Walk any decision tree.
Drop in a CatBoost model JSON and this site turns its splits into a guided questionnaire — one yes/no question at a time. At the end you get the predicted outcome plus the model's confidence across every possibility. Works for any tree, any topic.
How this works
A machine-learning model reached its prediction by asking itself a chain of yes/no questions. Instead of hiding that, this page asks you those same questions, one at a time, then shows what the model concludes.
🧺 Why pick a range, not an exact number?
The model only ever checks thresholds (“is it above 30?”), so a range is all it can actually use. Choosing a bucket gives the model exactly as much information as typing a precise number would — no more, no less.
🤷 “I don’t know” is allowed
Not sure? Pick “I don’t know”. The model then tries every possible answer and averages the outcomes, weighted by how common each was in its training data. The final number stays an honest probability instead of a guess.
📊 Reading the result
The big percentage is how confident the model is in its top pick. The bars below show every possible outcome; for a classifier they add up to 100%. A close race means the model isn’t sure.
Nothing you enter leaves your browser — it all runs locally.
What am I picking?
Each step is one question the model needs answered to decide where to go next. For numbers you choose a range; for categories you pick the matching option (or Other if none fit).
The small tag (e.g. “Depth 2 of 4”) just shows how far into the model’s decision chain you are — you don’t need to think about it.
Unsure? Hit “I don’t know” and the model will weigh up every branch for you.
How to read this
The headline is the model’s top pick and how confident it is. Each bar is one possible outcome and its probability — a taller bar means more likely. For a classifier the bars add up to 100%.
If a yellow note appears, some answers were “I don’t know”, so the number is an average across every branch those unknowns could lead to.
Your path through the tree
This is the trail of answers the model used, in order. Striped “unknown” rows are the questions you skipped — the model explored both directions and averaged them.