All due respect noted and appreciated, Mapu. You are a gentleman and a scholar, and a scoundrel worth matching swords and wits with any day.

Despite the fact that a ratings system can not be perfect, it's certainly better than the old points system and a step in the right direction. For the most part, the weighting already in the formula addresses some major concerns, and many of the points raised were considered when Oli first implemented the system, which is not to say that the system can't be improved, but that we're going to have to live with some apples to oranges weighting and other items are likely to be left in so as not to exclude too many game types that are regularly played. It's never going to be entirely precise, but I think it's a huge improvement from the old points system. Good job Oli!

Like every system this one has it's flaws too.
But we had quite a lot of players playing 100-point-games with their friends. Winner get all the points, looser get back to 100. Even if unintentional it caused quite a lot problems.
It's much more difficult to game this new system.

The biggest flaw of this system is the fact, that once you reached your final score (compared to all other people playing here) it will stay there.
Also I'm not 100% sure if (and when) we get rid of the DPoints, but I just thought putting our better representation of a players "strength" behind his name instead of the not very accurate Dpoints makes more sense.

Considering some questions in this thread:
(1) Are gunboat and regular communication games weighted the same? I
Yes. They use different skills, but I think nobody would say gunboat need fewer skills.

(2) Why aren't games ruined by metagaming/cheating removed from the stats and formula?
I'm working on this, but if your score does not reflect your actual skill, it will even out after a few games. If you lost too many points because of a cheater you will gain more points if you win in the following games (because of the bigger difference to your opponents), or you will loose fewer points if you loose.

(3) + (4) Different RR-requirements and maps...
Again, this will even out just over a few games, as your result is totally dependent on the games you played before - see (2)

(5) Why different Victory conditions are not kept in the count?
They are. A 20% win-condition change the rating not so much as a 51% win condition

(6) Does to rate games that lasted only a few turns make any sense?
Not so much, but see (2) again. It's hard to draw a line about what's acceptable and what not.

(7) Should you have a minimum number of games to be rated? I mean, in the top 100 we have players that play less than 15 games per year.
I'm all for this. As a new player the rating changes quite a bit in each direction till it settles on a number. I think 5-10 games should be enough, this is on the todo. Technically that means you get a rating, but you are not in the HoF, and the points are not shown on your name. In more official systems new users are considered a rating of 1000 for some (usually 3-5) games, and after these the average is calculated. I discarded this idea for too complex, but maybe it's worth a try.

@Devonian:
You got the numbers quite right. An explanation of the match value mV:
- In a WTA this is always 1.
- In PPSC this is the difference of the SCs%.
Example: 18SC vs 16SC in a PPSC is 53% vs 47% => mV=6% (or numeric 0.06)
So winning 18SC vs 16SC will not gain many points in a PPSC setting (like the same example with DPoints)
If you play in a 7 player game the winner will not get many points from the runner-up (as they have similar SCs, but much more from the defeated players or players with a low SC count. You can say he earned the points from the defeated 100%, and from players with SCs proportional to how much more SCs he had.

@diatarn: just questions I wonder, you know. I'll sleep like a log tonight, even without answers! :)

@raapers. Mostly agreeing with you, but from a different perspective. If you ask me my opinion on your previous post, I can't say I love the switch to v-points and I can't say I hate it. None of the systems is accurate enough, so the switch doesn't make much difference to me.
Instead I liked your questions, which pointed out a couple of real problems with this - or should I say "any" so far - rating system (and yes, your intention to be productive was clear and commendable).
So I tried to add some questions to your list that IMHO make sense in order to undeline that v-system isn't accurate enough and probably no system could ever be here, on a site where games to be compared are incomparable.
IMHO you shouldn't expect a meaningful result from a rating system here, hence no rating system should be taken too seriously (looking at diatarn_iv), but it's fun trying to make one and it's fun talking about it. Thanks Oli for providing us some fun.

