Merge pull request #56 from bale-go/rank_AI

rank_AI_board_size

README.md

@@ -62,6 +62,7 @@ This section describes the available AIs, with strength based on their current O
     * **[~4d]** **Policy** uses the top move from the policy network (it's 'shape sense' without reading).
     * **[~5k]** **Policy Weighted** picks a random move weighted by the policy, leading to a varied style with mostly small mistakes, and occasional blunders due to a lack of reading.
     * **[~8k]** **Blinded Policy** picks a number of moves at random and play the best move among them, being effectively 'blind' to part of the board each turn.
+    * **[3d - 12k]** **Calibrated Rank Bot** was calibrated on various bots (e.g. GnuGo and Pachi at different strength settings) to play a balanced game from the opening to the endgame without making serious (DDK) blunders. Further discussion can be found on [this](https://github.com/sanderland/katrain/issues/44) thread.
 *  Options that are more on the 'fun and experimental' side include: 
     * Variants of **Blinded Policy**, which use the same basic strategy, but with a twist.:
         * **[~5k]** **Local Style** will consider mostly moves close to the last move.

katrain/config.json

@@ -118,6 +118,9 @@
             "threshold": 3.5,
             "line_weight": 2,
             "endgame": 0.4
+        },
+        "ai:p:rank": {
+            "kyu": 4.0
         }
     }
-}
+}

katrain/core/ai.py

@@ -20,6 +20,7 @@ from katrain.core.constants import (
     AI_TENUKI,
     AI_TERRITORY,
     AI_PICK,
+    AI_RANK, 
 )
 from katrain.core.engine import EngineDiedException
 from katrain.core.game import Game, GameNode, Move
@@ -60,10 +61,13 @@ def ai_move(game: Game, ai_mode: str, ai_settings: Dict) -> Tuple[Move, GameNode
         policy_grid = var_to_grid(cn.policy, size)  # type: List[List[float]]
         top_policy_move = policy_moves[0][1]
         ai_thoughts += f"Using policy based strategy, base top 5 moves are {fmt_moves(policy_moves[:5])}. "
+        len_legal_policy_moves = len([(pol, mv) for pol, mv in policy_moves if not mv.is_pass if pol > 0])
         if ai_mode == AI_POLICY and cn.depth <= ai_settings["opening_moves"]:
             ai_mode = AI_WEIGHTED
             ai_thoughts += f"Switching to weighted strategy in the opening {int(ai_settings['opening_moves'])} moves. "
             ai_settings = {"pick_override": 0.9, "weaken_fac": 1, "lower_bound": 0.02}
+        if ai_mode == AI_RANK:
+            ai_settings = {"pick_override": (0.8*(1-((size[0]*size[1])-len_legal_policy_moves)/(size[0]*size[1])*.5)), "kyu": ai_settings["kyu"] }
         if top_5_pass:
             aimove = top_policy_move
             ai_thoughts += "Playing top one because one of them is pass."
@@ -103,7 +107,8 @@ def ai_move(game: Game, ai_mode: str, ai_settings: Dict) -> Tuple[Move, GameNode
             )
         elif ai_mode in AI_STRATEGIES_PICK:
             legal_policy_moves = [(pol, mv) for pol, mv in policy_moves if not mv.is_pass if pol > 0]
-            n_moves = int(ai_settings["pick_frac"] * len(legal_policy_moves) + ai_settings["pick_n"])
+            if ai_mode!=AI_RANK:
+                n_moves = int(ai_settings["pick_frac"] * len(legal_policy_moves) + ai_settings["pick_n"])
             if ai_mode in [AI_INFLUENCE, AI_TERRITORY]:
 
                 thr_line = ai_settings["threshold"] - 1  # zero-based
@@ -164,6 +169,14 @@ def ai_move(game: Game, ai_mode: str, ai_settings: Dict) -> Tuple[Move, GameNode
                     for y in range(size[1])
                     if policy_grid[y][x] > 0
                 ]
+            elif ai_mode == AI_RANK:
+                n_moves = int(round((size[0]*size[1])/361*10**(-0.05737*ai_settings["kyu"] + 1.9482)))
+                weighted_coords = [
+                    (policy_grid[y][x], 1, x, y)
+                    for x in range(size[0])
+                    for y in range(size[1])
+                    if policy_grid[y][x] > 0
+                ]
             else:
                 raise ValueError(f"Unknown AI mode {ai_mode}")
             pick_moves = weighted_selection_without_replacement(weighted_coords, n_moves)

katrain/core/constants.py

@@ -20,11 +20,12 @@ AI_LOCAL = "ai:p:local"
 AI_TENUKI = "ai:p:tenuki"
 AI_INFLUENCE = "ai:p:influence"
 AI_TERRITORY = "ai:p:territory"
+AI_RANK = "ai:p:rank"
 
