My full support to removing CD's after 1 year or 20 games (whichever is larger) is here.

im open to anything that can alleviate some of my older games from when i first joined the site.

Well, you would, Mr 94% RR.

That was for HR, not ssorenn. <3, ssorenn.

Actually 94 > 92, not <3

<3 RR is a really bad RR to have

i almost included you HR in my post, but you ninja'd me anyway

+1 I support this

Something to consider. FIFA's ranking system for international soccer (futbol) has a sliding scale where older performance is weighted less than more recent performance. For example, the last year's games count as 100% towards your score, and decreases over five years until the fifth year is only weighted as 20%.

I am not proposing a five-year sliding scale for RR, but instead present an example of weighting it. For our purposes, let's say 6 months of history counts 100% toward RR, the previous 6 months as 60%, the 6 months before that as 25%, and then anything older than that would not longer be considered.

The length of term before each step can be changed to whatever is thought best. 6 months could be a year, or each step could have a different length of time, say 1 year is 100%, 9 months at 50%, and 4 months at 25%, or whatever the Universal Hive Mind of WebDip determines is suitable.

I say last 6 months or last 20 games whichever score is worse.

Also once the coding is in i would assume it's pretty easy to change 6 months to 12 or whatever and same for last X games. If we find its too short/long change it.

Against the idea... What is to say someone who goes on an unexpected hiatus wouldn't repeat his actions again?

This is covered by the "last X years or Y games, whichever is more games". If you disappear from the site even for more than X years, then your past behaviour still counts.

I think you just have a model weighting newer games more.
(Old RR x .95) + (Latest game RR x .05) = new RR score.
*. The 5% weighting pulled from thin air.

This means bad results are never completely removed, but a result is always most influential right after it happens.

I would do just a straight-up weight based on number of games. I don't really see why years matter at all. Either you play frequently and they will drop off naturally due to the number of games, or you don't play frequently and it doesn't matter because of your "whichever is more" clause. Simplifies the equation without taking much away.

So that you can't play 25 live games in a month to fix your RR, whilst NMR'ing repeatedly in longer games, Abge

Yes. The time constraint also prevents players from (reliably!) killing themselves in the first three years just to up their RR.

Is that really a problem, though?

First, as people have said, live and nonlive RR would be really nice (and address this issue).

Second, if someone messes up their RR and is willing to play 25 live games to fix it, is that such a bad thing? Will people really play 25 live games just so they can CD in a nonlive game?

Abge, my objection isn't restricted to live games.

I don't think I understand what your issue is.

If it is just "last N games", then players can join games and intentionally kill themselves to have an easy reliable game.

"last N games, or Y time, whichever is more games" does not suffer from this problem.

Oh, suiciding in game to quickly rack up games. Yup, that indeed would be an issue. Of course, this would tank their "GR" so that approach would have its own consequences.

It is very important that whatever we choose doesn't incentivise perverse behaviour in game. The same goes for the skill ratings.

Agreed. What were you thinking of for a length of time?

""last N games, or Y time, whichever is more games" does not suffer from this problem."

Technically, people could just suicide N games and then wait Y time, although this is probably much less likely.

I think that if the Rating system integration discussed in the other awesome thread was implemented we may not have a need for a Y time. As if players suicide themselves their rating would go down. If both RR and "GR" were gates to game entry then suicide wouldn't technically open more games for you. As a high level game with a RR requirement of 80% and a "GR" requirement of X would be impossible for you to enter if you destroyed your "GR". Although I am not against Y time as a requirement; simply curious if removing Y time would make the implementation easier on the devs.

So, the gunboat tournament was bullshit. I, along with other players, were excluded because of RR, a criteria introduced simply because it exists. This would be well and fine if I felt I was actually unreliable - but I'm not. Almost every one of my CDs or NMRs is from a live game, which is a notoriously difficult thing to never miss a phase in - and additionally, I haven't had a civil disorder in a non-live game since... wait for it... 2011. (I will admit I NMR'd a lot in the most recent batch of Lusthog games, but this was mainly due to the holidays).

The RR system as is is bullshit. Players are punished for games that happened years ago, for live games where your internet cut out or you got tied up dealing with a dog that won't come in, or when you really had to go do something. It happens, it's a fact of live game life, and it's not indicative of actual reliability.

My proposals:
1. Separate RR into a live-RR and a non-live-RR. Live RR will be more fluid, with more recent games being weighted heavier. The scale would be something like this:
6-12 Months Old: Counted at 0.10
3-6 Months Old: Counted at 0.20
0-3 Months Old: Counted at 0.70

Basically, your live game RR would reflect your average reliability over your recent games, but not completely ignore past unreliability. However, any game over 1 year old will be dropped from calculations, to allow for a gradual fix over time.

2. Have non-live RR decay by dropping any game older than 2 years old, and have games older than 1 year be weighted by half.

3. Weight RR against phase length for non-live games. An NMR 10-day/phase game would be much worse for your RR than an NMR in a 12-hour/phase game, for example. This makes it so if someone NMR's on occasion in short-phase games, it's not super detrimental to their RR, but when someone NMRs in a long-term game, it's super punishing, because it should be.

However, the most important thing here is some sort of decay on RR. Even if it is as simple as only counting games from the last year or two. Some of us have been on this site for 6, 7, 8 years and a lot of those old games are from when we were in high school (or elementary as in bo_sox's case, probably).

I agree with Chaqa that a lack of 2 RR (love/nonlive) is the biggest concern