Docs  /  AI Studio Targets  /  ATR-Normalized Return

AI Studio Answer Guide: ATR-Normalized Return

Status: Per-family deep reference (companion to the consolidated Fusion AI Studio & ONNX User Guide) Audience: first-time AI Studio user Answer family: ATR-Normalized Return Prediction style: Regression

1. The Plain-English Question

ATR-Normalized Return asks:

“How far up or down is price expected to move over the next N bars, measured in ATR units instead of raw percent return?”

This is one of the most useful AI Studio answer families because it keeps the output in a market-relative unit instead of a raw percent move.

Plain-English translation:

2. What Fusion Is Actually Building

Example:

Meaning:

Target_ATRReturn10[t] = (Close[t+10] - Close[t]) / ATR14[t]

Plain-English translation:

That future ATR-sized move becomes the answer column for training.

3. What The Model Output Means

If the trained model outputs:

Very important

ATR-Normalized Return is:

That means:

On a strong trend day, values beyond +1 or -1 can happen.

4. Why This Family Is Often Better Than Raw Future Return

This is the main reason AI Studio tends to promote it.

Raw Future Return can change scale a lot across:

ATR-Normalized Return keeps the output in a volatility-adjusted unit.

That means:

This is why the Swing Oscillator trading goal in AI Studio prefers it.

5. What Look-ahead Windows Means For ATR-Normalized Return

This family uses the normal Step 1:

If Step 1 says:

1,5,10,20

Fusion can create:

Then in Step 2, you choose one of those horizons for the actual model.

So the workflow is:

Step 1: create several ATR-normalized answer columns
Step 2: choose one horizon for this model

6. Does The ATR Section In Step 1 Affect This Family?

This is one of the easiest places for a user to get confused.

Short answer

No.

The Step 1 ATR section:

is for the ATR Stop / Target Race answer family.

It is not the control section for ATR-Normalized Return.

That means:

Current product truth

The current formula for this family uses:

That ATR period is part of the current built-in target design, not a separate user-tuned field in Step 1 for this family.

7. Does The Flat-Zone Threshold Affect ATR-Normalized Return?

Short answer

No, not directly.

The Step 1 flat-zone / neutral-band field is for direction-style label creation, mainly:

It does not create a dead-zone for ATR-normalized regression outputs.

If you later use the model output in user-defined conditions with:

that is your workflow rule, not the Step 1 neutral-band setting.

8. What 0 Means Here

For ATR-Normalized Return, 0 really matters.

It means:

This family is one of the cleanest “around zero” outputs in AI Studio.

That is why it fits the idea of a bias oscillator better than:

where 0 means something else or is not a true centerline.

9. Good First Output Interpretation

The simplest first condition pattern is:

positive-side condition when model output > 0
negative-side condition when model output < 0

But a more practical workflow condition is often:

positive-side condition when model output > tested positive threshold
negative-side condition when model output < tested negative threshold
neutral condition near 0

Examples:

The exact thresholds should come from testing, not guesswork.

10. When ATR-Normalized Return Is A Good Choice

Use it when you want:

This is often the best first answer family when you want:

11. When ATR-Normalized Return Is Not The Best Choice

It is usually not the best first choice when you really mean:

12. Step-By-Step ATR-Normalized Return Walkthrough

Step A: Prepare the raw dataset

Open Export and make sure:

Important:

Step B: Create the prepared dataset

In AI Studio Step 1:

Fusion can now create:

Step C: Choose the actual training question

In Step 2:

Example:

Now the model is learning:

Step D: Choose the features

Pick the inputs that should help answer this question.

Good first idea:

features often pair well with this family because the target itself already uses a volatility-adjusted output scale.

Step E: Train

Choose:

Good first algorithm:

Step F: Read the output

Think of the output like this:

+1.0  = about +1 ATR expected over the horizon
 0.0  = no directional edge
-1.0  = about -1 ATR expected over the horizon

13. Good First Session

If you are new to AI Studio and want the clearest first ATR-normalized workflow:

Export

Step 1

Step 2

Step 3

That gives a very strong first continuous signed-score-around-zero workflow.

14. What Not To Do

Do not assume this family is capped like an oscillator

It is not bounded to -1 and +1.

Do not assume the ATR race settings are the same thing

The ATR race fields in Step 1 are for a different answer family.

Do not assume every large positive value means a reversal

A large positive value simply means the model expects a large positive move in ATR units over the horizon.

If the current market is falling, that can imply a rebound from here, but the target itself is not a “reversal detector.”

Do not use the line slope alone as the condition

The usual interpretation is the model output value itself relative to:

15. Current Product Notes From The ATR-Normalized Return Manual Pass

These are the places where the product could still explain this family more clearly.

A. The ATR section in Step 1 needs a stronger family boundary

Suggested copy:

“These ATR fields only affect ATR Stop / Target Race answer columns. They do not configure ATR-Normalized Return.”

B. The Look-ahead Windows help text should mention ATR-normalized return by name

Suggested copy:

“For fixed-horizon families such as Future Return, ATR-Normalized Return, Future Log Return, Direction, Up / Down, and Future Volatility, this creates one answer column per listed horizon.”

C. The product should keep emphasizing that this family is centered on zero but not bounded

This is one of the most important user expectations to set correctly.

D. A small visual scale example would help

Example UI note:

-1.0 = about 1 ATR down
 0.0 = centerline
+1.0 = about 1 ATR up

That would make the family easier to understand immediately.

16. One-Sentence Rule To Remember

ATR-Normalized Return means:

“Predict the forward move over N bars in ATR units, where 0 is the centerline and the value tells you how much upside or downside the model expects in volatility-adjusted terms.”

← All target families
NinjaTrader® is a registered trademark of NinjaTrader Group, LLC. Fusion Indicators LLC is an independent software vendor and is not affiliated with, endorsed by, or sponsored by NinjaTrader.