More Scoring Changes
by Xxav on July 9, 2009, 2:05 am, under World of Warcraft
I’ve made a few more significant changes to the scoring system that will reward guilds more from progress rather than quantity of achievements. The site is called guildprogress.com after all 
- See http://wow.guildprogress.com/US/Ysondre/Exodus/score for the latest scoring schematics ( or your own guild’s scoring breakdown page ).
A new algorithm is now being applied that grants a percentage of the above values based on kill order. This algorithm is NOT applied to minor achievements. It is roughly as follows:
- World 1st – 100%
- World 4th – 90%
- 18th – 80%
- 76th – 70%
- 326th – 60%
- 1385th – 50%
- 5884th - 40%
- 25000th – 30%
Special thanks to Wasserjunge for coming up with the algorithm. Thanks for the feedback so far!
July 13th, 2009 on 8:37 am
“These “fun” achievements should do little more than act as a tie-breaker to differentiate two nearly equal guilds”
I have to disagree, but I’m not sure there’s anything I could say to convince you. Call it “hardcore vs casual” or whatever you’d like (my guild is working on ‘normal’ Yogg at the moment), but who’s to say that Crazy Cat Lady is easier than FL+1?
One of the reasons I loved this site was because the value of an achievement was based on how many guilds had done it already. So there was no objectivity of casual vs hardcore players; the numbers spoke for themselves.
It also provided a natural decay as more guilds made progress. Constantly increasing tiers isn’t sustainable unless you’re just going to reset everything come 3.2. It makes more sense to have a set value for a bosskill and for each achievement, and then have them scale DOWN as more guilds make progress. Which is what you had.
I don’t want to discount all the work that Wasserjunge has done, because it’s great work, but can’t we just re-calibrate to make the value of a bosskill be relative to the number of guilds who have accomplished it, and scaling down instead of up?
IE: Having Alone in the Darkness worth 22000 and Flame Leviathan worth 100 is kinda odd (especially considering the “higher tiers of difficulty” we’ll have in 3.2); make Alone in the Darkness worth 200 and Flame Leviathan worth 1 (if so many people have done FL that your scaling gives that number).
July 13th, 2009 on 11:33 am
Ebs, first of all, I don’t think you understand the mechanics of regression algorithms. They act exponentially, so you must tier difficulty based on what you are trying to actually rank. The point of this site is to rank PROGRESS! Which means the greatest accuracy must apply to most progressed guilds.
If you wanted to re-calibrate the system based on individual bosses, you could, and it isn’t an altogether bad idea. The problem is (and I am not the code-designer is that you can end up with a situation that calculations become so highly variable that downtime for calculations gets too high. I DESIGNED this system around the number of guilds that have taken a couple of bosses, so you cannot state that it is not representative of that.
It is also set up such that if you are a guild still working on Yogg normal mode, but have spent time working the other achievements in the process, that you have a SIGNIFICANT rank above other guilds at the same point in progression.
I am not the first (I am sure), nor will I likely be the last to explain to you that PROGRESSION is a moderately hardcore concept. There are MANY guilds on my three primary servers that have Yogg-Saron down. Of those, most have some progress into the hard modes. I need this site to ascertain a rank discompassionately and accurately. A guild that has done 20 achievements but no hardmodes has no business being ranked above a guild that has done three hardmodes but only 5 of those marked achievements…to do that is not a ranking of PROGRESSION. Most achievements are not ACTIVELY sought by guilds. Many of these are because they slow down or detract weekly time to work progression fights, or because they interfere with performing the hardmodes of those encounters. As such, I don’t want to PENALIZE guilds for not putting an active focus on these…what you are suggesting WOULD do just that. Under this system, which is now tiered very well (Comparing the values, I have to agree with the rankings as they are falling…other than the fact that I am working on algorithm adjustments that will keep the same curve boundaries with slightly less diliniation.)
I am open to what you are suggesting, but you need to review the math before you make suggestions. Making FL worth 1 point and Alone worth 200 is not really different in any system than 100 and 22000. Keeping the ratio the same lets say 1 and 220 for instance, it won’t change the system AT ALL. But then everything would have to be scaled down in point values by a factor of 100. If you want to rank order of achievements by when people do them, you insert penalties that you are not fully aware of…for instance, my guild did +towers on second week, we are NOT going to go for Unbroken or Shutout until we have decided we don’t want the hardmode loot from FL anymore….we will get them, but a progression ranking should not punish us for getting those achievement later. I have already set up the system to penalize us, but not in such a way that it misrepresents our progression rank. Will it put us behind many people that are at the same point of progression as us?…most likely…yes. But it should. Those guilds that have these things and the fights we have down, should be ahead of us…which is why these achievements are not affected by order.
