An adaptive moving average that self-tunes its responsiveness — fast in clean trends, smooth in chop. Built on Ehlers-style signal-processing principles with phase and vector parameters that tune the lag/responsiveness trade-off per instrument.
Section 01 — Overview
BWT Precision AI-MA is an adaptive moving average — a smoothed price line that automatically adjusts its responsiveness based on real-time market conditions. During clean trends, the AI-MA accelerates and tracks price closely. During chop and consolidation, it smooths aggressively and flattens out. This adaptive behavior is the inverse of a fixed-period EMA or SMA, which is either always too slow during trends or always too noisy during chop.
The adaptive logic is influenced by the work of John Ehlers, who pioneered the application of digital signal processing techniques — Hilbert transforms, dominant cycle detection, MAMA/FAMA — to moving averages. The AI-MA exposes two tunable parameters from this lineage: Phase, which controls the trade-off between lag reduction and smoothness, and Vector, which adjusts how aggressively the MA responds to detected trend strength. Tuning these per instrument is what separates a generic adaptive MA from one calibrated to your market.
The single most important read this indicator delivers is slope state. A clearly-sloping AI-MA tells you the adaptive logic has detected a trend and accelerated to track it — that is your trade signal. A flat AI-MA tells you the adaptive logic has flattened because no trend is present — that is your "stand aside" signal. This is fundamentally different from how a fixed-period EMA behaves: the EMA is always sloping something regardless of whether the slope is meaningful, while the AI-MA flattens out when the trend dissolves. Use the flatness itself as actionable information.
Screenshot
Image provided by you
Section 02 — Concept Reference
A working glossary of every concept underlying the AI-MA — the signal-processing principles, the tunable parameters, and the visual signals you read on the chart. Master these concepts and the indicator becomes a precision tool rather than a generic average.
A moving average whose smoothing constant changes bar by bar based on a measure of trend strength. When the measure is high (clean directional move), smoothing reduces and the MA accelerates. When the measure is low (chop), smoothing increases and the MA flattens. This is the foundational concept for every adaptive MA family — KAMA, MAMA, FRAMA, and the BWT AI-MA all share this principle.
John Ehlers's body of work applying digital signal processing techniques to financial time series. Includes dominant cycle detection (Hilbert transform), zero-lag filters, MAMA/FAMA (the MESA Adaptive Moving Average), and the use of phase and amplitude as tunable parameters. The AI-MA inherits the philosophy: treat price as a signal, smooth adaptively, and expose the trade-offs as parameters rather than hardcoded constants.
Controls the trade-off between lag and smoothness. Lower phase values produce a more responsive MA — closer to price, more signal whip during chop. Higher phase values produce a smoother MA — further from price, less noise but slower to react. The phase parameter is the primary tuning knob: adjust per instrument so the slope state is stable enough to read but responsive enough to catch turns.
Adjusts how aggressively the MA responds to detected trend strength. Higher vector values cause the MA to accelerate more sharply when a trend forms; lower values dampen the response. Use higher vector for fast-moving instruments where you need the MA to catch up quickly during impulsive moves. Use lower vector for slower instruments where smoothness matters more than reaction speed.
The direction and steepness of the AI-MA. A steeply sloping AI-MA confirms an active trend; a gently sloping AI-MA suggests weak or fading directional bias; a flat AI-MA confirms no trend. Trade only when slope is clearly directional. The dynamic color coding makes the slope state instantly readable — the color is the signal.
When the AI-MA flattens, the adaptive logic has detected the absence of trend — the smoothing has increased to its maximum and the MA is essentially a long-period average. This is not a bug or a parameter problem; it is the indicator correctly telling you that no directional setup is available. Flat AI-MA = stand aside. Do not try to read the line; read the absence of slope as a signal.
The transition from a flat AI-MA to a clearly-sloping one is itself a high-quality signal — it marks a regime shift from range to trend. The first pullback to the AI-MA after acceleration begins is typically the cleanest trend entry of the session. Watch for visible steepening bar over bar; that is your "regime change confirmed" cue.
A visual marker placed on the bar where the AI-MA slope changes direction. Because the AI-MA is adaptive, dots only print when the underlying signal-processing logic detects a meaningful directional shift — making AI-MA dots more reliable than dots from a fixed-smoothing MA. Treat each dot as a candidate trigger for a reversal, but always confirm with structure.
