Yes, but I will remove this old rating soon and replace it with something new to be discussed.... :-) I will make a better CD-takeover stat for a better visibility of this.

And the CDtakeover stats are screwed now anyways and I will need to rebuild them.

Oh, ok

So can't you remove them for now? Make a quick fix before figuring out what exactly we want?

kaner-
I was only making a list of clues that make me believe that majority of members don't give a f*** about integrity of games and, hence, about related issues like RR.
The point of my post was that only people typing here cares about it and I would be surprised they're more than 30 in the whole. Or do you think they're 400?

The wider theme of all my posts in here is that no RR system will work like it's hoped until we don't do something about awareness of players.
So while we work to make up a better system, we should also work to form/inform our members.
I'm wondering why I have to explain this to *you* since you're the one who is making a great work on the wiki to form/inform our members! :)
(talking about "Howto" section)
That's why drano just got a +1 from me just because he wrote "Education is the answer."

_____
PS. Since you talk about extends, of course I believe a player should look for a sitter first of all. But what if he doesn't (or he didn't find any) and votes extend in a gunboat? I'm for the integrity of the game. Majority of players is not.
Just a thing I noticed.

Oli -

First of all thanks for your improvements. As usual, you're fantasic! :)

But there's something I'm missing about how personal stats are working now.
Let's take Grand Admiral Thrawn's (userID=1346) own stats [sorry GAT, nothing personal, just my example] and see:

Game stats:

Won: 21 ( 26% )
Drawn: 17 ( 21% )
Survived: 19 ( 24% )
Defeated: 22 ( 28% )
Resigned: 1 ( 1% )
Total (finished): 80

Well we all know that, for some reasons not important here, he abandoned 7 games.
But only one is in count: Resigned: 1 ( 1% )
The issue with this is that having only 1% in the worst result, better results don't match the truth, because they're higher than real.
The point is that 6 games disappeared from these personal stats. CD worked like an auto-cancel. IMHO this must not happen, because
"when you start a game, it'll ever be somehow on your record (unless it'll get canceled)."
Sorry for the self-quote, but they liked it.
These are the real personal stats:

Won: 21 ( 24% )
Drawn: 17 ( 20% )
Survived: 19 ( 22% )
Defeated: 22 ( 26% )
Abandoned: 7 ( 8% )
Total (finished): 86

My point is that some players intentionally go CD to avoid a bad result - Defeated - on their stats. The game simply disappear: that's why the current system encourages CDs.

If real stats were displayed, players willing to trick the system wouldn't have this option. Because the game would anyway be shown on their record and "Abandoned" is much worse than "Defeated".

This is what I meant. Would it be feasible? Do you see any flaw I don't?

Yes. I will add this stat soon.

TY! :)

I still wonder if this might work to address the re balancing they must play a game which started after the CD to its conclusion without missing a phase. That way each CD takes much longer to work off and everyone could know that they were dedicated to the site.

Darkarus, a point that has been made in this thread--one which is undeniable, I think--is that TOs have nothing to do with CDs.

It is a good idea to encourage TOs, but a different method would be more appropriate. One way to encourage TOs is to change the way they are recorded such that they are not recorded the same way as games I play from the start. I, for example, have no NMRs on my record and no CDs. So when I have occasionally perused the list of joinable games, I simply choose not to join because I can see that the vast majority of them are impossible to win or even draw. While it may be fun to take on the challenge, it sucks that I will get a defeat on my record for doing a service to the community. Moreover, it is an inaccurate reflection of my actual skill as a player. When I step into a role that is doomed from the outset, it is pretty unfair for it to be counted as a defeat the same way as when I start on an even playing field. Another relevant point is that with the new stat changes being discussed, someone else will already have gained a "CD" or an "abandoned" on their record for the country I TO, so the defeat is already recorded before I even take the country over. When I get a defeat because I was doomed, the stat is recorded twice.

Separate stats for TOs would be interesting. For example, if I have a TO win, that sure would say something about my skill as a player.