 AI_CONFIG_DEFAULT = AI_SCORELOSS
 
 AI_STRATEGIES_ENGINE = [AI_DEFAULT, AI_SCORELOSS, AI_JIGO]
-AI_STRATEGIES_PICK = [AI_PICK, AI_LOCAL, AI_TENUKI, AI_INFLUENCE, AI_TERRITORY]
+AI_STRATEGIES_PICK = [AI_PICK, AI_LOCAL, AI_TENUKI, AI_INFLUENCE, AI_TERRITORY, AI_RANK]
 AI_STRATEGIES_POLICY = [AI_WEIGHTED, AI_POLICY] + AI_STRATEGIES_PICK
 AI_STRATEGIES = AI_STRATEGIES_ENGINE + AI_STRATEGIES_POLICY
 AI_STRATEGIES_RECOMMENDED_ORDER = [
@@ -38,6 +39,7 @@ AI_STRATEGIES_RECOMMENDED_ORDER = [
     AI_TERRITORY,
     AI_INFLUENCE,
     AI_JIGO,
+    AI_RANK, 
 ]
 
 
@@ -52,6 +54,7 @@ AI_STRENGTH = {
     AI_TENUKI: "8k",
     AI_INFLUENCE: "8k",
     AI_TERRITORY: "5k",
+    AI_RANK: "12k - 3d", 
 }
 
 

katrain/i18n/locales/en/LC_MESSAGES/katrain.po

@@ -556,3 +556,12 @@ msgstr ""
 "Picks moves biased below the `threshold` line and plays the best one. "
 "Increase `line_weight` to penalize moves near the center more. Stops "
 "strategy after the 'endgame' fraction of the board is filled."
+
+msgid "ai:p:rank"
+msgstr "Calibrated Rank"
+
+msgid "aihelp:p:rank"
+msgstr ""
+"Picks moves at random from a limited selection of moves and plays the best "
+"one. Stronger settings select the best move from a larger selection. Since "
+"there is no 0 kyu/dan, 3 dan = -2 kyu."

katrain/i18n/locales/fr/LC_MESSAGES/katrain.po

@@ -597,3 +597,14 @@ msgstr ""
 "Le coup {move} a été annulé car il perdait {points_lost:.1f} points. Vous "
 "pouvez survoler le coup pour voir la réfutation. Veuillez jouer un nouveau "
 "coup."
+
+#. TODO
+msgid "aihelp:p:rank"
+msgstr ""
+"Picks moves at random from a limited selection of moves and plays the best "
+"one. Stronger settings select the best move from a larger selection. Since "
+"there is no 0 kyu/dan, 3 dan = -2 kyu."
+
+#. TODO
+msgid "ai:p:rank"
+msgstr "Calibrated Rank Bot"

katrain/i18n/locales/haha/LC_MESSAGES/katrain.po

@@ -529,3 +529,12 @@ msgid "analysis:nextmoves"
 msgstr ""
 "ㅋㅋNext\n"
 "Moves"
+
+msgid "aihelp:p:rank"
+msgstr ""
+"ㅋㅋPicks moves at random from a limited selection of moves and plays the best"
+" one. Stronger settings select the best move from a larger selection. Since "
+"there is no 0 kyu/dan, 3 dan = -2 kyu."
+
+msgid "ai:p:rank"
+msgstr "ㅋㅋCalibrated Rank Bot"

katrain/i18n/locales/ko/LC_MESSAGES/katrain.po

@@ -538,3 +538,14 @@ msgstr ""
 msgid "teaching undo message"
 msgstr ""
 "{points_lost:.1f}집 손해이므로 {move}을/를 물렀습니다. 커서를 올려 상대의 대응책을 보십시오. 다시 두어 주세요."
+
+#. TODO
+msgid "aihelp:p:rank"
+msgstr ""
+"Picks moves at random from a limited selection of moves and plays the best "
+"one. Stronger settings select the best move from a larger selection. Since "
+"there is no 0 kyu/dan, 3 dan = -2 kyu."
+
+#. TODO
+msgid "ai:p:rank"
+msgstr "Calibrated Rank Bot"