[Suggestion] Elo decay and Elo from AI

Yes.

 

I was not attempting to exaggerate at all. I did not fully understand that you wanted to take away all skill ranking and instead are just counting the number of games someone has played through a two steps forward one step back process. I had figured there would still be actual calculations done based on the difference between your ranking and your opponents which would determine your ranking gain/loss.

The fact that I could be ranked higher than a highly skilled player so long as I can win 50% of my games and play twice as much as they do is, to me, very silly and pointless.

My main concern was a new player coming in and having 0 points, them seeing the scoreboard and seeing a very high number like (going with real current leaderboard numbers from just beta) 2426. They then play a game and win. They now have 2 points and realize they will never be near the top of the leaderboard because they would have to win over 1400 games without a loss to get there in the shortest amount of time. They might (I would certainly) decide to not even try to compete, lose interest etc.

 

since when do you get 2 points when you win you first match?

Not to mention, how is this different from whats happening right now? They see people with 1k-1.5k ratings all the time.

Except that won't happen.

If you get ranked too highly due to your # of plays, you start meeting players who are actually better than you. You then get knocked back down.

A higher ranked player will only need to return and win a few games to catch up. Its 15-25 to 1, or even higher, depending on the formula adjustments. Him winning once = you playing 15-25 games. If inflation happened during his hiatus, then he should be winning easily when he comes back and is matched up against less skilled players (who have been playing) initially.

You start at 1000, he starts at 1000. You play 500 games, you get to 1500. You start meeting highly skilled players due to inflation, you lose subsequent matches and drop to 1200. Lets assume thats the equilibrium (the play too much and get stomped thing is constantly happening). He comes back, he wins easily (because his elo is actually is pre inflation) and he wins 20 matches against rating 1000 players (who would have been to be rating 800 in the past), he gets bumped to 1200, same as you again.

This makes his elo incomparable, which is the whole point. It also doesn't penalize him too harshly when he comes back. Neither does it allow for infinite inflation. As long as players are relatively active, the "per game gain leads to stomping by better skiller opponents" mechanism keeps them on relative equal elo with people of the same skill level. there is no way for you to actually gain elo from just playing and not lose it when the impact of that inflation bits you in the ass by pairing you up with better opponents.

this is very simplified. But it illustrates the point. you can optimize it further, maybe add seperate sort of decay to the win +1 bonus. but the general idea isn't unsound.

 

Starting at 0 and seeing the top players at or around 1600 is a lot less intimidating than 2500

 

At some point, your ELO no longer measures your skill (around 1400 or so) and the leaderboard is definitely there to give glory to top players. If you are a 1200 rated player and you lose 100 points due to ELO decay then there is probably no issues getting that back. BUT, if you are a 1600 player and there is a 1700 inactive player in first place it is almost impossible to farm up the points to overtake them. Why should we allow someone to farm up easy points at the beginning of the live release and sit on top without any competition (which is exactly what has happened in beta)?

 

The point I was trying to make is that both goals of the ranking system (ie. accurate matchmaking + glory for top players) can be satisfied by including an estimate for skill atrophy. So we're not in disagreement about this. I'll repeat my suggestion again - because I think it would work well:

I originally said 10%/week, but I think 5%/week would probably be better. The drop doesn't have to be rapid to be effective - and I'd expect the real level of skill atrophy to be quite slow.

(And the other part of the suggestion, to address the skewing of AI player ratings, was this:

In that post, I elaborated a bit on what that means and why I'm suggesting it.)

 

and those are exactly what we need.
slowly lowering inactive players rating to match the average rating will certainly not result in any uneven match ups.
A player that hasn't player for a month probably also lost some "skill" in that time, or is just not up to date with the current meta.

preventing elo gain for multiple consecutive wins against AI players also solves the potential AI abuse issue.

i really think that those 2 mechanics should be implemented.
the first one is far more important than the AI issue though.

 

Im not that much into elo-rating. Nevertheless its obvious that even the "cheating" mom gets very easy to beat after you got the cards and know how to counter her.

