AMGesus AI team builder

This is a tech preview of an experimental team builder that uses AI. We trained a binary classification model to predict if a battle between a Team_X vs Team_Y results in a win or loss. It was trained on hundreds of thousands of samples collected by our fine GO Battle Log members.

To build teams using this trained AI model, we try different combinations of mons to form candidate teams. Then, we ask the AI model to score these candidate teams against the most popular meta teams. The results are teams the AI model says have the highest probability of winning.

great ultra master

Teams to counter. This list is auto-filled with the most popular teams. The order of the list matters. We try harder to beat the teams at the top of the list. A smaller list will help the AI build more specific counter teams. A larger list will have more trade-offs. Try playing with this list to see how it changes the teams the AI builds.

Candidates. This list is auto-filled with the top-ranked pvpoke mons. Try editing this list to control which pokemon the AI team builder uses.

Number of teams to generate

Generate teams

---

Team vs team win predictor

Test the trained AI model -- ask it to predict the outcome of a battle between two teams. You may enter 1, 2 or 3 pokemon on each side.

Your team

Opponent's team

Predict outcome