In a clearly-sloping AI-MA regime, the line itself functions as a dynamic support or resistance level. Price typically pulls back to test the AI-MA before continuing the trend. Because the AI-MA is adaptive, this dynamic S/R behavior is most pronounced precisely when it matters most — during strong trends — and disappears during chop where the AI-MA flattens and provides no level reference.
Section 03 — Workflow
Every AI-MA-based setup follows this sequence. The adaptive nature of the indicator means the first two steps — tuning and slope reading — carry more weight than they would for a fixed-period MA.
Section 04 — Parameters
The AI-MA exposes only four parameters because the adaptive logic does most of the heavy lifting internally. Tune Phase first; Length and Vector are secondary.
| Parameter | Default | Description |
|---|---|---|
| Length | 14 | Base lookback period that anchors the adaptive calculation. Longer = smoother baseline |
| Phase | 30 | Balance between lag reduction and smoothness. Lower = responsive (more signals), higher = smooth (filtered) |
| Vector | 2 | Response gain — how aggressively the MA accelerates when trend strength is detected |
| Show Dots | On | Marks each turning-point bar with a visual dot for reversal monitoring |
Section 05 — Trade Setups
Each setup is a pre-defined sequence that takes advantage of a specific AI-MA state. The adaptive nature of the indicator means most setups are filtered by the AI-MA itself: if the slope state is wrong, no setup applies.
The simplest, most reliable AI-MA setup. The adaptive line is steeply sloping in one direction — confirming an active trend has been detected — and price has just made a new structural high (in an uptrend) or low (in a downtrend) above/below the AI-MA. Enter on the close of any same-direction candle while slope remains directional. The AI-MA filtering does the work: if the line is steeply sloping, the trend is real; if it is flat, this setup does not apply.
When the AI-MA is clearly sloping, it acts as dynamic support or resistance. Price typically pulls back to test the AI-MA before continuing the trend. The pullback to the AI-MA is the entry zone; rejection from the AI-MA with a same-direction candle close is the trigger. This is the cleanest single-indicator setup available with an adaptive MA — the indicator filters non-trends out automatically, so every valid pullback is in a confirmed trend.
The highest-quality AI-MA setup. After an extended flat-AI-MA period (range), the line transitions to clearly sloping — a regime shift from chop to trend. The first pullback to the AI-MA after this transition is typically the cleanest entry of the session because price has just established a new directional bias and the adaptive logic has confirmed it. Watch specifically for the slope to visibly steepen bar over bar.
When a clearly-sloping AI-MA flattens and then reverses direction, the adaptive logic has detected meaningful momentum loss followed by a directional shift. The turning-point dot prints on the reversal bar. Combine with structural confirmation (price taking out the most recent swing in the new direction) for a high-probability reversal entry. Avoid this setup when the AI-MA is bouncing between sloping states in chop — confirmation is everything.
The most profitable "setup" most days is not trading. When the AI-MA is flat for an extended period, the adaptive logic has confirmed there is no trend to trade. Stand aside. This is not a passive recommendation — it is an active discipline. The single highest-impact use of the AI-MA for many traders is as a filter that prevents low-quality counter-trend trades during midday lulls and lunchtime drift.
Section 06 — Best Practices
Adaptive MAs reward traders who treat slope and flatness as primary signals — not as cosmetic features. The following practices distill the most consistently effective rules from years of adaptive-MA use.
The Phase parameter is the single most important tunable. Lower Phase produces a responsive AI-MA (more signals, more noise); higher Phase produces a smoother AI-MA (fewer signals, more lag). Tune per instrument: faster instruments like NQ benefit from lower Phase to keep the line close to price; slower instruments like ES or daily-timeframe analysis benefit from higher Phase to filter noise. Trade only after tuning — defaults are a starting point, not a destination.
Read the AI-MA in two states: clearly sloping = trade, flat = stand aside. The flatness is not a missing signal; it is an active "no trend detected" message from the adaptive logic. Do not try to read price action against a flat AI-MA — the indicator is telling you the market is in a regime where its own outputs are not actionable. Stand aside until the slope returns.
When the AI-MA transitions from flat to clearly sloping, that is the indicator confirming a regime shift from range to trend. The first pullback to the AI-MA after this transition is statistically the cleanest entry of the session — a fresh trend has just been validated by the adaptive logic. Watch specifically for visible steepening bar over bar; once the slope is clearly directional, the regime shift is confirmed.
The Vector parameter controls how aggressively the AI-MA accelerates when trend strength is detected. Higher Vector = faster catch-up during impulses (good for fast-moving instruments). Lower Vector = smoother during impulses (good for instruments where smoothness matters more than reaction speed). Most traders never adjust Vector after the initial tune — set it once per instrument, leave it.
