Calcul des points : la faille V2.0
163 Nachricht(en) normal
Zeige 141 bis 160 von 163 Einträgen
26/12/2013 13:21
#141
The new ranking is now open!!
26/12/2013 13:23
#142
Merci pour la mise en place de ce nouveau classement :)
26/12/2013 13:25
#143
Ouwéééééé ^^
26/12/2013 14:12
#144
+63 places pour moi, back in the top 100 !!
Formule encore améliorable au niveau des "gros" jeux je pense : sur Mario Kart DD je suis 51/161, ce qui reste honnête et n'est pas atteignable en 2 minutes, mais ça rapporte très très peu de points (dans les 500).
Formule encore améliorable au niveau des "gros" jeux je pense : sur Mario Kart DD je suis 51/161, ce qui reste honnête et n'est pas atteignable en 2 minutes, mais ça rapporte très très peu de points (dans les 500).
26/12/2013 14:15
#145
pfff... je suis 101ème.
Merci pour ce nouveau classement.
Merci pour ce nouveau classement.
26/12/2013 14:23
#146
GJ! Looks good. 
26/12/2013 16:06
#147
Very good job !
Il faut juste mettre aussi sur le profil : Rang pointsRecord
Il faut juste mettre aussi sur le profil : Rang pointsRecord
26/12/2013 16:59
#148
Djon :
Very good job !
Il faut juste mettre aussi sur le profil : Rang pointsRecord
Je ne vois pas de quoi tu parles
26/12/2013 17:48
#149
Bravo pour cette maj
26/12/2013 17:55
#150
Merci pour la MAJ :P
Par contre, faudrait aussi mettre la version anglaise, parce que comme ça :
Medals rank : 5 / 3,898
Points rank : 3 / 3,898
Proofs rank : 7 / 3,898
Badges rank : 3 / 3,898
Rang pointsRecord : 2 / 3,898
C'est un peu bizarre
niconoe :
Djon :
Very good job !
Il faut juste mettre aussi sur le profil : Rang pointsRecord
Je ne vois pas de quoi tu parles
Par contre, faudrait aussi mettre la version anglaise, parce que comme ça :
Medals rank : 5 / 3,898
Points rank : 3 / 3,898
Proofs rank : 7 / 3,898
Badges rank : 3 / 3,898
Rang pointsRecord : 2 / 3,898
C'est un peu bizarre
26/12/2013 18:40
#151
Haha ! Bien vu nico c'était rapide !
Par contre :
Top 5 PointsJeu
Gaz8886 0
Tout nouveau membre devient premier du top jeux
Edit : Apparemment c'est corrigé !
Par contre :
Top 5 PointsJeu
Gaz8886 0
Tout nouveau membre devient premier du top jeux
Edit : Apparemment c'est corrigé !
26/12/2013 20:32
#152
Je demande à barty de lancer un scan pour la VA. Je traduis au plus vite.
EDIT: translations are all ok for now.
EDIT: translations are all ok for now.
27/12/2013 20:48
#153
Asep :
+63 places pour moi, back in the top 100 !!
Formule encore améliorable au niveau des "gros" jeux je pense : sur Mario Kart DD je suis 51/161, ce qui reste honnête et n'est pas atteignable en 2 minutes, mais ça rapporte très très peu de points (dans les 500).
Comme le dis Asep, la formule doit encore être ajusté surtout sur les gros jeux.. ou sur les petits jeux
Exemple :
35/195 - 1k4 est égal à : 14/14 - 1k4
Mais sinon GG c'est un très bon début.
27/12/2013 21:18
#154
Apparemment, je suis 32 :O
GG à Packattack qui est dans les meilleurs avec seulement 49 jeux.
GG à Packattack qui est dans les meilleurs avec seulement 49 jeux.