As far as resetting tiers…that is always the idea. Do you see any Naxx fights factoring into Ulduar numbers? I am currently beginning work on setting up (only for 25 man progress) numbers that will incorporate normal and heroic stuff for Argent Coliseum. There is still not quite enough information to finish…but it will be a whole new tier, and a whole new number system. There is no reason to apply more than a modifier of 0.0-1.0 modifier to the ranking of someone from the prior tier, and that modifier only helps define initial standings.
I am going to sum this up with this, and you can flame you all you want as a result. If it takes your guild much time (if any) to be able to kill Algalon or do Alone in the Darkness, then you are not the ones I am trying to set up an accurate system around. I want the system in those lower ranks to be as accurate as possible, but it must be chosen to have high full-scale inaccuracy or an accurate high-scale and a semi-accurate lower scale…as I said before, I am working on altering the dilineation of my regression algorithm to improve full range accuracy, but this will take a significant amount of time…I am hoping to have this done by 3.2…there are about 12 layered functions, I am happy to give you the systems and tables I am working with if you think you can crunch the numbers better than I can, I just know that a single algebraic error brings me back to square one, and explaining logarithmic scales to people is something I don’t have the time to do.
July 13th, 2009 on 12:44 pm
Just fyi, noticed a typo on Knock,Knock on Wood. It’s coming in as 7700 points instead of 700.
July 13th, 2009 on 1:00 pm
“I DESIGNED this system around the number of guilds that have taken a couple of bosses, so you cannot state that it is not representative of that.”
The problem with your design is that it is only a snapshot of difficulty right now. You’ll have to do the work again when 3.2 comes out, then when Icecrown comes out, etc. Why not build in “this boss/achievement is worth a value based on how many people have killed him/completed it”?
We can debate the definition of what constitutes progression and what doesn’t until the cows come home. To me, it’s progression to do something that’s measurable before other guilds do. It doesn’t matter if it gives better loot or not.
As for resetting tiers, I look back to TBC. Most guild progress sites included both T5 and T6 progress, because there was a large number of guilds who hadn’t completed T5 when T6 came out. When 3.2 comes out, many guilds will not have killed Hard-Mode General/Mimiron (let alone Yogg+0 or even Algalon), and I’d imagine they’d continue to work on this. Naxx is a bad comparison because everybody had finished the majority of the bosses when Ulduar came out, and Ulduar gave us 14 new bosses (and numerous hard-modes and achievements) to work on. 3.2 will bring us only 5 bosses.
Having a max value that decreases as more people get the kill will keep you from HAVING to reset. In this case, I’m only suggesting that you invert the current values. Decide on a max value for the hardest achievement (alone in the darkness right now, and whatever we see in 3.2 once that comes out), and extrapolating from how many people have downed Algalon and decrease it. It allows you to:
1) Not have to revisit the numbers when each content patch is delivered
2) Not have to reconsider the weight each fight is giving if you’ve underestimated or overestimated the difficulty of said fight
3) Point to a formula that answers the question “where does this number come from?” that removes any possibility of bias
And it also removes the “subjective” nature of how difficult an achievement is. I’m not trying to decry your work, or the numbers you came up with based on how many guilds gotten Mimiron or General hard-modes. I’m sure the numbers are very accurate right now. But instead of subjective constants assigned based on current progression, it should be very easy to factor in the number of people who have gotten the achievement. This would be a pre-processing step of updating, and thus should NOT be a major performance hit.
Consider this: you’ve had the hindsight to know exactly how difficult a fight is. From where are you basing this information? How many guilds have killed it. How would your numbers have changed had you tried to come up with numbers the day Ulduar went live? The first week? The first month? Would you have had to be constantly updating them?
(And you don’t have to explain mathematical systems to me. I work as an engineer for a very well-known software giant, and have a degree in Mathematics and Computer Science. Just because I’m not as hardcore of a WoW player as you, doesn’t mean I’m stupid.)
July 13th, 2009 on 4:16 pm
As an example of what I was talking about:
Point value per kill/achievement =
C*(1-ln(x/1.85 +1)/9.2775)
C = Constant for what the most difficult boss should be worth, x=number of people who have killed this boss.
