diatarn -

Thanks for the explanation. I have a decent math background, but statistics I never liked :)

wow, blinded by science!?
Gaussian methods aside, what would be so bad about simply posting ones stats and letting people decide what they will (or apply Gaussian statistical methods to the data set?)
Seriously, not having ratings or even percentages show, simple data only is a bit less implied judgement upon others? 100% implies perfection but use me as an example, yes I am "perfect" so far but only 148 phases, hardly deserving of any praise is it? Instead of such formulas or ratings or percentages or what-have-you, a simple statistic that can be judged by the eye of the beholder is less "judgmental" is it not?

@Tomahaha:
After reading your first lines, I was fearing you'd start talking about Bayes' theorem ;-)

I agree with you that the best judgement is reached by looking at the "raw" data (and that's what I'll do personally). However, I also think that Oli has a point in looking for a synthetic "catchy" rating that might help with the most "unsophisticated" (?) players, and that there is no coexistence problem with the two approaches.

Perfect over 148 phases is someone that most of us would be happy to play with. I don't know how much more "perfect" or "praise-worthy" someone needs to be. Also, even if someone hasn't played a ton of phases, the "perfect" label might very well encourage them not to miss.

Maybe add a statistic that tells us how long it has been since a person missed a phase? Doesn't seem terribly necessary to me but, while we are discussing options...

Yes, Bayes, they make an excellent teak cleaner available at a Home Depot near you...

The exact stats are easy visible on everyones profile-page.

The "catchy" rating is just for easier recognizion behind the players name because posting all the stats stuff (including the Win/Loss/Drawn ratio) behind the DPoints would make the names-line quite difficult to read.

And it's not very mathematican. It uses the stats directly from the NoCD and noNMR, but stretches the range from 100%-80% to 100% to 50%, because nearly all players are in the 100%-95% rage. There is no point in using a 0%-100% rating and than use only the range from 95% to 100%.

Also we used a "Rookie"-rating for <20 phases played. Maybe we should apply a "Rookie"-rating for <50 phases played and <5 games finished.

Me sitting in that "rookie" category, I have no problem!
I would say 3 games would be enough time to determine reliability but 300+ phases would be the min! (besides, how many phases would it take to complete say 3 games? far more than only 50) or possibly only number of phases? If say one LONG game lasted say 500 phases compared to say 3 games that lasted what, maybe 300 phases? Isn't that long game enough for determining reliability? But I'm cool with whatever you come up with, and I am part of that group!

Ooooh, another idea...
I have heard a few people who do not like this new idea. While I like it and many do, I also understand the reason people want pristine statistics behind them. While we do not want to erase this history, is it POSSIBLE to maybe add a section where a player can post a brief explanation to their stats? When I click on someones name and it shows stats. Maybe they have a CD or two, is it possible for them to post something to help others understand the reason for the CD. this way, to those who think it matters, they ignore the reasons, to those who want a reason, they can read why and maybe this also can allow us to see the CD happened a long time ago? It seems like a best of both worlds solution, but only if this is possible that is!!!

Maybe we can have check boxes with all the usual excuses: "- I was busy at school. - I was busy at work. - My wife was sick. - I was sick. - I was traveling with no internet access. - The dog ate my keyboard."

oh, I agree 100% and to me the stats speak for themselves not to mention anyone can simply lie about the reasons. But it would help those who feel they lost something with this new system AND, anyone who actually bothered to post a reason, in my eyes tells me they at least care! (even if it were lies, they care enough to post a reason!)

Hey my dog literary can chew a keyboard if you serve it to him!

I agree that posting a reason next to your stats is silly. I surely wouldn't give it any credence without proof. My only problem is the time in which cd's affect your reliabily, and whether or not you made a consciencious attempt to have it resolved. In these cases, I think it should not go against your reliability ranking, since making an attempt to resolve it shows that you are reliable. For instance, my reliability ranking is 94 (acceptable) and I know that 2 cd's is really affecting it. These cd's were almost 2 years ago, and I voted "pause" in both games and requested an emergency sitter in the forum, as well as contacting the mods.

http://www.vdiplomacy.com/forum.php?threadID=31311

Since those games, I've taken over at least 10 cd'd nations, and missed only a couple turns. Even in most of those very few cases, I would try to get an extend if I could. When I look at my reliability ranking, it seems very low compared to the majority of players, but I don't consider myself an unreliable player. I remain an active and communicable player in each game, and if I am busy I communicate those reasons and situations, so I'm not sure whether this new system adequately represents a players' reliability. "Acceptable" seems a little harsh, especially if anyone has ever bought a book from Amazon in acceptable condition, we all know what that means.