27/12/2013 21:28
#155
même si j'étais relativement opposé à la mise en place de ce classement; je dois reconnaître que c'est une bonne chose.
bravo aux développeurs, et à tous ceux qui ont contribué à cette formule, VGR évolue dans le bon sens...
vous avez vu qui est en tête???
mon choix pour le membre légendaire 2013 n'est que confirmé par ce classement!
EDIT, j'étais 13° du classement pointsRecords, me voilà 13° du classement pointsJeux!
bravo aux développeurs, et à tous ceux qui ont contribué à cette formule, VGR évolue dans le bon sens...
vous avez vu qui est en tête???
mon choix pour le membre légendaire 2013 n'est que confirmé par ce classement!
EDIT, j'étais 13° du classement pointsRecords, me voilà 13° du classement pointsJeux!
27/12/2013 23:29
#156
Très bon nouveau classement ;)
29/12/2013 13:41
#157
Thank you niconoe and OkamiAma77 for the new Games ranking!
Now that this Top 100 Games is the new VGR main ranking, the formulas behind it should probably deserve some polishing. Here are a few thoughts and remarks:
- Score of the first player:
Let me consider the case with no ties (y=1). The shape of the proposed formula looks good, but its parametric expressions looks obfuscatingly complex. Indeed, a simple nonlinear calibration reveals that the following expression provides a decent approximation of it:
g(x)=1000+2986*(x^0.67 -1)
This power function is much simpler, smoother, and easier to understand and to tweak (1000 is the score of a player alone on one game, 2986 is a shape coefficient, and the power 0.67 sets the strength of the concavity, which promotes diversification).
- Speed of the decay:
The same remarks can be made for the decay. In a similar manner, the following simpler expression could be used:
d(z)=min( 0.99 ; 0.7 + 16.02*(z^0.0042 - 1) )
Again, it is easier to understand and to tweak (towards a less severe decay, according to the first feedbacks).
- Ties for first place:
Consider now that several players share the same Master Badge. As it is much less likely to tie for first place on a game than on one single chart, there is maybe no need for a fancy formula to deal with it, and the usual averaging may do the job.
If you really think such ties should be penalized, then there are probably simpler ways to do it. For example, one possibility would be to base the awards of points not on the number of players (x) but on the number of players beaten by the first player, plus himself (x-y+1). As an illustration, if all the players on one game were tied for first place, they would all earn 1000 points (as if they were alone on the game).
- Number of records:
The new Top 100 Games is a welcome contribution, and is a useful new tool to compare players (though it is closely related to the Top 100 Badges). However, a feature which is surprising (especially when one is used to the Top 100 Records), is that it does not take the number of charts into account at all: a game with only 1 chart is equivalent to a game with 1000 charts, as long as there are the same number of players on both games.
A first remark is that, on a game with 1000 charts, the fact that a player with one score counts as much for the point formula as a player with 1000 scores seems unfair and can have nasty side effects… (A possible fix would be to base the formulas on the average number of players per chart (Posts/Records), though the games with many charts would end up disadvantaged).
A second remark is that, even though the problem with the Top 100 Records was due to the large number of charts on some games, neglecting this factor altogether seems to be another extreme. An intermediate possibility would be to actually take the number of charts into account, but in a sublinear fashion. To illustrate the idea, I slightly modify the Top Records as follows: for each player, instead of summing the PointsRecords from every game ( Σs(i) ), I take it to the power of p before summing ( Σ100*(s(i)/100)^p , with p between 0 and 1 to promote diversification (by concavity)). Below is the resulting Tamed Top 10 Records with p=0.75 (for a Top 100, ask someone who has access to the database)
1: 33Tails (+1) : 2,087,161
2: Flothaboss (-1) : 1,656,922
3: Kirby 54 (+3) : 1,574,046
4: BlackShark (+1) : 1,529,987
5: wadoludo (+2) : 1,419,853
6: Triple HHH (+2) : 1,384,995
7: daftpunk (-3) : 1,376,367
8: OkamiAma77 (+2) : 1,060,370
9: narutokyuby (+11) : 1,058,125
10: thelegendarypsr (-1) : 1,051,921
Compared with the actual Top Records, narutokyuby would enter the top 10, while André Luis Fernandes would leave it, showing how this simple trick promotes diversification among many games.