With C=750 (to line up with Wasserjunge’s initial difficulty scaling), this will give:
achieve # killed (x) value
1 light 139 399.7566811
3x knock 163 387.03716
firefighter 181 378.6596196
saronite 377 319.7692232
steelbreak 815 257.6580566
illusion 1215 225.4377371
heartbreak 2793 158.217617
orbit-uary 2928 154.4041336
cache rare 1928 188.1558675
observed 28 525.1819563
alone 0 750
To arrive at this number, I averaged out the # of guilds who have killed each hard mode, and then plotted that against the difficulty weight that Wasserjunge assigned to their difficulty (150 for easy, 250 for medium, etc). I plotted that into excel and created a trend line to do a curve fit. Not the best technique, but I can’t be bothered right now to work out the full math.
Note that these formulas are for the “full” value of the achievement, and progressive ones should be split up using an algorithm similar to what Wasser posted earlier (taking the inverse percentage of # guilds who have done each tier).
Example calculations for 3-tiers of 3xKnock (1787 1+, 474 2+, and 163 got 3):
Temp[x] =>
x=3: 1787/163
x=2: 1787/(474-163)
x=1: 1787/(1787-474-163)
WeightC = Temp[1]+Temp[2]+Temp[3]
Weight[x] = Temp[x]/WeightC
Which would give a breakdown of:
3xKnock: 232.3355
2xKnock: 121.7707
1xKnock: 32.9310
(Again, assuming a 750 weight)
I honestly don’t see a problem with using this approach. The formula should be tweaked as I’m sure the numbers are a little off (once you hit ~20k people reaching an achievement, you go into negative weights which is obviously not desired). I could sit down and crank out a more reliable formula though.
July 13th, 2009 on 11:02 pm
One quick thing I will point out is that, currently, each “era”’s (10man Uld, 25man Uld, WoTLK Release) scoring is self contained in that era. 25 man Ulduar scoring will not affect 10 man Ulduar scoring. So once the new raid dungeons are released the progression rankings will essentially reset. Therefore, scoring would not have to be significantly changed for previous content.
I will admit that I was hesitant at first in hard coding in scoring values. I was trying to avoid this in original system. I do think, however, that the new tiered scoring does bring the rankings more in line with what the majority of people want to see when looking for guild progress.
It would be nice to have an equation that would be able to do the tiered scoring automatically based on the percent of guilds having completed content. But completion percentage does not seem to directly follow player’s opinions of difficulty and progression.
With that said, I look forward to reading more comments regarding progression tracking.
July 14th, 2009 on 1:35 am
It is very easy for me to rewrite a regression Algorthim which is fully variable on values based on number of guilds that have taken the achievement or kill.
The reason I don’t want to scale non-hardmode achievements differently…as I have said MANY times now..is that I don’t want to penalize guilds that are forced to skip those achievements to do the hardmode kills. This needs no further justification. Most sites don’t give ANY credit for these, you should be content just in the fact that this one does, and it DOES make a difference, as you can view on any server if you look at the rankings.
The problem with an algorthim which is constantly variable based on number of guilds that have taken it is that the value for a kill changes each week. This means less information is pulled from a database and more calculations must be done each week. Yes, I can accurately draw one up, yes it might be more accurate…but I know enough about website building to know that the more active calculations and the less you have databased into an .asp or .mdbx (god bless xml microsoft format for databases) the slower the site will run and that is not ideal.
As for your progress concerns for 3.2 content…I think you need to read the patch notes a little better. There will be a 25-man “easy-mode” which will only be a small step up from normal mode 10-man Ulduar content. I would expect that guilds are more than capable of accomplishing some of this material fairly quickly, and as it has better loot, this will be the trend of movement for guilds regardless of where they are in Ulduar (just as, if you recall, once they removed attunements, guilds all began MH/BT regardless of where they were in SSC/TK). As such, progress can be tracked fairly easily without concern of Ulduar content. Essentially, I will have to construct a similar tiering system for 25-man normal mode and heroic mode, which is far less difficult than you might think.
Before you continue to bash the system in place without SEEING it work…why don’t you come up with REAL suggestions by pulling up your own server, looking at guilds on that server and finding something that you actually feel is being misrepresented in the final results. If you find something that you feel is misrepresented, post a link to it here, tell me what it is that is misrepresented and I can see if there is a logical way to fix it that won’t destroy the system as a whole…but don’t bash a system without evidence of it not working…
All you have done is come here and whine without showing any evidence of what you feel is wrong. Your suggestions to “fix” what you haven’t even shown to be broke are not really effective unless you want massive site-lag and server downtime for calculations. To make this make better sense to you, basically EVERY time ANY guild gets a new achievement or kill under what you are suggesting, EVERYONE that has done it have to be recalculated in order to maintain accuracy…and that kind of calculation-intensive system is precisely why I have used a non-linear regression curve against a highly tiered system..because by design it *theoretically* does the same thing without being as taxing on the system.