The "Acceptable" was a wording error.
I used A = "Excellent", B="Good", C="Acceptable" D="Bad", E="Very Bad" and F="Fail" and later introduced introduced (on Mapus request) an A+rating, so the naming from A shifted to A+, from B to A and so on.
Should be fixed now. BTW: As always is a grade right at the border to the next naming a bit harsh, but it you should be in another grade rather soon...

To keep everyone updated on the most recent changes.
We have now:

noCD = 1-(CDs/games played)
noNMR = 1-(phases missed/phases played)

If we combine these 2 (noCD+noNMR)/2 We get a rating that's very top heavy and most players have a rating between 100 and 90. That's why we apply a (^3) to this percentage to stretch the top 100%-80% to 100-50.

Result:
(CD+NMR/2) = 100%-99,6 => R100-R99 (Excellent)
(CD+NMR/2) = 99,5%-98% => R98-R95 (Very Good)
(CD+NMR/2) = 97%-94% => R94-R85 (Good)
(CD+NMR/2) = 93%-88% => R84-R70 (Acceptable)
(CD+NMR/2) = 87%-80% => R69-R50 (Bad)
(CD+NMR/2) = 79% => R49 (Horrendous)

The wording is just added in the profiles-page, so people can get an overview how to interpret the numbers without asking all the time on the forums.

You can set game-limitations based on the RR, the noCD and the noNMR.
As long as you have not completet at least 2 games and 100 phases you are considered a Rookie

Some numbers: http://www.vdiplomacy.com/hor.php

Oli, wouldn't it be nice to add one more grade ? Indeed, I propose that the players who have both missed 0 Phase and 0 CD get the grade "Perfect" instead of "Excellent".

Oli -

1. Current grade scale.
I'm aware you said it's just a temporary scale you'll keep "for some time" to let you "get some real usability-results."
Anyway you can imagine my thoughts on how it is built.
Keeping in mind that the purpose of such scale should be making quickly aware people of their reliability degree (= if and how they use to miss too many phases), I can't approve you didn't abandon school wording. It's misleading, it's not a skill rating! And it's dangerous, you could find cases (IE, userID=3159, 13 missed phases in 14 games played) in which a player who statistically NMRs once in almost every game he plays gets called "Very Good". Are we encouraging NMRing? This wording - sorry Oli - sucks.
Also, keeping in mind the players are concentrated on the top percentiles, my thought on how the the grades are spread is that while the first 2 grades might be ok, the other look really too wide. I mean that R85 is much different than R94. And R70 has nothing to do with R84. They can't be in the same grade, this is what "too wide" means.

2. Stats
Thank you very much, Oli, this "hor" is very useful! :)
Though I wonder why only 251 players are in it... how the sample was built? I expected about 400 active players.
As we predicted, concentration is on the top. Players with R98 or better are 44% and a clear majority (60%) has at least R95.
On the opposite side, Players with R80 or worse are only 17%. This percentage drops to basically nothing (4%) for players with R60 or worse.

Most of all, we now know how many possible joiners each degree has when you're creating games with the Custom option.

3. Scale I propose
It's what follows. You won't believe it, but I made it the least harsh I could think, hoping most of you can like it this way. Grade's wideness and name are consequential to my thoughts about the meaning of each single degree. Last column just shows possible "population" of each grade basing on Oli's stat.
Grade..............R-Range...........Numerosity
Reliable..........R100-R99...............34%
Normal............R98-R97.................16%
Acceptable.....R96-R94.................12%............positive grades = 62%
Inadequate.....R93-R90..................10%
Bad.................R89-R84.....................8%
Very Bad........R83-R75.....................6%
Horrendous....R74-R65.....................5%
Please talk to Mods..R64.or.less.....9%
This is what I mean with a scale that calls normal what is normal and it calls bad what is bad. The player I mentioned before would be Acceptable, not really "very good", and if he doesn't improve his behavior and goes on missing a phase in almost every game, he'll soon get to negative grades. This scale talks to players and tell them the truth, IMHO.
This is because I kept in mind that its purpose is to guide players who lack in reliability - and not to coddle and satisfy the ego ("Excellent"? What?) of players who just do what's expected.
This system final purpose should be to keep integrity of games improving the quality of play here, right?
So let's focus on players who need to improve their commitment. System's goal is to tell them they should put more effort in it. Most of players can do better than now, so let's tell them so. Let's tell them the truth.