As for CD forgiveness, there are only two methods I can think of that would work: (1) employ a non-CD filter like the current RR filter, but don't allow a 100% choice. So if you get your non-CD stat above 90 or 95%, you'd qualify for the top filter. (2) Admins grant CD forgiveness on a case-by-case basis, mostly for extenuating RL situations.

The first one is subject to complaints by those who want to choose 100% but can't, and the second one is subject to abuse by members whose grandparents suddenly start dropping like flies.

I think that the Reliability Rating is a bit flawed. Users can just save their moves as hold and not get penalized when they more or less give up.

I've added the idea of Guaroz and combined the stats for "Resigned" and CDs to one "Abandoned" that's in your gamestats.

Ragarding the new RR I think I will just combine noCD and noNMR by using
RR=(noCD + noNMR) / 2
Thats easy and simple.

Now we just need to assign school-grades to the different levels.
A first draft to spawn some discussion:
A=100,00 - 98,00
B=97,99 - 94,00
C=93,99 - 88,00
D=87,99 - 80,00
E=79,99 - 70,00
F= below

Sorry Oli, I've missed what school-grades would be for.
Would they be used just in game-creation, as an option to select reliability of players you want into it?
Or would they imply also other things, IE players with low grades can't join all the games they want?

I think it's to show how bad your rating is, there's no reason to not just leave the new game settings in %, right? Maybe naming them with grades too?

And I would make C, D and E to 85, 75 and 65 respectively.

The grades would just be used for an additional more convenient way to set a game requirement during the newgame creation and it would be displayed behind each users username in game as it's now.
I will not base any penalties on t his rating.

I'm planning to but some penalties on the CDtakeover stat. Maybe something really easy like <0 = you can't create new games and <-3 you can't join new games.

Ah, ok.
In this case I believe that how those grades are drawn is not crucial if you just add a "Custom" option. Likewise phase-lenght or NMR-policy, if you don't like the pre-set options shown, you make your own choosing "Custom" and putting your favourite value in.
Would it be feasible?

That's how it should work.

Oli,

Do you realize that your new formula
RR=(noCD + noNMR) / 2
will actually give a higher result to every player (except those who have 100% or more of course) compared to the current RR ? As a consequence, that formula will only statistically "hide" the more reliable players !

For example (no offense meant to the players) :

- steephie22 : current RR = 15% ; RR with new formula = 85.69%
- Grand Admiral Thrawn : had a RR of ± 40% a few days ago ; RR with new formula = 91.48%
- drano019 : current RR = 99% ; RR with the new formula = 99.64%

I think that it would be a TERRIBLE mistake to give to players higher percentages than what they have now. What we need is to educate players so that they understand that a 93% (it is always that same number that is given as an example) is absolutely terrible in reliability !

With your new formula, Oli, Steephie would get a boost of +70% on his RR and he would get a D mark. But that player has already asked for a system that gives him a C mark ! Obviously, he does not understand that his reliability is simply terrible !

With your new formula, Grand Admiral Thrawn would get a 91.48% RR and would get a C mark. By the way, it seems that Grand Admiral Thrawn, who is at the origin of all this conversation, has managed to make 1 more CD during these last days !
Obviously, he is not a reliable player.

Oli, your new formula is way too generous. Your A-F system is way too generous.
As a matter of fact, that new system will not encourage at all unreliable players to improve their reliability. That new system would only resassure them by showing them that they are more "reliable" than some other players. But the right way to formulate it is in fact that they are just a little less unreliable than the least unreliable players.

As a conclusion, Oli, I would please like to tell you respectuously that your new formulas would be terribly counter-productive !

Once again, what we need is to educate players about reliability.

Personally I think it might be a good idea to 'wipe the slate clean' and start afresh. Given that this is a new system being implemented, start without any prior NMR or CD record and consider Feb 2014 the new start date for the new record keeping.

Thoughts?

@ kaner406 :

Of all the ideas that have been proposed here, I find yours simply the worst !
You are proposing to "erase" the data that we have !

@Retillion: The old RR is no longer correct. The numbers are totally off, as my calculation does not work properly at the moment.

