AI Analysis

In 2016, DeepMind's AlphaGo defeated world champion Lee Sedol, marking a turning point in both AI research and Go study. Since then, AI has become an invaluable tool for players at all levels.

Modern Go AI doesn't just play strong moves - it can evaluate positions, identify mistakes, and suggest improvements. This makes it an ideal training partner and teacher.

Sente uses KataGo, one of the strongest open-source Go AIs available. KataGo provides:

• Move evaluation: How good or bad is each move?
• Score estimation: Who's winning and by how much?
• Territory analysis: Which areas belong to which player?
• Variation suggestions: What are the best alternative moves?

Sente analyzes real games played by humans to find instructive positions. When a player makes a significant mistake (a "blunder"), Sente extracts that position as a training scenario.

Each scenario includes:

• The game position where the mistake occurred
• The best move according to KataGo
• Territory analysis, and score estimates for the top playable moves

To get the most from AI analysis:

• Try before checking: Attempt to find the best move yourself first
• Understand the why: Don't just memorize moves - understand principles
• Focus on your level: Study mistakes from games at your skill level
• Practice patterns: Repeated exposure to similar positions builds intuition

AI reviews are not perfect teaching tools. They should be used alongside human reviews and curated problem sets.

For example AI is much better at surviving in small spaces than human players. If your group is being surrounded, AI engines may be able to find a continuation that saves it, but in practice it wouldn't be possible for a human player. In review, this could show up as a later move showing up as the blunder when it would be better to learn from the earlier mistake.