Ranking Suggestion

[Tasselfoot]

Re: Ranking Suggestion

be warned that i'm going to be a harsh critic of any potential algorithm.

[Dave&Wave]

Re: Ranking Suggestion

Well , the algorithim would use the average rank , but it should have something to do with how many songs you've done . This doesn't take an algorithim wiz , but just a smart person . It will use the average rank and the tier points , and propably the amount of songs you have done . Maybe if it subtracts the positive tier points from the average rank . Or ... I'm not sure . But that's just to get started .

[StringsnKeys]

Re: Ranking Suggestion

To Dave and Wave, a new account with no songs played has an average rank of close to 40,000 having an average rank of 40,000 because you triple AAAd one song, isn't going to put you too far ahead of everybody.

To Dore, I fail to see how -
FFR Overall Rank: 404
FFR Average Rank: 3,278
makes you a top 1000 player based on my previous suggestion? (I could be wrong though)

In any case, I get it......dragging down statistics for close to 1.6 million players and better yet, doing a bunch of calculations with them, will take a lot of server power. I suppose the goal is to play the game and enjoy yourself which won't be fun if everything starts lagging. For the sake of the argument though, I will start working on some algorithm for a "master ranking" system.

[dore]

Re: Ranking Suggestion

To Dore, I fail to see how -
FFR Overall Rank: 404
FFR Average Rank: 3,278
makes you a top 1000 player based on my previous suggestion? (I could be wrong though)

I'm 983rd in average rank

[StringsnKeys]

Re: Ranking Suggestion

Fair enough but as "strange" as you think that may be, you're sitting in that spot for a reason and it's probably not because you're bad...........I don't see why with 293 AAAs, you don't deserve that spot. I'm around 590 (rank) but even though I've played virtually every song, I only have 61 AAAs. I'm sure once you've played every song you will soar past me and lots of other people in rank. I don't think the concept is that difficult to deal with. Those who play more songs, will have a better rank.

[ieatyourlvllol]

Re: Ranking Suggestion

I would agree that the usage of the term "overall rank" for describing grand point total is completely misleading, particularly to new members. However, like Tass has stated time and time again, switching to average rank is difficult because of the implications it has in terms of server workload.

I'm not sure how FFR's score recording/updating/referencing system works, but there is almost definitely a less server-intensive method of data management.

For anyone interested in attempting to find a solution...

Currently, FFR probably uses something similar to the following: take, for instance, song 'A', which links to a descending array of the personal records of each member. To obtain a person's rank based on a given score 's', the server searches the array for the two elements 'a' and 'b' for which 's' is intermediary, and it then uses the corresponding ranks (unique or shared ordinal placement) 'a.r' (lower value = higher rank) and 'b.r' (higher value = lower rank) to give the person rank 'r', where 'r' = 'a.r' + 1. Afterwards (or perhaps before), all array elements with ordinal placement equal to or greater than 'r' are incremented by 1. I'm not sure, however, as to how FFR's storage/memory infrastructure works.

In any case, an average rank is obviously calculated by compiling the sum of levelranks and dividing it by the number of ranks. This operation is performed whenever a person looks up his/her average rank. However, since even if the individual doesn't personally acquire a new rank on any song within a certain time interval, it does not imply that others will not. For this reason, every time anyone else makes a request [ex: looking up a user profile or the top 1000 (if you're a subbie)], the operation/s must be repeated anew. Of course, I'm assuming that FFR does not use old storage values.

On the other hand, overall rank is much friendlier (to the server) because it's based on single incremented values for every user (rather than a comprehensive and communally dependent ordinal system).

If I have some spare time, I'm probably going to try to devise a method through which the ponderous calculation of average rank can be bypassed (to an extent). I encourage everyone else to give it a shot as well.

[jimerax]

Re: Ranking Suggestion

Average Rank isn't too meaningful unless you have enough credits to get all purchased/secret songs and have motivation to keep playing all regular songs.

[devonin]

Re: Ranking Suggestion

Every system for measuring the relative skill of players fails in some way simply because of the possibility that a skilled player simply won't want to do whatever specific playstyle is required to generate the number being used.

Average rank only works if you play on every song
Tier points only work if you play all the tier point songs
AAAc and FCs only work if you play every song

None of the usefully generated stats mean anything to someone who elects to not take part in that portion of the play experience, and so we could argue that NO ranking is remotely accurate or useful.

[devonin]

Re: Ranking Suggestion

A system that gives more weight to harder songs would be generally more accurate to measuring skill simply because skilled people will/ought to be playing harder songs more often than easier songs.

I also have suggested for ages that credits should scale to the difficulty of the song rather than just a straight score/50000 It shouldn't ever be the case that you could get more credits for a middling combo boo-heavy 6 or 7 that just happens to have 2000 arrows in it, than for AAA'ing a 12 that just happens to be 30 seconds long.

