I feel like, there will be other threads about this issue, but I could not find them.
If I get it right (according to https://www.remi-coulom.fr/Bayesian-Elo/#history
), the system of "Bayesian Elo Rating" is made for one tournament in order to calculate the elo changes more accurate.
This means, that the system how it is implemented at the moment, states that all games played for the ladder in the last 5 month are one tournament.
This has extrem impact on the ratings:
All games are counted equal. First this might seem logical, but the skill of a player may change extremly within the 5 month and then this becomes a big problem. Newer games give a more precise impression on the players skill at the moment.
The current rating simply state the average skill level of a player for the last 5 month and I have to confess:
I do not like that, as the skill level of many players does change extemly within 5 month.
So I would ask to change the rating-system.
If there are other threads about these topic, please link them here.