A turning-point dot tells you the AI-MA slope has changed sign — not that a meaningful reversal is occurring. Confirm every dot with structure: a price break of the most recent swing in the new direction, a candle close past a structural level, or a test of an obvious S/R. Dots in the middle of choppy ranges still print and unprint; dots accompanied by structure are real signals.
In a clearly-sloping AI-MA regime, the line acts as dynamic support (rising) or resistance (falling). Mark it mentally as your "buy zone" in uptrends and "sell zone" in downtrends. This dynamic S/R behavior disappears completely when the AI-MA is flat — do not pre-position limit orders at a flat AI-MA; price will cross through it without any reaction.
Adaptive moving averages reduce lag — they do not eliminate it. The AI-MA confirms trends; it does not predict them. The first bar of a new move is often missed; the dot prints two or three bars after the actual reversal. This is the price you pay for noise filtering. Use the AI-MA for entries on pullbacks and trend continuations, not as a leading indicator at exact tops and bottoms.
A moving average is a derived signal. The strongest setups occur when the AI-MA pullback aligns with a structural level — prior swing high/low, session open, daily pivot, or a level from BWT Core Levels. Confluence between an adaptive signal and a static price level filters out the highest-quality entries. Use the AI-MA as your trend filter, the level as your entry trigger.
A sloping AI-MA is not a green light to take any setup in the slope direction. Slope steepness matters: a steeply-sloping AI-MA is a strong-trend signal; a gently-sloping AI-MA is a weak-trend signal. In weak-trend conditions, demand more confirmation (structural break + candle close + level confluence) before entering. Treat slope as a continuous signal, not an on/off switch.
If your instrument changes character — volatility expansion, contraction, news regime — revisit your Phase and Vector settings. Parameters tuned during a calm summer market may produce too much noise during an active earnings season. Adaptive logic adjusts within the session, but the parameters that govern that logic still need to match the broader regime.
Section 07 — Common Mistakes
These are the recurring failure modes specific to adaptive moving averages. Avoiding them is, on its own, a meaningful edge.
The flatness is not a missing signal — it is the indicator telling you no trend exists. Traders who try to read price action while the AI-MA is flat take low-probability counter-trend trades that the indicator was designed to filter out. Trust the flat state. Stand aside.
Using a high Phase value on a fast-moving instrument like NQ produces an AI-MA that lags every meaningful turn. Using a low Phase on a slower instrument creates noise. The Phase parameter exists precisely so you can match it to instrument character — leaving it on default for every chart wastes the indicator's main feature.
A gently-sloping AI-MA is not the same trade signal as a steeply-sloping one. Treating both equally — entering with the same conviction in both — leads to losses on the weak-trend setups. Slope is a continuous signal; demand more steepness for higher-conviction trades.
"AI-MA up = take longs" is too coarse. The adaptive logic provides nuanced information through slope steepness and the flat-to-sloping transitions. Reducing it to a binary up/down filter throws away the precision the indicator was designed to deliver.
Even an adaptive MA is a smoothing filter — it lags by construction. Traders who expect the AI-MA to print exact tops and bottoms get frustrated when the dot prints two bars late. That late-by-two-bars print is the cost of noise reduction; the trade is to enter on pullbacks afterward, not to anticipate the dot.
A turning-point dot in the middle of obvious chop is noise — the slope flipped because the smoothing relaxed momentarily. Without a structural break, prior-level test, or candle confirmation, the dot is not actionable. Discipline yourself to wait for structure.
Placing stops at or just inside the AI-MA itself guarantees they get hit on routine pullbacks during otherwise valid trends. Stops belong beyond the AI-MA by at least 1 ATR, or beyond the most recent structural swing — never inside the noise the AI-MA is designed to absorb.
Adjusting Phase until the line "looks nice" is not tuning — it is decoration. Tune by behavior: does the slope state stay stable for several bars at a time during trends? Does it correctly flatten during chop? If yes, the tune is good. If not, adjust further.
BWT Precision Indicators require a valid BWT license for NinjaTrader 8. The adaptive moving-average concepts described on this page draw from publicly available work by John Ehlers and the adaptive-MA literature. This page is provided for informational and educational purposes only and is not trading advice. Trading futures and other leveraged products involves substantial risk of loss and is not appropriate for all investors. Past performance is not indicative of future results.