It's still nice to have the AI there if for some reaoson there isnt enough human players so you can keep on looting the mp chests. Sure you can remove the elo-rating changes from AI opponenets, thats only fair for the more competetive players.

Summary:
Keep AI-opponents (you can check the box still to disable them)
Remove ELO-raiting changes from AI-opponents

 

Claiming my bias: i like to frustrate people who play for epeen because i think excessive epeen makes people behave badly, so rank for matchmaking is fine, but leader boards annoy me. So although i am sympathetic to people who want to go for top rank and are stymied by inactive players, as noted above, i don't actually like the whole idea of leader boards.

Maybe it would be enough to decouple the leader boards from the rank system. Don't include an absolute rank leader board - maybe just have daily, weekly and monthly ones that only count wins and losses during that window and leave it at that?

I suggested a rank loss floor in my original post to prevent intentional deranking and newb stomping. That kind of got ignored, probably because it was too low. Think about ways to calculate a floor that make it better. Someone suggested putting it at the average rank of all players, or maybe it should be half the distance to 1k (1600 floors at 1300) - whatever works.

Maybe do something drastic like hide ratings completely (keep them behind the scenes for matchmaking), and just maintain results from periodic tournaments?

I don't know - maybe my bias should disqualify me from being taken seriously on this issue. Maybe the people who are really invested in rankings and leader boards should be the ones who influence policy here. Maybe Blue Manchu needs to specifically ask those people for feedback instead of just relying on those of us who are active on the forums.

 

I am with Mutak on the idea of short-term leaderboards.

1) You don't have inactive people hogging the epeen.
2) You DO have brand new to the game players who happen to play a match right after reset get to see themselves in the top 10/50/whatever.

This second can be a HUGE recruitment incentive to PvP. I very rarely bother with PvP in any game I play outside of milking any "low hanging fruit" gains that I can use elsewhere. But in one of the games I had played they had a few competitive events which were very short window (weekly PvP windows, and a 2 hour "who can do the most damage" event every day). Happening to see myself in the top 10 on one of those lists one time made me actually start caring about competing. I didn't make the connection until long after I stopped playing for why I played so differently than normal, but that was what caused it: Because I was, for a brief moment, top ranked.

And to defend Turin's idea from the end of last page, he wasn't saying to give out free points, just to lose 1 less point when you lose a match. So you would not face people of immensely greater skill than yourself, when your elo got up above your actual skill (which it will do in any elo system, by design) you would just play slightly more games against those better opponents before you are back in your "comfort zone" elo range. The elo system as a whole would creep up to be centered on higher numbers, but no one person would be sitting at an anomalously high value (remember, you "gain" the extra point only when you LOSE points from a failure, so you still wind up rated lower than you had been, just not AS much lower).

Of course, CH doesn't work on a "start at 1200, and every match exchanges elo from loser to winner" system, so it is largely pointless to have thought up and debated. There is elo creep already in place (at one point, PvP had just been added, and the "average" ranking was 0... current average is certainly higher than that)

 

Without overhauling the current system, I think a light elo decay that stops at a certain point would be nice. Something like:

1% loss after each week without playing, with a maximum loss of 10 percent.

I think the loss limit is important. Imagine being a 1500+ player not playing for a year, and then coming back and having to wade through new players to get your rating back up. That's not fun for you, or the player's you have to wade through. So if you didn't play for a pretty long time, and the loss limit was 10%, your rating would be:

1600 -> 1440
1500 -> 1350
1400 -> 1260
1200 -> 1080
1000 -> 900

Those numbers seem about right for a player who's probably going to be a little rusty. The only issue I can see is a player coming back after a long time, playing a game, and then not playing again for a while. Would the loss limit be reset? hmmm...

 

What about just hidding people on the all time ranking if they have for example less than 3 games this week or 10-15 this month?

 

Just steal from other competitive games, small amount of elo decay over time when inactive, change up map pools and reset stats after a couple months, implement regular tournaments, i.e season end tournaments, and supplemental weekly/bi-weekly tournaments and/or qualifiers