@sqrg: The chance of missing a phase was calculated that way, but we counted the missing phases twice because missing a phase does much more harm than not missing a phase.

The "chance of not leaving" is a good idea, but I want each CD have the same weight as it really screws the games and instead of finishing many games I think it would help more if you just take an open spot. Also that way you can't earn a bonus if you finished many games before you go CD in a game.

PS: You meant "phases missed / phases played, right"?

I have 1783 without a miss, but figured since I was gone for a few months I was still way behind.

@Oli: no i meant what i wrote down. #phases missed / # games finished.
What you describe is the OLD way the rating worked.
To calculate the chance of someone missing a PHASE in your GAME. You need to get the ratio of PHASES missed to GAMES played. The old system did not wor because the ratio of phases missed to phases played is always very small. Playing a game against a player will take multiple phases so there is a chance he will miss every phase. My calculation takes the whole game into account.

Ah, ok. I see now.

@Decima Legio: Thanks for the comment! About your question of # phases played are ignored. Theya re in fact not, haha. So nice of you to notice, i was wondering if someone would... let me explain.

I assume you've read my post above this one (adressed to Oli). There I am talking about a game taking multiple phases. Well how many does a game take? On average that is: # phases played / # games played. There is your missing variable. So where does it go? Well, when calculating what the chance is that a player will miss a phase in your game you can do it directly as i did in my original post. Or you can do it like this:
First you calculate the chance the player will miss any given phase. That is: # phases missed / # phases played. (Note that this is the old reliability rating!) Then you multiply this by the number of phases in a game. Well actually, the average number of phases in a game. But we calculated that before, it was: # phases played / # games played. When you multiply these two, the # phases played cancel out. And there goes the variable, it's not needed to calculate the chance we want.

Hope this algebra made any sense ;)

But it's not that easy, because the average number of phases of a game is approx "3*"avg. turns played" of that specific variant (you can get this info on the variant-page)". And not "# phases played / # games played"

Hmm. This made me think.
@Oli: can you calculate the average number of phases per game for all the games ever played? That number could also be used for and estimate of reliability. You could also calculate the average number of phases missed per game or the average chance of a player leaving a game. Interesting stats and we could use those to get comparative reliability ;)

I'm overdoing it. Maths nerd for life yo!

@sqrg: Sorry, no. There is no way to get a stat for all phases played (at the moment). That would require a new database-entry...

@Oli, true. But reliabilty can NOT be per variant. It had crossed my mind earlier. You need to average at some point, because there is not enough data otherwise. It also makes sense: How i behave in other variant does say something about how reliable i will be in a new game, even if i have never played that variant. You must average.

You know, most people have over 90% reliability here anyways. No matter what system ;)

I was wondering what the letters was for. Looking pretty awesome!
The only advantage webdiplomacy has over this is that it has GR system, WHICH wouldn't work here since we have unbalanced maps.
But dammit, it's fun. I enjoy a lot of games in this website and I intend on leaving regular webdiplomacy for this.
Thanks Oli! (whose name is the same as mine)

Know a thing, sqrg? You are much more expert than your profile says.
Little games played, but many interesting posts on the Forums! :)

I like the ideas in your first post and I'm working on them.

although I like the joining games increases you RR, I think it should never allow you to get to 100. If the CD aspect of your RR is calculated now as CD-TO, then you should set a limit so CD-TO>0. This means someone who takes over 100 games can't have a perfect RR despite CDing in 20 games...

I also think NMRs should phase out over time:
Every NMR in the last 10 turns causes a 10% loss in RR
Every NMR in the last 20 turns causes 5% loss in RR
Every NMR in the last 50 turns causes 2% loss in RR
Every NMR in last 100 turns causes a 1% loss in RR
Every NMR in last 200 turns causes a 0.5 loss in RR
and so forth.

So rather then looking at the total number of phases missed over phases played it would be based mostly on recent games, that means if you missed a few turns when you were near but eventually started making every phase, you could eventually get back up to 100%.

Thoughts?

@fasces: That can't be done in an easy way... :-/ The database can't tag each NMR with a timestamp... That would be really hard to do... But with many phases played they fade automatically...

@fasces: The formula i proposed is a ratio in that sense. Once you've missed a phase you can never get back to a full 1 (or 100% if you prefer). But we are rounding the numbers. Suppose you've missed 1 phase, then you would need to finish 200 games to get rating = 1 - 1 / 200 = 0.995 = (rounded) 100%.
And that's fine I would say that player is actually really reliable!

@Guaroz: Thanks I hope I can be useful. I've always loved breaking games apart and making my own versions. Lego's were my favorite toys (the simple bricks and sticks, not the fancy modern stuff) haha. Vdip fits right in.
About my games... I think Oli moved the site and stuff or maybe lost the database in some other way? I haven't been playing webdip for a year almost. Came back for more ;0

@ Oli: Where are TakeOvers expressed? Or is it a hidden stat I can calculate with:
unbalanced CDs - Left = TOs

@sqrg. haha yes I loved Lego's bricks and sticks. And Meccano? Please come back soon. :)

How about A+ for those with really high reliability scores (eg 98% and above)?

"@fasces: That can't be done in an easy way... :-/ The database can't tag each NMR with a timestamp... That would be really hard to do... But with many phases played they fade automatically..."
Fair enough, i have very little experience with coding so have no idea how easy/hard stuff is to impliment.

@Guaroz: If someone takes an open spot (and has a unbalanced CDs) it adds one to the balance.

What is an "unbalanced CD"?

@sqrg,
your formula hits (and hits strongly) on NMRs. I'm fine with that.
Nevertheless, if you elaborate it you'll obtain:

RR = 1 - ( # phases missed / # games finished) + ( # games left / # games finished)

...considering we have a 1st turn CD rule, there's something wrong with the # games left penalization

@Oli. I'm not sure how what you said was different from what I said, but I guess that there's a way to know how many times you Took Over a CDed Country. That was the important thing, thanks.

42 Open games, now. 3 less than 15 hours ago. Just a coincidence or is new RR starting to work?

Nice feature! Now, in game board, your RR Group is reported near your name, close to your points!
Question: what does a question mark mean?

New player (no phases played)

How many phases does the question mark apply to?

only 0.