TOP Players / Tier Requirements Thread - Updated 7/8/09

[jimerax]

Any "non-FC(or low combo) PA scores" can't be recorded or counted in current FFR system.
And Tier Point is only calculated by the highest Total Scores which are actually recorded.

We have to count the points of those PA scores by SS manually?
No, get over it.

also, stop posting your tier point list here.
There's a thread in Brag Board.

[gnr61]

Motion to start a manual calculation system based on a GOOD scoring system.
list will have WAY fewer songs (to offset things like 2 challenging AAAs being as good as a Crowdpleaser FC), thinking like 75ish.
Songs would be select and covering different areas of skill, rather than just every challenging+ song thrown together. Bit of a hassle it may be to calculate but it would be much more accurate I feel. AND NO DAMN AAA BONUSES rofl

[MrRubix]

Jimerax, your response surprises me -- I am fairly certain that it can be automated. Saying that you'd have to count points from screenshots is a bit of a straw-man argument: It isn't what I am proposing at all. Why else do you think levelranks.php takes a while to load? It's a php file -- it's performing calculations and comparisons on data extracted from the score database for every single song in the game. Such things could also be done for more specific criteria for songs in the Tier Points list -- right now the Tier Points calculator is using scores and from this deriving an approximation for full combos and the number of goods. Granted, this approach is easy to calculate and use as a metric, but it's hardly useful for a measure of skill.

Yes, doing it the way that ieat, gnr and I have been referring to would take a little bit more coding work, but it would not be difficult -- just a bit more tedious with respect to customizing appropriate prerequisites, and really it would not differ so much from the current system aside from the fact that it is using different numbers.

If I am wrong about any of this please let me know.


EDIT: I just asked Synthlight about how such values are derived -- what I've assumed above is correct. It's very possible to change the system over to something more appropriate.

[PsYcHoZeRoSk8eR]

I love how I have 16 AAA's and I'm on the list. That is by far the lowest AAA count. The next lowest person has 76.

[TheSeventhSign]

Originally Posted by MrRubix

I've always called it Perfect Accuracy (even though I know many say Attack)

ME TOO

And personally I think Tier Points based on PA would be a much better idea. Leave average rank to combo and Tier Points to PA..... at least in that way, they would be MORE different from each other..... I still think Tier Points follow average rank at this point.

[ieatyourlvllol]

The current coding behind FFR scoring simply follows this semi-pseudocode:

//
variables p, g, a, m, b
variables combo, total

time windows : hit : increment corresponding variable
combo = greatest distance between m increments

total = variable*value + 1000*combo
//

The current tier calculator uses the total value, which is anywhere from 1 to 7 digits.

For a PA-based system, you would omit the totaling procedure and could instead have a storage string that represents the accuracy increments. Ex: 5G1A1M1B would indicate a score of 5 goods with an average miss. Of course, the tier calculator would need to support the interpretation of variable-length strings, which I assume can be accomplished through basic case statements.

The whole process would be a bit tedious, which is kinda what JX was getting at.

[CR_fLiP_fOaM]

I got 4.1,1,2 on choprite and no tier points lol. Now that's stupid.

[remedy1502]

Wow, I'm sticking around until my tourney is finished, and I was Tier 4 when I had 252 points, now Tier 4 is 360 :S I have 266 LMFAO.

EDIT: I guess I could AAA a bunch of songs from subscriber week, I haven't played most of them. I only played VC and harder, and a few Challenging.

[cetaka]

rock the boat

[MrRubix]

Well I mean scores right now are calculated by 1000*Greatest Combo + 550*Perfects + 275*Good + 55*Average - 210*Miss - 20*Boo.

So take Pants for example. It figures out what an SDG's score is by Max combo*1000 + (Max Combo-9)*550 + 9*275, so when Max Combo = 3024, you get an SDG score of 4684725.

The problem isn't string length of whatever. It's just that the Tier Point calculator I think checks the score for their requirements against your max score from your levelranks. So if your score for Pants, for example, were more than 4684725, you'd get the SDG award, although technically it can be a misnomer. Say you get 9 goods but a bunch of Boos. You would technically NOT get the SDG req even though you have SDG, but not PURELY SDG.

Therefore, the Tier Point calculator should do what levelranks.php does -- look at all your scores for a given song. So, for example, let's take flip's Choprite score of 4 goods 1 Average 1 Miss 2 Boos. Let's say he missed at combo 1400. For a song with a max combo of 1796, this makes his score 1400*1000 + (1796-4-1-1)*550 + 4*275 + 1*55 - 1*210 - 2*20 = 2385405 out of a max of 2783800. According to a req that asks for 15 or fewer goods, this score is 1796*1000 + 1781*550 + 15*275 = 2779675.

Even though flip clearly got far less than 15 goods, his score of 2385405 doesn't really come close to 2779675. The system's obviously flawed somewhere. He is clearly capable of 15 or less goods (and, technically, he DID get 15 or less goods), but not the score for PURELY 15 or less goods (in the absence of any other variable).

