just telling you things after they have stopped being actionable at your pace.

Another diagnostic: if your regime flips almost never revert on the next bar, your smoothing is quite heavy. Some stability is good — a regime flip that holds is more useful than one that chatters. But if flips only appear well after the move is obvious in price, you have passed the point of useful stability and entered the territory of consistent delay.

How to handle it: Shorten RSI Smoothing or Signal Length first, because those affect individual slots and the blend simultaneously. Reserve Master Smoothing for situations where you specifically need a calmer composite and you are willing to pay for it with additional delay. If the chart still feels too noisy after reducing smoothing, the problem may not be smoothing — it may be that you are watching a timeframe that is faster than your actual decision speed.

Cross-ticker blend interpretation

When you blend RSI from different tickers, you are averaging momentum readings from different instruments. This is technically meaningful — RSI is already percentage-based, so a +40 reading from one instrument and a +20 reading from another can be averaged — but the result does not describe the momentum of any single instrument.

What you get instead is a custom composite: "the average RSI momentum across the instruments I chose, weighted the way I configured." This can be a useful summary for comparing correlated instruments or tracking a reference. But it does not have the same interpretive clarity as a single-ticker, multi-timeframe blend. If the cross-ticker blend flips regime, you cannot immediately tell whether your primary ticker caused it, the reference ticker caused it, or both contributed.

How to handle it: If you use cross-ticker slots, keep their weights deliberate and consider using the individual slot lines (or per-slot alerts) to track which instrument is driving the blend when the composite changes.

Weight-zero slots are not off

Setting a slot's Blended Weight to 0 removes it from the blended calculation. It does not disable the slot. The slot still computes its RSI, still plots its individual line (unless hidden), and still fires all three of its alert conditions (Is Bullish, Is Bearish, Regime Flip).

This catches traders who set a slot's weight to zero expecting it to go silent. If you are getting unexpected alerts from a slot you thought you "turned off," check whether it is still enabled with a zero weight. To fully silence a slot, use the Enable toggle, not the weight.

Weight-zero is a feature, not a gap. It means "I want to monitor this slot separately, but I do not want it influencing the blend." The distinction is useful once you know it exists. The problem is discovering it by surprise.

Session boundaries and gap behavior

The indicator does not filter by session. RSI is computed on whatever bars the chart provides — including pre-market, after-hours, and overnight sessions if your chart displays them. Session boundaries can produce RSI spikes if the price source has a large gap (e.g., a gap-up at market open). The RSI calculation treats the gap bar like any other price change, which can push the oscillator sharply in one direction for a few bars before the lookback window absorbs the gap.

If your chart includes extended-hours data and you notice sharp RSI spikes at session boundaries, that is the indicator correctly processing the data it receives. It is not an error, but it may not reflect the kind of momentum you are interested in. Some traders prefer to trade only during regular hours and use a chart that excludes extended sessions.

What the oscillator structurally cannot do

It cannot tell you why momentum is moving. RSI measures the proportional size of up-moves versus down-moves over a lookback period. It does not know about news events, earnings, volume surges, order flow, or structural levels. When the oscillator shows strong bullish momentum, you are looking at a statistical summary of recent price direction — not a diagnosis of what caused it or a prediction of whether it will continue. A sharp RSI spike from a single gap bar and a steady RSI climb from eight consecutive bullish closes can produce similar readings. The oscillator shows you the resulting momentum state. It does not distinguish between the different kinds of price behavior that created it.

It cannot tell you where momentum will go. Regime state describes the present relationship between RSI and its signal. Even the alignment alerts — all slots bullish — describe the current state of the evidence you configured. The oscillator has no forward-looking mechanism. An all-bullish stack that has held for twenty bars says that momentum has been directionally consistent. It does not say the consistency will continue for a twenty-first bar. Regime persistence is common but not guaranteed, and the longer a regime holds, the more tempting it becomes to trust the pattern — which is exactly the moment to remember that the oscillator is reporting, not predicting.

It cannot replace verification against price action. A bullish RSI stack on a chart that shows price making lower highs and struggling at resistance is a conflict between the indicator and the price structure. The indicator does not know about the resistance. You do. RSI is one input. The chart is another. When they disagree, the oscillator is not wrong — it is just incomplete. The regime reading is accurate within its own scope: RSI really is above its signal on those timeframes. But the scope of what RSI measures does not include whether there is a ceiling above price. Your job is to hold both pictures and decide which one carries more weight in that specific context. The oscillator cannot make that call for you, and it should not look like it can.