Added tournaments
ROTC 2 - London Elite/Open divisions
ROTC 2 - London Womens division
ROTC 3 - Waterloo Elite/Open divisions

*You can see any direct changes to your rating at the bottom of this post.

Previously, determining a new player's starting rating was very straight forward: If your first tournament was in the highest possible Open division, you were assigned an initial rating of 1000, otherwise your assigned rating was 500.

This became slightly complex with the introduction of Ontario's Elite division in 2022. If a new player were to begin playing in Ontario and was not immediately invited to the Elite division (eg. they recently moved to Ontario), they would be forced to play in the next lowest division. With the mindset of 'it's not their fault they can't play in Elite' we decided that if your first tournament is in the highest division EXCLUDING any closed divisions (ie. Ontario Elite), you would still be assigned a rating of 1000.

This means brand new players would be assigned a rating of 1000 in Open Advanced if there was an Elite division present

With the re-introduction of the Elite division in Ontario for the 2024 season we noticed some unwanted side effects after ROTC 2 - London. Specifically we noticed relatively good players rated 1000 or higher who happened to not be playing in Elite for whatever reason were gaining much much more rating than deserved in the open division, simply by beating first time players that were inaccurately rated at 1000 because the top division was closed. This was happening to the extent that players who are not playing in Elite were creating their own closed off ecosystem (or rather ELOsystem xD) and were underservingly surpassing players who play in Elite.

The Solution:

To combat these unwanted side effects of good players playing against new players, we have decided to retroactively make any new player's starting rating 500 if their initial tournament division played is not the highest division. This includes tournaments such as the ROTC stops in 2022 and 2024 where the top division is invite only.

Along the same vein, we have also made the decision to subject the RO University Championship to a similar treatment, based on a subjective analysis of the skill present at the tournament. This means if your first tournament was the RO University Championships, your initial rating has also been retroactively changed from 1000 to 500. This also means if you happened beat a lot of new players at this tournament, you will have gained less.

These are the 102 players that have had their starting rating directly changed from 1000 to 500

These are the overall effects of the initial rating change on active players as of May 31 2024 (Post 2024 ROTC 2 - London)

One of the biggest flaws of this rating system we aimed to fix was the effect of new players on the system. Not only could a new player heavily influence the ratings of players they play against, but their rating would also be inaccurate, had they not played many tournaments.

A trivial solution to this would be to eye test each and every new player when inserting them into the system. This method is laborious, potentially biased, and even with best intentions in mind, would be very inaccurate in some cases.

Rating Reruns are an attempt to resolve this and quickly adjust those players who are far away from their deserved ranking. This concept was originally introduced in the Roundnet Germany Index (RGX).

How it works:
When a player that is new or inactive plays a tournament, they are subject to a Rating Rerun. We will use the results from this players tournament to assign them a more accurate starting rating before we calculate everyone else's rating change. Let's look at the following example:

Lets say, for sake of simplicity, we have a tournament with 2 teams who are playing against eachother. On one side we have a veteran team where both players are ranked at 1000 (team average of 1000). On the other side we have a player named Bike Burray. Bike, being the community leader and teacher that he is, decides to play with one of his students. This is the student's first time at a tournament, and realistically should be rated around 500. Regardless, they get automatically assigned a rating of 1000, as per usual.

Player	Rating
Veteran 1	1000
Veteran 2	1000

Player	Rating
Bike Burray	1000
New Student	1000

Over the span of an entire tournament, Bike Burray's team loses a lot. Lets say for example at the end of a long tournament the final ratings look like this.

Player	Rating
Veteran 1	1200
Veteran 2	1200

Player	Rating
Bike Burray	800
New Student	800

This final result isn't as accurate as it should have been. Bike Burray shouldn't be losing this much rating just because a new player starts at 1000 rating.

Now with Rating Reruns, we can better estimate this new (or inactive) player's rating. We will take this players rating after the tournament (800) and rerun the entire tournament with this updated value but all other players will return to what they were originally. This happens with all new or inactive players in the same run. Now the ratings of the players at the start of the tournament look like this:

Player	Rating
Veteran 1	1000
Veteran 2	1000

Player	Rating
Bike Burray	1000
New Student	800

And at the end of the tournament, after running the results of all the exact same games again, the final ratings look something like this:

Player	Rating
Veteran 1	1125
Veteran 2	1125

Player	Rating
Bike Burray	875
New Student	675

This is not perfect, but it in system unlike chess, for example, where there are much fewer games played and more games played with and against the same people, trying to more accurately predict a players starting rating is important. This same example is applicable to a case where 'New Student' is replaced by a player that 'used to be great but now is literally over-rated'

Similarly, the opposite affect would take place if we replaced 'New Student' with a player who is 'inactive but got really good since the last time they played a tournament'.

This change is not retroactive, for the sake of not confusing players when their ratings are changed after playing no games, but is in effect indefinitely from now on.

There is some room for experimentation with this feature. We could continuously Rerun a players rating until the results of their first tournament converge to a certain rating +/- a threshold. However this would put an immense value on a player's first performance and would be easily influenced by the rating of players they play with and against. There is a balance to be found between a single rerun and an infinite converging number of reruns. After testing and examining some specific cases where veteran players return to the scene overrated, we believe a single rerun is helpful and more accurate than not.

Added March Madness and Spring Spikes tournaments for effective rankings for April 1 2024.

Updated Open 'active' requirements to require at least 2 tournaments played in the last 9 months and to require at least one ROTC tournament played in the last 9 months. Previously was at least 1 tournament in the last 6 months and at least 2 tournaments in the last 12 months.

Updated Women's 'active' requirements to require at least 1 tournament played in the last 9 months. Previously was at least 1 tournament in the last 12 months.

This should better reflect who is truly active in the sport, while considering limiting factors such as the off season, important tournaments, and lack of Women's divisions.

Added Family Day Open tournament. Generated effective rankings for March 1 2024

Implemented feature allowing for generation of tournament reports by player.

Added ability to generate lists including or excluding inactive and/or nonlocal players.

Implement use of tournament multiplier when inserting and calculating tournaments. Retroactive.

Pushed all ranking lists and raw csvs to remote.

Fixed a bug where recalculating rating on a given date would not accurately pull the players starting rating.

Initialize players in the top 31 2020 rankings to have a scaled value based on community voted results. Help kickstart the system with more accurate data then simply starting everyone at 1000.