BTW, another very good point was made by Ruffhaus, about the "head hunting". Perhaps, can I re-elaborate it as the following?:
(8) V-system makes harder for Top Players to do well in non-Anon games, because the system rewards defeating those players mostly and hence they may get targeted. Which is *not* fun.
The reason look obvious to me: it's the essence of that system to pretend that a multiplayer game is just a series of many different and unrelated 1v1 games. It's its biggest pro and its biggest flaw at the same time.
So I wonder: consider a Modern. If you want to account for strength of opponents, isn't there a way to weigh differently a game joined by champions (hard to win) vs. a game joined by noobs, without assuming they're just playing 45 unrelated 1v1 games? This assumption has a big impact on some games!

Thanks Oli,

Can you add that explanation to the description when it is added to the site. It is much more clear than what is written in the wiki regarding the mV.

Also, in my example, the result would have been 1.8375. But, I notice all the totals are listed in whole numbers.

Is the exact fraction added to our score but rounded for display, or is the adjustment rounded at the end of the game, and the rounded number added to our totals?


To all others,
Yes, there are drawbacks, but it seems like a very good solution. It was the result of probably 2 years of discussion, feedback, heated debate, and compromise. As a result, no one got what they really wanted, but that's solid evidence that a fair compromise was the result.

@Oli. Thanks for your kind replies, you told me things I didn't know. The (5) was fantastic.
Though the only thing I'm not much convinced about is the "it will even out after a few games". Which has its importance, I'm afraid.
First because it implies that the System isn't reliable itself; there will be never a moment in which you can say: "the rankings are now up-to date".
Second because nobody knows how long will last a series of "bad-rated" games, so I'm not sure whether this "soon or later your rating will auto-fix" can be an explanation for any flaw.
For example, a good and very reliable player, who uses to play games with a strict NMR-Policy because he doesn't like corrupted games, is disadvantaged _forever_ towards an equally skilled and equally reliable player who uses to make easier points in games where he can take advantage of NMRs.
Anyway my point is: The more are the flaws, the less the system is able and fast to auto-fix his own errors.
Cheaters (2), games with NMRs (3), unbalance of the maps (4), games ended too soon, think how many games have been drawn instead of canceled: should a KW901 ended in a 10-way draw be rated?(6), idle players (7) and non-Anon games basically unplayable(8), look too many issues for the auto-fix theory to work good & fast enough, because too many games get badly rated.

To not mention we're weighting "apples to oranges" (1), but that's unfixable and I agree we have to live with it. I can even agree it's less inaccurate than the D-system, so... thanks! :)

I really like that the site has changed over to this system. But then I never did care about d-points, so can't speak to what has been lost (or de-emphasized).

Is there any consideration of adding a decay rate to peoples' ratings? This would discourage "sitting" on a high rating, and encourage people to keep playing; the ratings would also tend to reflect the current active players.

5-10 games minimum to be rated is a good idea. An old but similar Diplomacy rating system (run by the Vermont Group I think, or something like that?) would consider your rating official once you had finished 7 rated games.

How does the Ghost Rankings deal with the problem of a player sitting on their 'rank' ? Is their a type of point decay in that system?

Players are taken off the ranking boards if they have not completed a game in 3 months. There is no penalty for rejoining. There is natural inflation I believe however.

This system has the same natural inflation as the ghost rating system, even though both are zero sum, strictly speaking.

I like the idea of hiding players from the top 100 if they are inactive and reactivating them when they return.

There're very good in the comments: why would all games (except 1v1) rank the same: it's not the same defeating 11 players than defeating 3.

IMO the following measures should be taken:

1. All games are anonymous (prevents metagaming to some extent)
2. NMRs → replacement rule is default
3. You begin (new account) with zero points and can't lose points (unless a penalty for DCing/NMRing is implemented, but never for losing).
4. Winner gets as many points as factions are in the game: 7 in standard, 3 in Hundred, etc. Or maybe better make it x10 for easier fraction count: 70 in standard, etc.
5. Drawers get equal fraction of winning points divided times 2 (so drawing has a serious penalty). For example: a 3-way draw in standard gives 35 / 3 = 11 (rounded down) for each drawing player.
6. PPSC survivors get the fraction of winning points without modifications (you joined a PPSC knowingly and some maps are so imbalanced that they require PPSC rules to compensate)
7. Scores get "rusty": on January 1 2015 (for example) the fraction of your score earned before Jan 1 2014 is divided by 2. Cummulatively it means that your last year's (and present year's) score is worth 100%, your score from 2 years ago only 50%, from 3 years ago 25%, etc. That way you can't live off past glories if you're not playing often (your score from 2010 is worth only 6%, for instance).

