Variant ratings: Constant sum of rating

Whenever a variants gains/loses points now,
we ensure that the sum of all variant ratings of that type stays
constant by adjusting all other variants in the opposite direction.

install/database_schema.sql

@@ -265,12 +265,12 @@ CREATE TABLE variant_ratings (
   /** This should reference the game that triggered the elo update */
   league_id   INTEGER  NOT NULL REFERENCES games (league_id),
   variant_id  SMALLINT NOT NULL REFERENCES variants (id),
-
   num_players SMALLINT NOT NULL,
+
   change REAL NOT NULL,
   value_after REAL NOT NULL,
 
-  PRIMARY KEY (league_id, variant_id)
+  PRIMARY KEY (league_id, variant_id, num_players)
 );
 
 DROP TABLE IF EXISTS user_base_ratings CASCADE;

src/constants.py

@@ -43,7 +43,7 @@ FORBIDDEN_GAME_OPTIONS = [
 
 # Cache time (in seconds) for history requests of players
 # In case of frequent reruns (especially during development), we do not want to stress the server too much.
-USER_HISTORY_CACHE_TIME = 5 * 60
+USER_HISTORY_CACHE_TIME = 60 * 60
 
 # Fraction of seeds which is assumed to be unwinnable
 UNWINNABLE_SEED_FRACTION = 0.02

src/ratings.py

@@ -196,13 +196,52 @@ def process_rating_of_next_game() -> bool:
     # Calculate changes in rating
     user_changes, variant_change = compute_rating_changes(user_ratings, games_played, variant_rating, score == 5 * num_suits)
 
-    # Update database for variants
+    # Update database for the played variant
     cur.execute("INSERT INTO variant_ratings (league_id, variant_id, num_players, change, value_after) "
                 "VALUES (%s, %s, %s, %s, %s)",
                 (league_id, variant_id, num_players, variant_change, variant_rating + variant_change)
                 )
     # Note: We do not commit here, only after players have been processed as well
 
+    # Now, we get a list of all other variants of the same rating type, we will also need to update these.
+    cur.execute(
+        "SELECT DISTINCT ON (variants.id, variant_base_ratings.num_players)"
+        "  variants.id,"
+        "  variant_base_ratings.num_players,"
+        "  COALESCE(variant_ratings.value_after, variant_base_ratings.rating) AS current_rating "
+        "FROM variants "
+        "LEFT OUTER JOIN variant_base_ratings"
+        "  ON variants.id = variant_base_ratings.variant_id "
+        "LEFT OUTER JOIN variant_ratings "
+        "  ON variant_ratings.variant_id = variant_base_ratings.variant_id"
+        "  AND variant_ratings.num_players = variant_base_ratings.num_players "
+        "WHERE"
+        "  variants.rating_type = %s"
+        "  AND NOT"
+        "  (variants.id = %s AND variant_base_ratings.num_players = %s)"
+        "GROUP BY ("
+        "  variants.id, variant_base_ratings.num_players, variant_base_ratings.rating,"
+        "  variant_ratings.league_id, variant_ratings.value_after, variant_base_ratings.num_players"
+        "  ) "
+        "ORDER BY variants.id, variant_base_ratings.num_players, league_id DESC",
+        (rating_type, variant_id, num_players)
+        )
+
+    # This ensures that the sum of ratings of variants of the same type stays constant (up to rounding).
+    variants_of_same_rating_type = cur.fetchall()
+    other_variant_change = -variant_change / len(variants_of_same_rating_type)
+    other_variant_vals = []
+    for (other_variant_id, other_num_players, current_rating) in variants_of_same_rating_type:
+        other_variant_vals.append((league_id, other_variant_id, other_num_players, other_variant_change, current_rating + other_variant_change))
+
+    # This updates the variants now
+    psycopg2.extras.execute_values(
+        cur,
+        "INSERT INTO variant_ratings (league_id, variant_id, num_players, change, value_after) "
+        "VALUES %s",
+        other_variant_vals
+    )
+
     user_ratings_vals = []
     for user_id, change in user_changes.items():
         user_ratings_vals.append((league_id, user_id, rating_type, change, user_ratings[user_id] + change))