Limitations and Trust Boundaries

Limitations & Trust Boundaries

Every tool has a region where it helps and a region where it starts to mislead. This page maps those regions for Axiom MA Osc Lite — not as fine print, but as real instruction you can use to judge when to trust the oscillator, when to doubt it, and when to ignore it entirely.

If the oscillator just showed you something that felt authoritative and you are about to act on it, read the section below that matches what you saw. If you are studying the tool before you use it, read the whole page. Either way, the goal is the same: you should be able to name what the oscillator is good at, where it stops being good, and why — before the stakes are real.

Boundary 1: Saturation and resolution loss at the extremes

What happens

The tanh bounding function that converts raw distance into the −100 to +100 scale has a mathematical property that shapes everything about how the oscillator behaves at the extremes: it flattens. As the input gets larger, the output approaches ±100 asymptotically — meaning it gets closer and closer but never quite reaches the bound, and the rate of change drops off sharply.

In practical terms: the difference between a reading of +50 and +60 might represent a change of 0.5 ATR in actual distance from the MA. The difference between +95 and +98 might represent a change of 3 ATR or more — a vastly larger actual change compressed into three points of oscillator space.

Why it matters

Near the center of the range, the oscillator is informative. Small moves in the reading correspond to small moves in the actual distance, and you can meaningfully distinguish between readings. Near the extremes, the oscillator is saturated. Large moves in the actual distance produce tiny moves in the reading, and you cannot tell from the oscillator alone whether the distance is "big" or "enormous."

This means the oscillator loses resolution exactly where many traders want it most — when price is far from the average and the question "is momentum still accelerating?" matters for the trade. The oscillator at +95 cannot answer that question. It can sit at +95 while the actual distance doubles.

How to think about it

When the oscillator is in the saturated zone (roughly above ±80, depending on sensitivity), treat the reading as a binary statement: "price is far from the average." Do not try to extract gradations of strength from small differences in the reading. The information density in the saturated zone is low by construction — not because the oscillator is broken, but because tanh compresses that range.

If you need to understand whether momentum is still accelerating in the saturated zone, the oscillator cannot help you directly. Look at the price chart, look at the rate of change of the underlying distance (if you can approximate it), or use a different tool that preserves resolution at extremes. The oscillator's job in this zone is to tell you "price is stretched" — not "how stretched."

<!-- TEACHING ASSET: Trust-boundary visual — annotated screenshot showing the oscillator pinned near +100 for many bars while price continues to move higher, with annotations marking entry into saturated zone, the extended stay, and the point where the oscillator finally begins to drop. A second panel or annotation showing a sharp oscillator drop during ATR expansion, not price reversal. -->

Boundary 2: The blend hides disagreement

What happens

The blended Fast line is a weighted average of all enabled slots' Fast values. When slots disagree — some positive, some negative — the average can land near zero or at a moderate positive/negative value that looks calm and unremarkable.

Why it matters

A blended reading of +10 can mean "all three slots read approximately +10" or "one slot reads +70 and two read −20, and the weights averaged them to +10." These are completely different situations: the first is genuine consensus near moderate stretch; the second is active conflict between timeframes that happens to produce a tidy-looking composite number.

If you act on the blended reading alone — "it is near zero, the market is neutral" — you may be ignoring the fact that your short-term and long-term timeframes are pointing in opposite directions. The blend has averaged the disagreement into a number that suggests agreement.

How to think about it

The blend is a summary, not a verdict. Use it for a quick read of the pane when you have calibrated your weights and understand what the blend represents at your settings. But whenever the blend is near zero and you are making a decision, or whenever the blend does something unexpected, check the individual slot lines. The slots contain the real information; the blend is a compression of it.

Think of it this way: if the blend is strongly positive and all visible slot lines are also positive, the blend is a faithful summary. If the blend is near zero or moderate, it might be a faithful summary of neutral conditions — or it might be masking a tug-of-war. You cannot distinguish these from the blend alone.

This problem intensifies when slots run on different tickers. Cross-ticker disagreement can represent a macro dynamic — equities dropping while treasuries rally, for instance — that the blend averages into a calm-looking number. See Multi-Ticker Mixing for a full treatment of when cross-ticker blending helps and where it masks the picture.

Boundary 3: Alignment is not a signal

What happens

When all enabled slots are in the same regime (all bullish or all bearish), the alignment alerts fire and the visual impression is of complete agreement — all lines colored the same way, all regime states pointing the same direction.

Why it matters

Alignment is the condition most likely to be treated as "confirmation" that the trade is safe. It feels like it means something structural: every timeframe agrees, so the direction must be strong.