My point is that, between the Top Games that emphasizes the number of games, and the Top Records that emphasizes the number of records, a balanced Top 100 VGR could, in my opinion, be reached between the current Top Games and this Tamed Top Records, with the number of charts being part of the equation (to encourage people to post on all the charts of a game), but not in a linear fashion as in the current Top Records.
(Something like multiplying the current number of PointsGame of each player on each game by a (concave) power of the player’s number of records on the game may do the job, up to some fiddling of the parameters)
These were my remarks about the new formula. Feel free to comment these points.
Slowly but surely, VGR is making progress towards better formulas!
Now that this Top 100 Games is the new VGR main ranking, the formulas behind it should probably deserve some polishing. Here are a few thoughts and remarks:
- Score of the first player:
Let me consider the case with no ties (y=1). The shape of the proposed formula looks good, but its parametric expressions looks obfuscatingly complex. Indeed, a simple nonlinear calibration reveals that the following expression provides a decent approximation of it:
g(x)=1000+2986*(x^0.67 -1)
This power function is much simpler, smoother, and easier to understand and to tweak (1000 is the score of a player alone on one game, 2986 is a shape coefficient, and the power 0.67 sets the strength of the concavity, which promotes diversification).
- Speed of the decay:
The same remarks can be made for the decay. In a similar manner, the following simpler expression could be used:
d(z)=min( 0.99 ; 0.7 + 16.02*(z^0.0042 - 1) )
Again, it is easier to understand and to tweak (towards a less severe decay, according to the first feedbacks).
- Ties for first place:
Consider now that several players share the same Master Badge. As it is much less likely to tie for first place on a game than on one single chart, there is maybe no need for a fancy formula to deal with it, and the usual averaging may do the job.
If you really think such ties should be penalized, then there are probably simpler ways to do it. For example, one possibility would be to base the awards of points not on the number of players (x) but on the number of players beaten by the first player, plus himself (x-y+1). As an illustration, if all the players on one game were tied for first place, they would all earn 1000 points (as if they were alone on the game).
- Number of records:
The new Top 100 Games is a welcome contribution, and is a useful new tool to compare players (though it is closely related to the Top 100 Badges). However, a feature which is surprising (especially when one is used to the Top 100 Records), is that it does not take the number of charts into account at all: a game with only 1 chart is equivalent to a game with 1000 charts, as long as there are the same number of players on both games.
A first remark is that, on a game with 1000 charts, the fact that a player with one score counts as much for the point formula as a player with 1000 scores seems unfair and can have nasty side effects… (A possible fix would be to base the formulas on the average number of players per chart (Posts/Records), though the games with many charts would end up disadvantaged).
A second remark is that, even though the problem with the Top 100 Records was due to the large number of charts on some games, neglecting this factor altogether seems to be another extreme. An intermediate possibility would be to actually take the number of charts into account, but in a sublinear fashion. To illustrate the idea, I slightly modify the Top Records as follows: for each player, instead of summing the PointsRecords from every game ( Σs(i) ), I take it to the power of p before summing ( Σ100*(s(i)/100)^p , with p between 0 and 1 to promote diversification (by concavity)). Below is the resulting Tamed Top 10 Records with p=0.75 (for a Top 100, ask someone who has access to the database
)1: 33Tails (+1) : 2,087,161
2: Flothaboss (-1) : 1,656,922
3: Kirby 54 (+3) : 1,574,046
4: BlackShark (+1) : 1,529,987
5: wadoludo (+2) : 1,419,853
6: Triple HHH (+2) : 1,384,995
7: daftpunk (-3) : 1,376,367
8: OkamiAma77 (+2) : 1,060,370
9: narutokyuby (+11) : 1,058,125
10: thelegendarypsr (-1) : 1,051,921
Compared with the actual Top Records, narutokyuby would enter the top 10, while André Luis Fernandes would leave it, showing how this simple trick promotes diversification among many games.
