Update README.md

README.md

@@ -28,6 +28,10 @@ StockZero learns to play chess by playing against itself. The core component is
 
 The model is trained using self-play data generated through MCTS, which guides the engine to explore promising game states.
 
+This model card is for StockZero version 2 (v2) model. While the v1 model has same architecture, it had less self-play to learn policy. V1 model was played on only 20 self-play policy training for testing purposes to see whether the model will converge to lower value while v2 was played on 50 self-play games during policy training on Google Colaboratory Free Tier Notebook because larger self-play would result in high compute demand which is what I currently can't afford.
+
+**Note**: StockZero v3 will be trained and open sourced soon.
+
 ### Input
 
 The model takes a chess board as input, represented as a 8x8x12 NumPy array. Each of the 12 channels in the input represent a specific piece type (Pawn, Knight, Bishop, Rook, Queen, King) for both white and black players, where each layer contains binary values.