| prompt_core = """You will be provided with a chess game in PGN format, along with a detailed comments. | |
| Task: | |
| 1. Your task is to rewrite the comments based on your role. You need to maintain the original meaning of the comments while adapting them to fit your role. | |
| 2. Only rewrite the comments, do not change the move numbers or the moves themselves. | |
| 3. Do not make up comments or variations, only rewrite the provided ones. | |
| 4. The player who played the move is also provided, use this information to adapt the comments accordingly. | |
| 5. In your comments, as necessary - specially after a sharp change - indicate overall better/worse position after the move from engine evaluation score after the move. Do not use the engine evaluation score rather interpret it in a human-readable way, e.g. "White is better" or "Black has a slight advantage". | |
| Do not bother if the score is not significant, e.g. less than 50 centipawns. | |
| 6. Your output must be a valid json object. | |
| 7. Your output must only contain the comment_refs and the rewritten comments, without any additional text or explanations. The additonal informations are provided for you to understand the context of the game and the comments. | |
| Game in PGN format: | |
| {pgn} | |
| List of comments below. Each comment contain the comment_ref, the move number, the actual move, the player who played the move, a comment, a suggested better variation (can be None), engine evaluated score before the move (in centipawns), engine evaluated score after the move (in centipawns). | |
| Rememeber, the engine evaluated scores are from the point of view of the white player. Positive scores indicate a favorable position for white, negative scores indicate a favorable position for black. | |
| {comments} | |
| """ | |