4. Rookies.
I followed this discussion and, at first sight, I was for 100 phases / 3 games.
But then I found newcomers like userID=3777. Worse than the previous one, he has missed 9 phases in 7 finished games. If the system calls "Very good" a player who statistically does NMR more than once in every game he plays, and even my scale would call him "Normal" (R97), then our formula has some problem with players who haven't finished many games yet. Their sample is still too small.
Solution other than changing the formula could be making wider the Rookie-range.
Even 5 games absolutely don't look enough. Maybe 10?
If 10 games look a too long time to keep a newcomer in the Rookie status, perhaps split it in 2. Call them anyway, say Rookie-Junior up to 5 games, with all restrictions (he can't create new games and he can play up to 3 games per time) and Rookie-Senior from 6 to 10 games finished and with smaller restrictions (say he can play 6 games, 2 of which created by himself).
Anyway, current formula has problems with newcomers and a solution must be found.

Guaroz,
regarding your last point, I don't really see why you say that the formula has problems with players who finished few games.
I mean, in the cases that you show the problem is not the small sample: in one case you are talking of 13 missed phases out of 552 and 0 CDs out of 14 games; in the second, of 9 missed phases out of 460, and 0 CDs out of 7 games.
The number of games is relatively small, but the number of phases is pretty decent; furthermore, things would not change if you were talking of 130 missed phases out of 5520, and of 0 (or even 1-2) CDs in 140 games: you'd still have ~ 1 NMR/game and ~ 96 reliability.

I might agree that the formula is too "accomodating" (btw, one easy way to fix it is to use an exponent higher than 3 - see what I was saying yesterday to drano), but sample size has little to do with it.

"not to coddle and satisfy the ego ("Excellent"? What?) of players who just do what's expected." But then "normal" in your scale is people who have missed a lot of phases and abandoned games.

I don't see why there has to be a normal bell curve distribution for this. If everyone does as expected, then it will indeed be top-heavy. It would be awesome if everyone on the site was in the Excellent category.

I like what Oli has done, I like the wording, and most of all I very much appreciate all of the time he has taken to do this. To tell him his work "sucks" isn't very nice.

Well diatarn, you could be right. I haven't analyzed more than 10-12 profiles and I noticed that the problem was more evident in newcomers. Then I believed it was so.
It could be instead a sheer coincidence and hence the sample size has nothing to do with it, yes. You look much more good at maths than me, I'm sure you'll find it right.

The rest... it's a matter of points of view. I understand that a great number of phases played mean a rather long time.
But if your "missed phases" are equal to (or greater than) the games you finished, then my idea is that you are likely to ruin every game you join. IMHO, this is not acceptable, whether you've played millions phases or not.

"Join less games. Join slower games. Set an alarm for deadlines on your cell. Do anything. You can't go on like this NMRing in basically every game!"
This is what the System should suggest to those players. They're everything but 'Very good'. They're not even 'Normal', if they NMR in every game they're in!

As for CDs, they're just a different way to call 2 consecutive NMRs. No NMRs = no CDs.
A middle power NMRing a crucial turn is usually much more devastating for the game than a 1-unit power CDing in the endgame. IMHO, we should not give to CDs more importance than they have: count them just as 2 NMRs.

NMR is the enemy. No NMRs, no problems.

Rookie until 10 games are complete?
If you play one game at a time, that could take a few years to get out of rookie status. Seems a bit harsh and also tends to ignore this type of player who may very well be reliable. Me for example, I have no problem with being a rookie status. Yes I have played the game for over 30 years but I have no problem with such status until I prove myself. But if it takes a few years of casual gaming then that will no doubt turn some off and it does not reflect ones real reliability very effectively ...shutting out many players for so long a period of time, it almost feels like an exclusive club and rewards those who play possibly TOO often. Getting a gauge of what a player is does not require stupid high numbers. What's wrong with say 300 phases played? That should give you a good sampling of ones reliability without making them feeling unwanted or excluded for so long a time.

Can we add an excuse box for I was drunk, celebrating my sports team's championship. My long missed phase was by 15 minutes while I got carried away watching the the Red Sox win the World Series. It was completely my fault, and it might have cost me a shot at a 209SC solo on the WW4 map, so I feel the pain more than the black spot on my previously perfect attendance card.

The best way to win the respect of your fellow players is to own your mistakes, apologize for them, and seek to prevent them from reoccurring. ((% of the "real life" excuses we get here are just hogwash and lies because no one is willing to be responsible for their actions any more. Everyone's a victim.

Circling back to the idea of an explanation. Players do have a personalized section right below their name on the player profiles. If you need to make excuses or defend your reliability reputation, there is a place provided for it. Post your excuses for sloth there, and see if anyone cares.

@Guaroz
perhaps you can get your way by using something like
modified noNMR = 1 - [(NMR + A*CD)/PHASES]
[where NMR is the number of NMRs, CD is the number of CDs, PHASES is the number of phases, and A is an arbitrary constant] as the base which is then raised to the Nth power (currently, N=3).
This would amount to consider CDs as "super NMRs": if you have CDs in your record, your modified noNMR drops; but you don't get the "diluting" effects of having 0 CDs.
so, the two cases you were showing would play out as
- 13 NMRs in 552 phases with 0 CDs --> modified noNMR = 0.9764, modified R-rating = 93.1 (with the current N=3) rather than ~ 96
- 9 NMRs in 460 phases with 0 CDs --> modified noNMR = 0.9804, modified R-rating = 94.2 (with N=3) rather than ~97
If you believe the change is too small, you can play with N: N=5 moves the two ratings to 88.8 and 90.6, N=10 moves them to 78.8 and 82.1.

