Aisde from the pressing concerns of where Retillion appears in the standings, which must of course be the number one concern for everyone, given his recent investment in daily 1v1 games there continues to be a systematic flaw in the way HoF points are being awarded.

Oliver you said that the focus would be on Win>Draw>Suvive, but the formula in place favors other factors above these. I want to say up front that I am absolutely a proponent of weighting the points awarded to reflect strength of opponent and size of the board. However, I feel that the present methodology focuses too much on these factors, placing them above the actual game results. This issue is so out of scale that it has some players receiving more points for surrendering a solo in one game than they receive for winning another game. Surrendering a solo is a loss for everyone except the player that solos. Thus the phrase solo, as in solo victory, as in one winner.

In no case should a player that loses by solo recieceive more points than any draw or victory results. The "survive" result is a euphemistic phrase for losing a game by solo rather than elimination. It happens to all of us, but it should be avoided. This is the very essence of the game of Diplomacy. The objective of the game is to win, and failing that to prevent anyone else from doing so. I'm not advocating some excessive punishment for it, but it should never be rewarded, and certainly not more than another player's draw or victory result in another game. Losing is losing.

Losing by solo is like losing the Super Bowl, or the FA Cup. There's no reward or glory for it. It's just failure.

Ruffhaus,

So this issue has been covered already and works correctly for WTA.

But for PPSC as it has been played on this site, solos can be more valuable than a draw (wrongly though that may be from a WTA perspective). So, if the calculation is to be adjusted, it should be adjusted from a universal reset sending everyone back to 1000 and adjusted correctly going forward (ignoring past results), since this is not how the games were played. (I'm not sure I'm in favor of a universal reset anyways..)



I floated a proposal at one point in this thread for scoring PPSC (this would be for an alternate ranking system rather than the current system, or the current system after a reset for the reasons outlined above) where the solo winner vs all survivors is weighted 75% to 25%. Thus in a 2 player draw each gets 1/2 while in a solo the survivor only gets 1/4 of the total. For a 3 player draw, each gets 33% while a strong second place survive gets a majority of 25% of the score, or less than the draw.
Unfortunately this can break down once you get to a 5 player draw, where if the strong second isn't willing to oust the remaining 3 players he may have close to 25% by letting the other strong player solo, rather than 20%.

So instead I would amend this earlier proposal to (where n is the number of players at game's end):
Draw = 100/n%
Win = 100*(1-1/2*n)%
Survive = 100/(2*n)%

For a two player draw: W = 75%, D = 50%, S = 25%
For three players: W = 83.3%, D = 33%, S = 16.7%
For four players: W = 87.5%, D = 25%, S = 12.5%
and so on and so forth such that a survive is only worth at most half of what a draw would be.

However once again I stress this would need to be implemented either after a universal reset, or for a brand new ranking system alongside the current one.

Thus while in principle I agree with you that the current system is flawed, it has been flawed from the beginning, and can only be fixed with a complete reset or having a separate ranking system (or avoiding playing PPSC all together or at least look at PPSC or WTA rankings separately once Oli gets those coded up..)

Just my thoughts on the matter.

+1 RUFFHAUS
+1 Leif

Okay, with that explanation of the logic behind the scoring system, why on earth would anyone ever play in a PPSC game? And further, why is the default game setting PPSC?

I guess what I'm driving at is what is the basis of rewarding this kind of play? I could see the value in rewarding points per supply center within draws, but a solo in any circumstances is a just that, a solo victory. I don't understand the logic behind awarding people points for failing. Can someone please explain to me why this was ever the case? What purpose does it serve to award points to people who participate in ultimate failure? This is simply encouraging players to learn to play the game improperly. Why on earth would anyone want this?

Ruffhaus,
I agree that a solo is a solo and should grab a vast majority of the points, but what about in large maps, that many of us have played in where the solo victory is 50 like in IV. Under your formula, a country with 50 solos and the country with 47 gets zero. Seems like a lot of hard work to gain nothing on a poorly designed solo requirement. I would agree with you if all solos were 50% + 1, which will drive more draws and then a true solo should be awarded with the remaining powers getting 0 if they survive, not negative, that doesn't make sense either.