Where it reads points it should read P.O.I.N.T.S.

^^ Oh, man, scrap that last thing. Why on earth does it translate "points" for "D"?

Louis, they don't all rank the same. In a 12 player game, winning against 11 players equals 11 wins, each getting v-points for the win. in a 5 player game, winning against 4 players equals 4 wins.

In addition to that, smaller games are given less value per win, 1v1 games aren't actually singled out, it is gradual for all game sizes. (See the formula for gV.) The gV formula for winning a 1v1 game is like winning +-1/8 of a win against a single opponent in a larger game.

Two examples that demonstrate the difference between playing WTA & PPSC in the new ranking system:

take the game 'silent fog':
http://vdiplomacy.com/hof.php?gameID=20976
I won this one and because it was a WTA game -> everyone is considered defeated and looses rank to me.

in this game 'team USA':
vdiplomacy.com/hof.php?gameID=20857
I won the game, but actually got less rank than lazynomad who survived. This was a PPSC game. lazynomad got a large boost in his rank because he was expected through his position in the ranking at the time to have been defeated in this game.

This is terrible - who's idea was this? Does anyone even believe in play-testing?

Obvious troll is obvious.

Sorry if this has been asked before, but would games in which you have taken over a CD power be rated different than those where you were present from the beginning?

I quite enjoy taking over CD's and the discount on the point bet made it a fun gamble to see if you could salvage the position without particularly big risk to your point total.

Whereas, if there is no discrimination, it creates a huge disincentive to taking over CD's.

I guess this also brushes up against the the issue with unbalanced maps, so a solution to the latter would fix the former as a bonus.

I don't recall the details, but something concerning CDs was taken into account in the V-points algorithm.

Maybe Oli or others can enlighten you LogicCure.

However, I am happy to see this "V-thing" finally coming true after two years of testing.

Thanks a lot Oli, this rating system is way superior to the old one.

If someone CDs he "lost" to all other players. So his rating will get a deep hit.
(some with DPoints, we do not refund DPoints here is you CD, even if you had below 100)

After congrats time, it's proposals time:

I remember I had tested this with two past SRGs and it turned out work well:
I am certain that using V-points instead of D-points in the game creation page settings would be another step forward for the website... It shall be a more reliable way to choose the level of competition. Ultimately it's just a matter of morphing some concepts a bit: we may not choose to "bet" some points anymore, but we may filter (or not filter, obviously an option!) the level of opposition.

In practice, I suggest this for the game creation page:
Instead of choosing a bet that does not do its job at all ( any bet below 101 means nothing due to D-points refund and games with bet higher than 100 basically never existed on V-dip ), we may use a "level of competition" option.
Basically 4 combinations of V-points lower and upper bounds (customizable numbers):

1) No bounds:
game open to everyone
2) Lower bound on:
no beginners
3) Upper bound on:
no experts
4) Lower and upper bounds on:
no beginners nor experts depending on the numbers chosen

+1 Decima Legio

Yes, I've seen that being the cause of a CD is counted as a loss, but what I'm asking is if there is some kind of consideration in the case that a new player takes over a CD power. If he is rated as harshly for his loss as he would be if he were present the entire time.

It seems to me that the takeover player shouldn't be rated at all. But, I don't think that's how it really happens.

The takeover is rated depending on the position compared to the starting position.
So if you take a position that is worth only 20% of a starting position each game does count much less.
But on the other hand if you take a winning position each match does not count that much too, because you did not earn most of your win on your own.

So if someone takes over a position with 5 SCs (in a larger game) and turns it into 7 or 8 SCs but has no chance in hell of being able to win, has he screwed himself?

For instance, my takeover in gameID=21663 has seen the position improve significantly since I took it over but there was never any way I could have won this despite my seriously improving the position from being on the verge of elimination (5 SCs caught between two powerhouses to 8 SCs and standing strong). I'm no where near the topped rank player left in the game, but have grown the nation by 60%.

Rinascimento isn't rated.