But alignment is a backward-looking observation. It tells you that all configured MAs at all configured timeframes have had price on the same side for long enough that the Fast line is above (or below) the Slow line on each one. This is consistent with a strong trend in progress — and it is also consistent with the late stage of a trend where the slowest slot has finally caught up just before the reversal begins.

Alignment can persist through an entire trending leg without adding information about timing. It can hold for fifty bars, providing no edge whatsoever about whether the next bar continues or reverses. And it can appear at the worst possible moment — right when the last holdout slot flips to join the others, which sometimes happens exactly as the move is exhausting itself.

Here is the scenario that costs people money: the 5m and 15m slots have been bullish for an hour. The 60m slot finally flips bullish. The alignment alert fires. The trader reads this as "all timeframes agree — this is confirmed." They enter the trade. What actually happened is that the move ran long enough for the slowest average to catch up. The 5m and 15m have been in the move for a while and may be nearing their stretch limits. The 60m just got there. Alignment arrived not because something new started, but because something old finally showed up in the lagging data.

How to think about it

Treat alignment as a description of the current state, not as a prediction. "All my configured timeframes show price above the average and the distance is increasing" — that is what alignment says. What you do with that information depends on the rest of your process.

The question to ask when alignment fires is not "should I enter?" but "when did each slot arrive at this regime?" If the 5m slot has been bullish for hours but the 60m slot just flipped, the alignment is fresh in a structural sense — the longest timeframe just joined. But "fresh" does not mean "early." The fastest slot has already been in this move for a long time. If all three have been aligned for days, the alignment is stale and tells you nothing new. Neither freshness nor staleness gives you a timing edge. Both tell you where you are, not where you are going.

Boundary 4: Sensitivity is not calibrated

What happens

The sensitivity parameter multiplies the ATR-normalized distance before the tanh function is applied. Different sensitivity values produce different mappings from actual distance to the bounded display. A reading of +70 at sensitivity 0.5 represents a larger actual distance than +70 at sensitivity 2.0.

Why it matters

The default sensitivity of 1.0 is a starting point, not an optimized setting. It was not derived from statistical analysis of any specific instrument or timeframe. On volatile instruments where price regularly sits 2-3 ATR away from the MA, the oscillator at sensitivity 1.0 may spend most of its time near ±100 — effectively saturated. On quiet instruments with low ATR-relative distance, the oscillator may never leave the ±40 zone.

If you compare readings across instruments without adjusting sensitivity for each, or if you interpret the oscillator's position on the ±100 scale as having absolute meaning ("+70 is always pretty stretched"), you are treating the sensitivity setting as if it were calibrated. It is not.

How to think about it

The sensitivity parameter is a lens that determines how the oscillator presents the underlying distance data. Different lenses produce different pictures of the same reality. Before you interpret the oscillator's position on the scale — before you decide that +70 is "stretched" or +30 is "mild" — you need to know how the sensitivity setting maps those numbers to actual distance on your instrument at your current volatility.

The practical step: after setting sensitivity, look at a few weeks of data and ask whether the oscillator uses a reasonable portion of the range. If it is always saturated, your sensitivity is too high for this instrument. If it never leaves the center, your sensitivity is too low. Adjust until the oscillator's range usage matches your analytical needs. See For the Geeks — Sensitivity experiment.

Boundary 5: OB/OS levels are not RSI thresholds

What happens

The overbought and oversold reference lines (default +70 and −70) are user-defined thresholds on the oscillator's tanh-bounded scale. They trigger specific alert conditions when crossed.

Why it matters

Traders with RSI experience import an intuition: "overbought readings mean the market is stretched and likely to pull back." RSI's 70/30 levels carry some empirical weight in range-bound conditions — decades of use have built a body of experience around what those zones tend to mean.

The tanh-bounded oscillator's thresholds do not carry the same weight. The construction is different: tanh can hold extreme values indefinitely during trending conditions without any tendency to revert. RSI has a built-in mean-reversion property in its calculation (it normalizes by average gains plus average losses); this oscillator does not.

A reading above +70 on this oscillator can persist for days or weeks during a strong trend. Crossing the OB level does not make a reversal more likely. It means the distance from the MA, at your sensitivity setting, is large enough to cross a line you drew. That is all.

How to think about it

Use the OB/OS levels as attention thresholds: "when the oscillator reaches this zone, I want to pay attention and check other factors." Do not use them as action triggers: "when the oscillator reaches this zone, I should take a countertrend trade."

If you set OB at +70 and the oscillator crosses it, the correct response is to notice and investigate — not to sell. Check the individual slots, check the rate of change, check the price chart, check your other analysis. The oscillator reaching your threshold is the beginning of an evaluation, not the conclusion of one.