[ieatyourlvllol]

Re: Ranking Suggestion

Tier points are more suited for the purpose of indicating pure skill, and (if I'm not mistaken) one the main benefits of using them is that since the requirements and point assignments are static (until the entire system is updated), the load on the server is substantially lower than what you would encounter with average ranks. Nonetheless, I still think there should be a way to publicly display average rank for the dedicated (and/or time-endowed) members who have made an effort to cut down on all of their songranks. That is, unless tier point count/level was implemented as a replacement - at that point, for people who strive for skillrank (particularly higher-level members), there would no longer be as much of a purpose in playing the easier songs, so the overall need for average rank would slowly fade.

devonin: Yeah - Kairon suggested earlier that there should be a metric comprised of all three ranks (tier, average, overall), but as you said, each one requires concentration in a particular aspect of play. Thus, the inherent nature of each is such that they must remain separate for the sake of avoiding ambiguity. The skill-based system of tier points could be effectively employed as a complement to the time-oriented metric of overall ranks, but that would still leave average ranks unaccounted for. For this reason, I might try to find a way around the problems posed by the calculation of average ranks.

In addition, though my opinion is probably slightly skewed, I strongly agree with your stance on credit assignment.

Sprite-: That's a good idea - also, in the upper range (9-12), we would probably want to divide even further due to the gigantic difference between "Very Difficult" and "For Guru's Only" songs.

[jimerax]

Re: Ranking Suggestion

Average rank only works if you play on every song
Tier points only work if you play all the tier point songs
AAAc and FCs only work if you play every song

None of the usefully generated stats mean anything to someone who elects to not take part in that portion of the play experience, and so we could argue that NO ranking is remotely accurate or useful.

True, so maybe there isn't a prominently good ranking system enough to change the current ranking system even though these are better than GTS ranking as "measure of skill".

Average Rank = quite easily influenced by inactiveness compared to Tier Points/AAAs/sum of best scores (that's why I removed it from TP bonus). but when you play all songs it's the least biased system.
Tier Points = by far the most accurate for TOP players, but it doesn't work well for not very skilled players. plus it requires huge server load.
AAAs = it's not influenced by inactiveness of players that much, but even good scores like bf/sdg aren't counted at all, so it's a bit biased. especially people who don't have AAAs won't like this system.
Sum of best scores = same with AAAs, it's biased with combo ability due to FFR's scoring system (1 mash FC score is far more important than 100 awesome PA scores). this system is quite fair and simple in terms of accomplishment though.

I always like to see new better ideas for the ranking system.
difficulty adjustment could be used for it.

[StringsnKeys]

Every system for measuring the relative skill of players fails in some way simply because of the possibility that a skilled player simply won't want to do whatever specific playstyle is required to generate the number being used.

Average rank only works if you play on every song
Tier points only work if you play all the tier point songs
AAAc and FCs only work if you play every song

None of the usefully generated stats mean anything to someone who elects to not take part in that portion of the play experience, and so we could argue that NO ranking is remotely accurate or useful.

Although average rank only becomes accurate if everyone plays every song, it should be considered that those people who actually care about their rank, will take the time to play every song. I know one of the arguments has to be that a player might only play 50 songs and AAA all of them and still have a crappy average rank. However, as inaccurate as an average rank for that player may be, it would be ridiculous to calculate an average rank strictly based on the songs played by a given player. Then everyone would just go play excite bike once and have average rank 1. So as, "unfair", or, "inaccurate", as that may be, it's really not that unreasonable. Unless, a system were to be used where, like tier points, only certain songs were used to calculate average rank. So hypothetically only 10 songs from difficulty 1-12 were used in average rank so that a player only had to play 100 songs instead of 621.

On the credit assignment note, I totally agree. Perhaps a new thread should be started to tackle that issue with a breakdown of how the "new system" should work.

[devonin]

Re: Ranking Suggestion

Although average rank only becomes accurate if everyone plays every song, it should be considered that those people who actually care about their rank, will take the time to play every song.

People who care about their average rank take the time to play every song.
People who care about their overall rank take the time to play high scoring songs.
People who care about their tier points take the time to play tier point songs.

That's the whole justification for suggesting that none of these are necessarily "accurate" because many people of skill greater than those who are ranked highest simply don't care to do what is required for that specific rank.

[StringsnKeys]

Re: Ranking Suggestion

I'm not sure that the argument of, "people can't be bothered to meet the necessary requirements to achieve a good rank, so the ranking system is inaccurate.", has any validity. That's kind of like complaining that it took you 4 tries to get your drivers license when you couldn't be bothered to learn the rules of the road.......and then accusing the ministry of transportation of an unfair system.

[NSane]

Re: Ranking Suggestion

It sounds like a good idea, for there are those players out there that think if they get Overall Rank 1 they are automatically the best (But sadly that is not the case). Both ranks exist for a reason, Overall is purely score based and can take longer to become #1 in. Average rank is skill based and can be lowered as your skill increases and your ranks go down. So technically the not-as-good players can still have a chance to be #1 at something :P.