Here's what I'd recommend. Say the 6/7 req for Choprite is "At least 1786 Perfects" (ten away from the 7/7 req, being AAA or 1796 Perfects). This could be checked as follows: Compare user's score (1790 Perfects) with req score (1786 perfects).

Therefore flip would get the requirement since he has the # of Perfects to meet the req. This is more fair than giving him 0 points just because he didn't combo it. Using the # of Perfects naturally accounts for avmissing (as it detracts from # of Perfects), but the only thing it does not really guard against is boocount (say flip had like 20 Boos or something). In my opinion, this is negligable, really (although there are ways to guard against this, it's a little much). IMO a higher boocount inevitably leads to a lower PA anyway.

Regardless, this is how I think things should be -- read the # of steps from a given songplay and use this to compare against the score req, NOT merely the score itself and deriving numbers from there.

[daringdaniel1111]

I missed the update because my wireless network adapter broke.

[jimerax]

Considering we have MP records and beta level leader, we already have a method for recording scores in detail (total score/perf/good/ave/miss/boo/combo), and probably it's possible to record all scores we've got and auto-calculating the points from those data technically (only server capacity problem).

Also from those data, it's still hard to determine how we should calculate "the most skill related points". Are there any practical ideas?

oh Rubix posted something..
If the "best perfect count" is recorded it would be definitely the factor of the point.
also, ''FC flag(no miss, boo count should be considered)", "AAA flag(no boo)" would be needed.

[MrRubix]

Actually jimerax, the comparison would be no more intensive than it is to check levelranks.php (which scans through 30 million scores every time you run it according to Synthlight).

Comparing specific values to specific reqs would not be a problem at all.

The only real issue would be deciding fair reqs for relevant songs. You can check against any variable you want in the database, so for higher-difficulty songs you can indeed throw combo in there as well in addition to PA.


"If the "best perfect count" is recorded it would be definitely the factor of the point.
also, ''FC flag(no miss, boo count should be considered)", "AAA flag(no boo)" would be needed."

Right, such things would be needed. My point is that going purely from combo is not indicative of skill as much as specific score breakdowns are.

[ieatyourlvllol]

Imo, it would really require revamping of the entire scoring system...

...unless you wanted to somehow merge the combo-count and accuracy-measure processes and write a calculator script for it. :doubt:

[MrRubix]

It would be a revamp, but I think it'd be a much-needed and more widely-accepted revamp, who knows. I have little faith in the current system -- there are many players who aren't ranked where they should be. I'm just trying to propose a system that accounts for this. Rewriting the system wouldn't be that hard (wouldn't take more than a day if you knew what you were doing. You'd just have to 1. Make the reqs 2. Code the reqs in a way that compares the specific scores to the reqs). IMO making the reqs would be a more time-consuming process (figuring out fair cutoffs). I just find it silly that people with great PA on certain songs get no tier points because they didn't get the FC in a game where combo can be acquired through antiskill.

[HammyMcSquirrel]

I apologize if this has already been brought up. I don't think it has been, but...

Couldn't the "Pre-Total" score just be used instead of the "Grand Total" score?

That doesn't take into account the combo, so why not use that?

Edit: That is...If the plan is to do some new form of coding, it may be easier to make a code to take that score. I don't know anything about how that would work and the difficulty, but it's just my thought on this.

[stlunatic0124]

Originally Posted by gnr61

PART of the game certainly. But why should it be ALL of the game, like it is now. On songs that are difficult to FC or even just to rack up a high combo, PA matters little to not at all. a 740 combo on crowdpleaser is like automatically top 15, but if you look at someone like ieat who has 9 goods or something on it but 492 combo, you can tell who deserves the higher rank.

I can't FC BB Revenge or Pants because of avmisses or brainblocks, but I can PA both better (about 30 goods best on each) than most people can.

There has to be some kind of balance. ONE miss shouldn't turn your score into crap.

I don't feel like reading the rest of the posts, but it's only ALL of the game for ****ty players. When it's a song that we're all going to FC, the difference for the top players is going to be the PA, whether it's AAA or 1 or 2 goods or whatever.

[gnr61]

Originally Posted by stlunatic0124

I don't feel like reading the rest of the posts, but it's only ALL of the game for ****ty players. When it's a song that we're all going to FC, the difference for the top players is going to be the PA, whether it's AAA or 1 or 2 goods or whatever.

This doesn't apply to A LOT of songs, mainly songs that CAN be FC'd just fine, but often aren't. Look at something like Kyuin. If I get 0 goods and an average miss, are you saying I'm a ****ty player because i can't compete with the people who DID FC it with an sdg or something similar?

ps - wtf hammy how is that cp rank 18? XS

[jimerax]

Pre-total score is quite dependent on PA, and it is sometimes used in unofficial tournaments. But I'm not sure it's actually recorded in the database (we can calculate it from other parameters).

Although perfect count would be better indication for accuracy IMO since it's purer.

[flipsta_lombax]

I like the ideas being presented in this thread. I mean 10 goods plus 2-3 avmisses in the middle of Choprite is just so annoying.