Boundary 6: Regime is direction, not quality

What happens

Regime color tells you whether Fast is above or below Slow for each slot. Bullish regime (vivid color) means Fast > Slow. Bearish regime (faded color) means Fast < Slow.

Why it matters

Regime can feel like a quality assessment: "bullish is good, bearish is bad." But regime is purely about the direction of the distance score relative to its own smoothed average. A slot can flip to bullish regime because the oscillator went from −80 to −60 (still deeply negative, but now above its Slow line). A slot can flip to bearish regime from +90 because the oscillator pulled back to +85 while Slow is still at +88.

If you read regime as "bullish means the oscillator supports buying" and "bearish means sell," you will misread situations where the oscillator is bearish at +80 (still strongly positive, just pulling back from an extreme) or bullish at −60 (still deeply negative, just bouncing within a downtrend).

A scenario that trips people up

The 60m slot has been at +85 for a while. Price is far above the session-level average. Then the Fast line drops to +78 while the Slow line is at +82. The color fades — bearish regime. The trader sees the faded color and thinks "bearish." But the oscillator is still at +78. Price is still well above the average. All that changed is that the stretch is starting to compress slightly. The slot flipped to bearish regime while deeply in positive territory. Bearish regime at +78 means "the distance is narrowing" — it does not mean "the market is bearish."

The reverse is equally common. A slot climbs from −80 to −55 and flips to bullish regime. The color brightens. But the oscillator is still at −55 — price is well below the average. The brightening means the distance is shrinking, not that the picture has turned positive.

How to think about it

Regime tells you the direction of the most recent change in the distance score, relative to that slot's own smoothed trend. It is useful shorthand for "is the distance currently increasing or decreasing?" But increasing from −80 to −60 and increasing from +20 to +40 are both "bullish regime" and they describe very different market contexts. Always read the regime alongside the absolute level of the oscillator. A regime flip at +80 and a regime flip at −80 both produce the same color change — but they do not mean the same thing, and treating them the same will mislead you.

When the oscillator is not the right tool

The oscillator measures one thing: how far price sits from a moving average, normalized by volatility, across multiple timeframes. There are situations where that measurement is not helpful, and using the oscillator in those conditions produces noise rather than insight.

Range-bound, choppy markets

When price stays near the averages on all configured timeframes, the oscillator hovers near zero with frequent small regime flips. It is working correctly — the distance is small and fluctuating — but the readings are not informative. The oscillator was built for conditions where price moves away from the average at different scales and the user wants to gauge that stretch. When there is no stretch to gauge, the tool has nothing useful to say.

Very short time horizons on very high timeframes

If your chart is on the 1-minute timeframe and your longest slot is on the weekly, the weekly slot's Fast line will barely move during an entire trading session. The reading is technically correct but practically useless — the weekly MA is so far from the 1-minute action that the normalized distance changes are invisible at the intraday scale. Match your slot timeframes to the time horizon of your decisions.

When you need absolute price levels

The oscillator is deliberately unitless. It tells you about distance in volatility-relative terms, not in price terms. If you need to know "price is $3.50 above the 20-period EMA," the oscillator cannot give you that. It can tell you that $3.50 is "moderately above" or "extremely above" relative to ATR, but the actual price distance is abstracted away by the normalization. For absolute levels, you need the overlay variant (Axiom MA Lite) or a raw MA on the chart.

When you need reversal timing

The oscillator does not predict reversals. It does not even suggest them. A reading at +95 persisting for forty bars is the oscillator doing its job — accurately reporting that price remains far from the average. The oscillator has no mechanism for detecting when that distance will shrink. If your process depends on an oscillator that tends to revert (like RSI in certain conditions), this tool will frustrate you. It was built to measure, not to forecast.

The honest summary

This oscillator is good at one thing: compressing multi-timeframe MA-distance into one readable pane on a consistent scale. It handles cross-ticker comparison through ATR normalization. It makes the repaint tradeoff explicit and verifiable. Within those boundaries, it does its job well.

It is not good at distinguishing "stretched" from "very stretched" once it enters the saturated zone. It cannot tell you when a trend will end. It cannot tell you when a reversal will start. It cannot resolve disagreement between slots — it can only average it into a number that hides the argument. It cannot tell you whether your sensitivity setting is appropriate for your instrument; that calibration is yours. And it cannot tell you whether the distance you are seeing is historically unusual, because the ATR normalization recalibrates with every new bar of volatility data.

None of these limits are defects. They are the consequences of the construction choices that give the oscillator its strengths. Bounded range means compression at extremes. ATR normalization means the ruler changes length as volatility shifts. Weighted averaging means disagreement gets smoothed over. Every strength has a cost. The oscillator is not going to warn you when you have crossed from the region where it helps into the region where it misleads. That boundary is yours to know.