I have notices also that a win in 1v1 game gives you 2 .
While a draw iin a ww4 amoong 7-9 good players only awards 6!

oh my, now that I look the HOF again numbers are different
my SOLO in a Viking game awards me 2 !
and my other solo in Tready of Vurden awards me 0 (!) points yes 0

Looking at your HOF games, I think I see the problem...

A draw awards 100% RR against any defeated players, so you will always gain... However, a solo in a PPSC game appears to give you an RR based on the amount of SCs you got, which means that a draw is quite often better than a solo in games.

I know I'm going to slip down the ratings a lot if this is fixed (because I draw a lot :D) but surely the soloer should score 100% RR against all other players?

@cypeg. IMHO, about WWIVs, it's a much bigger flaw that a win like this gameID=1623 gives you 30 V-points. Turkey & C-Asia NMRed several times at the beginning and strong players (except Russia) started ALL faraway from Iran, being basically nobody in Asia and North-Africa and being concentrated mostly in the Americas, fighting each other and scoring poorly. Even my sister's little daughter would have won this game with Iran, while playing with her dolls (sorry gantz, I know you're an excellent player and I don't need a Rating System to know it. What I mean is that if you drew up a north-american country, say Illinois or Mexico, Russia would have won this game, and scoring a positive number of V-points would have been much harder for you).
That said, I'm not even sure why they included games with less than 6 players...

@Cap'n. My answer for your question is in my previous post on this thread (6 Mar 2013).

perhaps a solo is a solo no matter if it is a PPRC or WTA! ?

Guaroz - That's not really the point I'm trying to make... I agree completely that PPSC should be treated as PPSC, but putting the soloer's RR to 100% against everyone else doesn't change that. It just makes sure you can't lose points for a solo.

The system of rating survivors against one another according to the current system wouldn't need changing, and that's the essence of PPSC. Survivors would get more for having more SCs than other survivors or those who were defeated, but the soloer wouldn't be penalised points for a lower-rated player controlling a portion of SCs at the end of the game.

why dont we give a PPSC solo a bonus? after all fiat money is what runs the monetary system.
And then perhaps I can make loans?

Cap'n - sorry - I can't understand. "putting the soloer's RR to 100% against everyone else" in a PPSC would change the goals each players has, transforming the game into a hybrid much closer to a WTA than to a PPSC.
My concern is not a flawed useless Rating System, in which lucky Iran facing basically no strong player took 30 times the V-points Texas took having to face many strong players. Wasn't it the reason why this thread was opened? A RS keeping in the count WHO you are facing? Mission failed. Completely.

My concern is not that this would get even worse with your proposal.

My concern is people starting playing PPSC games as if they were WTAs.

I don't care RSs. I only care going on having fun playing my games. If they're WTA, I expect my opponents playing them like WTA. If they're PPSC I expect they play them like PPSC.
Having people in my games playing differently because of a freaky RS, would ruin my fun a lot.

@cypeg:
You did gain only 2 in that Viking game, because your rating is much higher than the rating of your opponents. With this system you can't gain many points by playing much lower rated players. You need to play opponents of equal ratings (or better) to gain more points.
But I will check how this compares to a draw.

If you see the game as one and not all these sub-ratings you will realize that this game gave cypeg still a positive end-result.It's just a small one, because he did only play against low/normal rated players and the viktory-condition of the Viking-variant is not half SCs but much lower. This is expected using this system. Once you are in the top100 you need to play other top100-players to gain a significant amount of points.

I did a small simulation of the game if it would have been a draw, and this would have award 2 too... (http://vdiplomacy.com/HoF_9611.pdf)

On a related note: at one stage I had some kind of failsafe implemented for PPSC, so you can't lose points if you have more SCs than your opponent.
I removed this, because the impact was too small, but comparing this to a draw I think I will add this back. It seems more fair (but it won't change anything in this case)

@Oli Would it be possible to put a HOF minimum rating requirement in order to join games?