There remains to find an appropriate value for A: instinctively, I'd say A~3, so that a CD would cost you ~ 5 NMRs (the 2 actual consecutive missed phases - which are recorded in the number of NMRs - plus 3 "penalty" phases). But it should be looked in detail.

ps: At present, I am not advocating a change in the current formula, I'm just looking around to see if it is possible to get improvements that make it more acceptable to Guaroz (and possibly some other people).

Ruffie, we agree more than you think! I only mentioned this for those who want to plead for forgiveness and you said it yourself, "apologize for your mistakes" and that is what I was suggesting, a way to apologize and move on. But you mention a space is added already so yes, they can take advantage of that if they feel such a need!

something I am confused by...
I looked at the average number of phases per game based on some who posted on this page and it makes no sense to me,
Drano has 40 games played, 1103 phases = 27 per game
Raro has 27 games played , 1212 phases = 27 per game
Diatran has 107 games played and 2886 phases = 45 per game
Yet I am only in one game, granted a large one but we are in at 10-11 game years only, maybe a standard game length? (even somewhat short?) yet I show 148 phases.

Why the drastic number differences?
Drano and Raro = 27, but Diatran is almost twice that and I'm going to average, what 300??? Something aint right in what i see...

Maybe when you had the mods re-do your turns you got double phase credit.

Sorry, couldn't resist.

Tomahaha, the numbers you quote about me are quite wrong (assuming Diatran is me :-) ): on my profile I see I finished 27 games (but only 22 count for the RR rating) and 1212 phases (I guess in these 22 games).
Then, it's 55.1 phases/game
I went to check also other users:
Drano: 39 games / 1103 phases --> 28.3 phases/game
Raro: 79 games, 2886 phases --> 36.5 phases/game
you: 0 games, 148 phases --> infinite phases/game
Mapu: 48 games, 2632 phases --> 54.8 phases/game
Guaroz: 128 games, 4815 phases --> 37.6 phases/game
RUFFHAUS: 48 games, 3754 phases --> 78.2 phases/game
all of the above: 364 games, 16549 phases --> 45.5 phases/game

I guess that;
- your 0 games is because you haven't yet finished a game (though I must say that if you are in a single game, then you should be in its 30th year or so, not in the 11th: there are a maximum of 5 phases/year: weird)
- the difference in phases/game between me/mapu/RUFFHAUS and drano/raro/Guaroz is really large; and that's strange, too.

I understand some will get killed off early in some games, but all I checked had at least several wins so that should balance things out. And my stats... it's only year 10-11 count it as one game and the numbers are still strange.

It is absolutely logical that some players have an average Phases/Games ratio much higher than some other players.

Indeed, you will play more Phases in one specific Game if :

- the Variant played takes in average more Phases than some other Variant. For example, a WWIV Game has in average 31.94 Phases while a Classic Game has in average 17.16 Phases.

- the player is not eliminated. Obvioulsy, succesful players will have a higher Phases/Games ratio than players who gets more often eliminated.

- some players enjoy to play the game until it is really finished (i.e. until a very narrow Draw is achieved) while some other players are happy with a quick and large Draw.

- it may be wrong but it looks like Full-Press Games have in average more Phases than Gunboat Games.

And so, all those parameters combined together have the logical consequence that some players have a much higher Phases/Games ratio than some other players.


And we have here a perfect example of people who are surprised by a very simple statistics which is just a simple ratio : x/y.
It may *seem* very nice or very intelligent to calculate many things but it is pointless when people don't understand the numbers that are in front of them.

Oh, I shouldn't have written "pointless", I should have written "counter-productive" because what we want is that players, especially new players or possibly very young ones, understand that missing 5% of their Phase is a catastrophe !
AGAIN, what we need is to EDUCATE people by telling them that missing x% of their Phases is not considered as being reliable and that we expect players to be reliable.

Oh, I shouldn't have written "counter-productive", I should have written "very productive" ! ...But only for the ones who doctor and truly understand the statistics that they are creating and presenting to the world. That is called PROPAGANDA !
But we have no such thing in this conversation : what we have here is, in my opinion, people who love to calculate things but who fail to understand that the results of their calculations will be misinterpreted by most people.


One last thing : you should be very cautious about the word "normal" because most people wrongly understand that "normal" means "good".