Status: Per-family deep reference (companion to the consolidated Fusion AI Studio & ONNX User Guide) Audience: first-time AI Studio user Answer family: Swing Reversal Prediction style: Classification
Swing Reversal asks:
“Does this bar later become a confirmed swing low, a confirmed swing high, or neither?”
This is one of the most visually intuitive answer families in AI Studio, because most traders already understand the idea of:
But it is also one of the easiest families to misunderstand.
The important thing to know is:
this is a confirmed-turn label, not a zero-lag pivot detector
Example:
Target_SwingReversal5Meaning:
window = 5 bars on each side
Fusion checks whether the current bar becomes a confirmed turning point using bars before and after it.
Plain-English rule:
+1 if Low[t] is the lowest low in bars [t-5 .. t+5]
-1 if High[t] is the highest high in bars [t-5 .. t+5]
0 otherwise
That means:
+1 = confirmed swing low-1 = confirmed swing high0 = no confirmed swingThis family uses future bars for confirmation.
That means the label is answering:
“Will this bar later prove to be a confirmed turning point?”
It is not answering:
“Can I know in real time, on this exact bar, that the pivot is already confirmed?”
That is why this family is best thought of as:
not:
This family uses its own Step 1 section:
Enable Swing Reversal AnswersSwing Confirmation Windows (bars each side)For this family:
Fixed-Horizon Windows (bars) does not create these labelsThis is one of the places where the UI needs to be more explicit.
If the swing confirmation window is:
5
Fusion checks:
That means the full pattern spans:
11 bars total
This is the main design choice for this family.
This is a classification family.
The correct trader meaning is:
+1 = confirmed swing low
0 = no confirmed swing
-1 = confirmed swing high
The saved model contract may still need class decoding to translate the encoded class id back into the trader meaning cleanly.
So the correct mental model is:
treat this family as +1 / 0 / -1 turn labels
not:
treat the raw plotted encoded number like a smooth oscillator
0 Means HereFor Swing Reversal, 0 means:
So 0 here means:
“No confirmed turning point.”
That is different from 0 in:
Future ReturnATR-Normalized ReturnATR Stop / Target RaceThis family often matches how traders think about chart structure:
It can be easier to understand than raw return prediction when the real question is:
“Is this the kind of bar that becomes an important turning point?”
This family depends on:
HighLowIn practice, the safest rule is:
OHLCV on in ExportIf High or Low are missing from the captured export, this family cannot be created correctly.
This is a very important limitation for users to understand.
Because confirmation needs future bars on the right side of the pattern:
That is not a bug.
It is a direct result of how confirmed swing labels work.
Plain-English version:
you cannot confirm the last possible swing until enough future bars exist
Direction (3-Class)Asks:
“By the end of the horizon, is the move down, flat, or up?”
Swing ReversalAsks:
“Does this bar later prove to be a confirmed turning point?”
So Direction (3-Class) is about the final directional outcome.
Swing Reversal is about whether this bar later becomes a structural turn.
That is a different kind of question.
Open Export and make sure:
OHLCV is checkedIn AI Studio Step 1:
Enable Swing Reversal AnswersSwing Confirmation WindowsExample first pass:
Swing Confirmation Windows: 3,5That lets Fusion create both a faster and a slower swing-label version.
Click Create New Capture.
Now Fusion creates answer columns such as:
Target_SwingReversal3Target_SwingReversal5In Step 2:
Prediction Style: ClassificationAnswer Column Source: Generated Answer ColumnsGenerated Question Family: Swing ReversalExample:
5Now the model is learning:
Target_SwingReversal5This family is classification-only.
If only swing-reversal targets are available in the capture, regression should not be the active path.
Pick the inputs that may help answer:
“Is this bar likely to become a confirmed turn?”
Useful first feature types often include:
Good first rule:
Choose:
Good first algorithms:
LightGBM Classifier if availableRandom Forest ClassifierThink of the output like this:
+1 = confirmed swing low pattern
0 = no confirmed swing pattern
-1 = confirmed swing high pattern
And remember:
this family uses future confirmation
So the label itself is intentionally lagged relative to the bar being classified.
If you want a clean first experiment with this family:
Enable Swing Reversal AnswersSwing Confirmation Windows: 3,5Prediction Style: ClassificationAnswer Column Source: Generated Answer ColumnsGenerated Question Family: Swing Reversal5Algorithm: LightGBM ClassifierRandom Forest ClassifierThis gives one of the clearest “turn-recognition” workflows in AI Studio.
This family uses future bars for confirmation.
They may not have enough future bars yet for confirmation.
0 means a small continuous valueHere, 0 is a class meaning “no confirmed swing.”
This is a classification family built around bounded turn labels.
These are the product improvements this guide exposes.
Suggested note:
“Swing Reversal is a bounded class target for confirmed swing low / no swing / swing high. Use Classification, not Regression.”
Suggested note:
“Swing Reversal uses future bars on both sides of the bar for confirmation. It teaches the model to recognize setups that precede confirmed turns, not to detect the pivot with zero lag.”
Suggested note:
“Because confirmation needs future bars, the final bars in a captured dataset cannot be labeled for this target.”
Swing Reversal means:
“Predict whether this bar later becomes a confirmed swing low, no swing, or confirmed swing high using future-bar confirmation on both sides.”