Currently, the elo value and the value used to calculate the new elo is updated after each game.
This leads to bad scaling when two players repeatedly play against each other (player A wins 2, then player B wins 2 leads to player B having a higher elo value afterwards when both started at the same point).
This can be improved by introducing a base elo which will be used for the elo gain calculation. The current elo value will still be saved (for ranking etc) but we keep a fixed value as a basis for the calculation.
This base value could be updated once a week (e.g.) from the current displayed elo value. This mimics the chess elo somewhat (which batches tournament, for example).
Currently, the elo value and the value used to calculate the new elo is updated after each game. This leads to bad scaling when two players repeatedly play against each other (player A wins 2, then player B wins 2 leads to player B having a higher elo value afterwards when both started at the same point).
This can be improved by introducing a base elo which will be used for the elo gain calculation. The current elo value will still be saved (for ranking etc) but we keep a fixed value as a basis for the calculation. This base value could be updated once a week (e.g.) from the current displayed elo value. This mimics the chess elo somewhat (which batches tournament, for example).