My point is that, between the Top Games that emphasizes the number of games, and the Top Records that emphasizes the number of records, a balanced Top 100 VGR could, in my opinion, be reached between the current Top Games and this Tamed Top Records, with the number of charts being part of the equation (to encourage people to post on all the charts of a game), but not in a linear fashion as in the current Top Records.
(Something like multiplying the current number of PointsGame of each player on each game by a (concave) power of the player’s number of records on the game may do the job, up to some fiddling of the parameters)
These were my remarks about the new formula. Feel free to comment these points.
Slowly but surely, VGR is making progress towards better formulas!
29/12/2013 15:32
#158
wow 35eme pour le rang points o_O Jai jamais pensé etre aussi haut dans le classement apres la mise jour lol
29/12/2013 17:33
#159
TigrN :
However, a feature which is surprising (especially when one is used to the Top 100 Records), is that it does not take the number of charts into account at all: a game with only 1 chart is equivalent to a game with 1000 charts, as long as there are the same number of players on both games.
It doesn't take the number of charts into account at all and it shouldn't. I think that's the beauty of it. A game is a game. The competition can be more fierce in a 1 chart game than in a 1000 charts game. And a game with 1000 charts has more medals to fight for. And some records can be time consuming, as others are quicker. Number of records shouldn't be mixed in the new points rank.
The current system will encourage players to submit to more charts as they fight for better position in the game ranks. We just need to give it time.
But I do agree that the current formula needs some polishing. As Djon also pointed out.
The question is how much points should the player ranked at 1st place get compared to the player who's last. Or in a game that has 100 players, how much points should 1st, 10th, 50th and 100th get?
29/12/2013 18:10
#160
TigrN :
- Score of the first player:
This is the Okami part. I asked him to give a formula that keep giving points to the 500th on 500 people and have a logical way to reduce the given points exponentially from the number of 1st tied.
- Speed of the decay:
This is my part, and I wanted to give 0.75*1st point to the 2nd and up the 0.75 coefficient to reach 0.99 only for the 100th ranked. The way to move from 0.75 to 0.99 should be logarithmic to me and not linear, that's the aim of the decay. I don't know if your formula do the same trick but if it does, I need to compare the growing of the 2 functions.
- Ties for first place:
A simple avergare isn't sufficient and the point is if we have only 1st, no matter if there's tied 2nd or 3rd or others, the 1st place is considered as "hard" to get (the word "hard" gets more valuable for more people playing).
If we have 2 firsts on a game, then it's considered as "hugely easier". I'm not talking about 2 tied 1st on a record but on a game. That means tied for all charts (or similar). That's why the Okami's formula is only based on 1st and take in considerations the number of tied.
- Number of records:
I also wanted to make evolve the formula with a way to include the number of records. This is important to me because I think if you improve a score on a record but this don't give you a new rank, you should be congratulate but currently you're not... and I think it's a problem too...
I didn't find a way to do this fairly, but the thing is there was a poll, and the result said "only use the game ranking, not the number of record".
So that's what I did to respect the more people opinions I can on VGR.
The pointRecord ranking must stay like this for me. It really defines who's completly finish a game and doesn't simply post half of scores and don't send the rest because "I'm not first on it" or something like Zimer's behaviours. It's purely a quantitative ranking and it takes sense to have both quality and quantity on all VGR rankings
.Thanks anyway for your contribution and the time you passed to explain and develop your formulas. I wished I had your help with Okami when argueing on this new one.
Zeige 141 bis 160 von 163 Einträgen
You must be logged in to reply to this topic.