Bringing meaning to MMR

[LeeRomeno]

I personally love MMR, it has kept zone alive for euros and is generally around 1000000 times more entertaining then elim (which was only thing alive before that). It is interesting to see the ratings go up and down, but I have a feeling this could be improved dramatically if we bring in some sort of seasonality into it.
When i started in subspace, the main competitive thing (at least in EG), that everyone cared was winning points reset. So after a while public points were reset and results published, so there was a clear winner in both individual as well as squad level.

I think MMR would benefit from similar system. I do not think resetting scores is correct way to go in MMR, but what we could do is have either monthly or quarterly "seasons" with winners announced after each ends. Scores at the end are stored. Then next season starts and previous season results are starting point for new one.
It would also be nice to have some sort of stat page, where you could check different cool stats and awards (twdt "achievements could work as inspiration here). You can look at total scores (best rating fe), but can also look at who improved most during season, won most games etc.

[Rab]

Mostly agree with the sentiment of wanting more meaning. Other thoughts on the topic of "meaning":

1. Aliasing still happens. Players can have multiple names on the website ranking table which should never happen. If you change names your record should follow you.

2. Results are often determined by things other than how people played. e.g. pub noobs throwing games or people starting at 3k. There's a knock-on effect as some players recognise these situations and don't !p up.

3. Players who haven't played in x time should drop off the leaderboard.

[LeeRomeno]

Mostly agree with the sentiment of wanting more meaning. Other thoughts on the topic of "meaning":

1. Aliasing still happens. Players can have multiple names on the website ranking table which should never happen. If you change names your record should follow you.

2. Results are often determined by things other than how people played. e.g. pub noobs throwing games or people starting at 3k. There's a knock-on effect as some players recognise these situations and don't !p up.

3. Players who haven't played in x time should drop off the leaderboard.

1 - this i agree, otherwise u will have some people (not naming names, you know who), who will get to some score and then store it forever to see their name there. But emphasis would be not in all time scores, but actually scores based on last reset. All-time greatest achievements (like wins in a row or highest rating ever etc) could be used in some HOF board or similar

2 - this more or less happens to everyone, so over long enough time it balances out. But i certainly believe that new player starting rating should be lower (if that player is aliased checked) and there should be very clear rules of what determines throwing. I have seen some wild subjective actions from hosts. Some kill game, some ban the supposedly thrower, some just ignore etc.

3 - I think there should be some minimum games played per reset filter that would only count active players.

[Rab]

2. I will elaborate, to explain why even though it happens to everyone it doesn't balance out:

If I'm a player with 60% winrate under normal conditions.
If the winrate with Pubber123 signed up is determined by which team gets them. For argument's sake let's say that winrate is 50% because they might be on your team or not on your team.
In that example, if you sign up when Pubber123 is playing your winrate is lowered, so you're better off sitting.

That example is totally dumbed down, but is roughly true. More complicated logic:

You can sort of predict the draft outcome by knowing who's playing. You can lookup their ratings. Staff have an advantage of perfect knowledge.
You can learn this by playing, then knowing whether based on your rating whether you're likely to be teamed up with the newbie again.
e.g. if you're rated like vys, then you get shitter123 on your team almost guaranteed, because the algo is trying to balance the ratings out.

It also varies by arena and time.
jd3 eu = shitter on your team = loss
jd3 late eu / early usa = shitter on your team = a very strong rest of your team = maybe win
jd3 usa = shitter on your team = loss
jd3 when jd4 is running = shitter is in jd4 = do a calculation which arena you're likely to be in = if you're likely to be in jd4 with the shitter, do a calculation whether you can dunk all the other jd4 shitters and win regardless or not.

Players who sit these sort of situations will do better in terms of mmr rating. You also have to try to calculate who will do these strategies and who will play regardless. Because if you have the ability to draft fair teams, but a shitter shows up, so someone sits, that makes the teams less fair, and you have to recalculate any previous calculations with this new information.

I agree that hosts are weird sometimes. Defo more likely to have a game killed if the shitter is on the host's team. Which makes it more worth taking the 50% risk if you think you're likely to be on the hosts team, but only if the host isn't also a shitter.

Basically. this post is shit and gay. and I don't want to have to think about any of this.

And ss nerds have form - it's the same sort of thing as people leaving when they think there aren't enough in the arena, making the prophecy a reality.

This is the part where someone says "Rab you're a nolife loser and if you actually got good at the game it's the best way to improve your rating". yes I know.

[cubone]

This gives me an idea. It would be interesting to look at average scores for games based on the makeup of teams. For example, what's the average score of a game with a 4000 rating average where all players are in the 4000 range vs games where there's a 4000 average but you have someone like Ease (6k) paired up with someone in the 2k range. Look at the range of ratings (high minus low) and then look at the average scores.

[Rab]

tricky to do I bet, but a good idea worth trying.