Yes, this should be easy to add, but as the ratings are still work in progress I didn't add this.

By the way, what would a minimum rating for the Hall of Fame be?

and what about the treaty of verdun where my solo gets zero points :)
I do not disagree with the new system, but I feel some tweaks should be made.
For example, Ruffhaus will never gain points as he is in 2nd place but the amount of effort and time is still enormous, and whether I played lower ranked players or not, I still feel I won with a certain amount of sweat. I rarely get any advantage by playing new players

Quick scenario demonstration of HoF points:

Bozo (Player 1) solo's vs 34 other players each at 1000 in WWIV, default win conditions of 50SC's to solo, WTA:

Re1 = 1 / ( 1 + ( 10^( (Rating2 - Rating1) / 400) ) )
Re2 = 1 - Re1
Rating2 = 1000
Rating1 = 1357

Re1 = 0.886457624
Re2 = 0.113542376

Rr1 = Pl1SCs/(Pl1SCs + Pl2SCs)
Rr2 = Pl2SCs/(Pl1SCs + Pl2SCs)

Pl1SCs = 50
Pl2SC's = 0 (WTA means all other players sc's set to 0 in case of solo).

Rr1 = 1
Rr2 = 0

D1 = Rr1 - Re1
D2 = Rr2 - Re2

D1 = 0.113542376
D2 = -0.113542376

So for:
scoreMatchPlayer1 = K * gV * mV * D1
scoreMatchPlayer2 = K * gV * mV * D2

Where gV = ((numberCountries -1 ) / numberCountries )^3 * (1 - (numberCountries / 100))
numberCountries = 35
and
gV = 0.595862391

(assuming mV is 1, not exactly sure what this match value normalization does for PPSC or WTA, Oli can you explain that further?)

K = 30

scoreMatchPlayer1 = 2.029668954
scoreMatchPlayer2 = -2.029668954

ScorePlayer1 = sum ( score for each matches)
ScorePlayer1 = scoreMatchPlayer1*34
ScorePlayer1 = 69.00874444

So Bozo can still get 69 soloing against 34 new players in WWIV.

But what about a classic game or a 1v1?
The only thing that will change is essentially number of countries and thus gV, so
gV for classic is 0.585655977 (practically the same as for WWIV)
gV for a 16 player game (max that gV can assume) is 0.692138672
gV for a 1v1 is only 0.1225

For a 1v1 game:
scoreMatchPlayer1 = 0.417268233
ScorePlayer1 = 0.417268233
(interestingly bozo gets no points, if decimals are truncated instead of rounded, unless he plays 1v1 games with players who have 1187 or more)

For a 3 player game:
scoreMatchPlayer1 = 0.9789876
ScorePlayer1 = 1.957975201
So bozo should still get 1 point from soloing a 3 player game against 2 new players.

For soloing a classic match:
scoreMatchPlayer1 = 1.994903139
ScorePlayer1 = 11.96941884

So Bozo most definitely can gain points against other players including playing against new players who only have 1000 .

This was all from following the wiki about the HoF scoring.

So cypeg, looking at your treaty of verdun game, I see something right off the bat that looks weird to me.

Even though you solo'd, the PPSC calculation looks to have given you a -9% difference against rancher. This shouldn't ever happen in a solo even in PPSC, so I agree something is wrong. I could see the current calculation coming up with something where a survive for rancher gets him more points than a draw, but not a better result than cypeg's solo.

The issue looks to be in the calculation of Rr1 = Pl1SCs/(Pl1SCs + Pl2SCs) and then D1 = Rr1 - Re1 for PPSC solos

For the person soloing, the number of SC's that everyone else has should be effectively 0 with respect to the person soloing, thus it doesn't matter for the winner how many second place had.