@Retillion: that percentage is the true percentage though. And a D sucks I'd say.

I was suggesting the change because I think the difference between 0% and 70% is too big to be shown by one letter, IMO. If you think a B is terrible with this rating, you can choose to only play with A. That's the problem.

Right now, my reliability rating seems to suggest that I ruin about 85% of the games I join, while that's simply not true. I ruined less than 20%, because of serious shit IRL.

I think I deserve the terrible rating of 85% rather than a number that is by no means a percentage but just a number with a "%" behind it.

And my RR is 15% because I refuse to just join open games, which I consider ruining a game even further unless it's explicitly asked for.

Either take the actual percentage or remove the "%".

@Retillion:
The new RR should be only based on facts. No more "hiding" behind CDtakeovers.
That's exactly why I wanted some more opinions. I always said a 90% mark looks too good, but everybody keeps telling me, no, 90% is not that good, we all know this (That's why 90% "only" gives a "C").

How would you spread the grades based on
A = Excellent, B = Very Good, C = Good, D = Acceptable, F = Fail

I'm with Retillion on this one. Erasing data prior to Feb 2014 seems like an absolutely terrible idea. It goes against everything we argued for about having a permanent record of data. It would take months before we saw any patterns forming again, and you'd have a large period of time where unreliable would be seen as reliable, and reliable people, if they missed a mere 1 turn, could be seen as unreliable. And what about people who play only 1 or two games at a time? Missing one turn could cripple their RR for a VERY long time.

Whatever the grades end up, I'd like an A+ for those who have never missed a phase. We deserve a little recognition.

I agree. Deleting old data is not a good idea.
For about 2 years now I collect the phases played and phases missed.
We can't rely on the CDs and the CDtakeovers from back there, because the way it got added to the database, but I can retrieve some this information from the takeover-message in globalchat and the end-results (resigned).

This is what we have now:
confirmed: phases played, phases missed, CDs (from Spring 2012)
too complicated to programm and not used for the RR calculation: CDtakeovers.

So we start with the data we have and everybody will get CDtakeovers that match their CDs. In the next few days/weeks if your CDtakeover is negative I will add game-restrictions based on the CDtakeover-status that has nothing to do with the RR.

So basically we move the restrictions from the RR to a separate stat and do no longer need the CDtakeovers influence your RR. But the restrictions are still in effect, they are just no longer bound to the RR.

Well done Oli, the question put in those terms is much more understandable.

Though a stat, on how things currently are, would be very helpful to understand what is Excellent, or what is Good, etc.

IE, some could say based on
A = Excellent, B = Very Good, C = Good, D = Acceptable, F = Fail

that your previous scale
A=100,00 - 98,00
B=97,99 - 94,00
C=93,99 - 88,00
D=87,99 - 80,00
E=79,99 - 70,00
F= below

looks fair enough.
But if half out the 420 active players are between 100,00 - 98,00, then grade A would be not "Excellent" but it'd be "Average or better". Too wide, actually! So we'd have to re-draw the ranges.
Can you provide us a stat about how many active players would be atm into each grade?

Another reason why I'm asking for stats is that, from my point of view, who has a new-RR of 98 or better is doing nothing special. He's just playing the way he's expected to. When I saw for the first time how wide the range of grades were in the scale you proposed, I felt them like: A=Normal, B=Acceptable, C=Bad, D=Very Bad, F=Fail. Why should we call excellence what is just normality? Why should we call "good" an unreliable player with a C rating? Wouldn't it be misleading? He would actually think he's doing "good"! Because "good" is the way the System calls his (poor) commitment on games.

I understand it's a matter of feelings and points of view, so a stat would help us get how things really are.

PS: I've added the new code on the server so invalid data is no longer in the way of a good discussion. Again, nothing is final here.
If we could agree on some kind of Grades based on the calculation (noNMR + noCD) / 2
we might be able to apply a factor that moves the current RR-% to a more common system A=90+, B=80+, C=70+, D=60+, F=59 and below.