If the value of Rr1 can be reduced to less than Re1 even in a solo because of supply center differences, then the solo is not worth anything. A solo result for WTA looks to be a Rr1 of 1 and thus never less than Re1. Why should this be different for PPSC between the winner and the survivors? The supply center difference should only come into play for survive vs survive and survive vs defeat not survive vs solo or vice versa.

So, having worked through the math, I now agree there is still an issue with the calculation.

btw Oli, I really like the view of the HoF results by game ID:
http://vdiplomacy.net/hof.php?gameID=9540

This is tremendously helpful in seeing how the HoF ratings changes were calculated for a given game and for finding anomalies such as the one I outlined above. Excellent work, I heartily approve!!
Keep it coming (but don't neglect your family to do so), just letting you know how awesome your work is!

One question though in the match above, why is the match value so low for all players involved?

I see most players at 100% in most games except for what appear to be obvious takeovers.. what factors can affect mV and how is it calculated?

@Leif: But that's the way it's calculated in PPSC-games all the time.
And the -9% is something I will change. I checked a few examples and the end-result was to small to warrant a more complex coding (And it still is, we speak about 1 point, thats 0.1%). But I see this is still an issue.
Also another issue is the small K-Value. In Elo-ratings there is a KValue of 32 so in a match with 2-same-ratet-players you can loose (or gain) 16 ) in a typical game here it's only 5-10 . So this might need some tweaks too.

@cypeg: The new rating does not value points, it just compares each player against each other. If you are No.1 you have a hard time gaining more points. And you do not need to, as you are already No.1. Once you are no longer Top 10 it's much easier to gain more points again. So if you do not keep playing your score will devalue over time.

Because of this drawback (there is not much fun in gaining _no_ points) I always said I will keep the DPoints in place. They are just not an optimal measure of success (for an HoF for example). The Hof is much more accurate now than the old DPoint ones, as many people abused the DPoint-system to gain an amount of points disproportional to their knowledge.

Between 2 drawers each SC-count is the same, so the match-value is 0 and we do not count these games.
If one player has 50% of the SCs and another one has 30% the value is just 20%. This way the close the SC-count the less the value. (And that's why it's usually not a problem if a high rated player wins and has a close 2nd place with a much lower rating. Sure there is still some points to be lost, but it's a very small fraction compared to all the players he defeated.)

If you want to check the complete code you can take a look here:
https://github.com/Sleepcap/vDiplomacy/blob/master/lib/rating.php

I definitely do not agree with that logic Oli..

Re1 is the expected result, and Rr1 is the real result, correct?

When a player solos in WTA, the real result is 1, when a player draws the real result is 0.5 because the SC's between the two players are equalized.

Translating to PPSC, the real result should still be 1 (they won) for a solo, and 0.5 for a draw (they tied), and something between 0 and 1 for survives based on the ratio between the players.

In a large SC count game, say WWIV with the victory conditions set to half +1 with one 1 strong survive a solo and the rest defeated, the player who solos would get 50% + some small insignificant fraction, while the player surviving, would get slightly less than 50%, effectively a draw for Rr1. Even in PPSC this is wrong.

I think I understand now what Captainmeme was talking about in his post above because I've stumbled on the same thing here. He called it RR (which i misunderstood him to mean reliability rating), I'm calling it Rr1 from the wiki (real result for player 1).

PPSC should value a win in all circumstances with the difference between a draw and a strong survive being minimized compared to WTA and in some cases (based on our current interpretation, a strong survive being better for certain players than a draw).

I don't think there is or should be any debate about whether a strong survive is comparable to a solo. I know that in D points that the points won can be very similar between the strong survive and the solo victor, but this shouldn't be the case for HoF ranking at all. Here the two systems are very much not comparable and thus shouldn't be.

If this would require a reset to alter the formula (which I don't think it does) then I would lobby hard for a reset as the current calculation is very much incorrect for PPSC solos in my eyes and I would guess many other players as well.

and thus shouldn't be compared.*

player's eyes as well.*

I'm getting sloppy!!