I don’t want to get in a flame war, this is not my intent. Just, you are expressing that you see a problem with something without showing clear evidence that something is broke. Further, you are suggesting an intensely calculation intensive system which is not really ideal for web environments without understanding (at least it seems to me you don’t) what a proper non-linear regression curve does to accomplish the same thing with less calculations.
July 14th, 2009 on 2:34 am
“To make this make better sense to you, basically EVERY time ANY guild gets a new achievement or kill under what you are suggesting, EVERYONE that has done it have to be recalculated in order to maintain accuracy…”
To be fair, this is currently the case. Since we are not guaranteed to be discovering kills in the order they were killed the ranks have to be reordered every night before calculations are done. There is not really much that can be done to avoid this as long as the calculations depend on the kill order.
To give an idea as to how the mysterious calculations work, the main steps are:
1.) Rank each achievement/kill in order it was killed
2.) Get each guild’s kills/achievements, retrieve boss/achievement worth and apply algorithm based on kill order.
3.) Sum up each guild’s score and store by era and by zone as necessary.
Then various other housekeeping tasks are ran such as ranking guilds per era/zone by score, recording rankings change versus previous, update badges, remove defunct guilds, etc. This process takes about 25 minutes currently.
It would be almost impossible (and silly) to do these calculations real time. And like Wasserjunge said, “the more active calculations and the less you have databased…the slower the site will run and that is not ideal.” There are no calculations being ran during normal use of site. Everything is stored in the database.
Anyways, now that I’ve derailed a bit, I encourage visitors who may disagree with a certain guild’s ranking to post the guild here and your reasoning so that we can hammer out any problems. Don’t forget you are able to view the scoring breakdown of any guild and compare it with any other guild via the link on the guild sidebar.
July 14th, 2009 on 3:34 am
Moving onto your calculations..and…your degree and job are irrelevent to the issue at hand…no offense, it took me 2 seconds to look at your equation and recognize why they aren’t working for you for the range you want it to. Get off the spreadsheet. This doesn’t require any algebra, just a basic understanding of log functions and how to set them up.
Your log root and constants have no basis, so basically, you are using a constant regression curve that is based on no binding information. You are failing to induce bounds to these values. That is the reason that scores go negative…and why they don’t bind where you want them to. If he were to change the way he runs calculations, a similar equation could be inducted. But looking at the source for his page, he is sending data back and forth between the server and an XML database which seems to be opening individual guild files altering necessary information saving the file and sending it back. This is not an ideal setup for applying a constantly variable equation. Your tiering mechanism could work with proper bounds, but again, these calculations are far more intensive than you seem to understand. I would expect with a background in Computer Science that you could see this.
If you want to see the real way to do what you are looking for, there are a few things about log systems that you need to apply in order to do this.
Take the equations:
1-a(logc(xc)-1)
let c=R^((1/(b/a))
where R=total number of guilds that are being considered into the equation
a=ANY VALUE (this will not change anything which you can tell by solving for the value c at any point a, the curve is always the same.
b=the percentage of the curve that you want to be expressed over range (R)
x=the only obvious independant value in this situation which is the rank of the achievement (not rank of the guild, this regression curve is to determine the value of the kill or achievement, not to determine the points the individual gets from it)
What you are suggesting could be accomplished using a variable value “R” (as I said it would be REALLY easy to incorporate what you are suggesting ;p) With that variable value R setting the regression curve for each achievement, the formula is solved, and other factors simply must be maintained.
To use this, a database would need to be incorporated that ranks the major achievements or kills from those with the least guilds completed to most, and those values become “x” in the curve. (0 guilds would have it in a deadlock tie for rank 1 which is ok so long as if there are say 10 that 0 guilds have taken, the 11th that 1 guild ahs taken is still ranked 11, and not fluke-ranked 2 for some reason)
There is no way to use this for minor achievements, as again, you would be creating a penalty for certain guilds that skip them (which is a REALLY big deal if they are skipping them to do hardmodes that achievements may forbid.)
So, let’s say that I am ok with a major achievement being worth anywhere between 5 and 1000 points, so we would want to see .005% as the lower bound of value range. I can set a=1 in this case for ease.
The equation for what an individual guild gets for a major achievement would be as follows…
1000*(2-LOGC(V*C))*(1-.1(LOGD(X*D)-1)
C=(Guilds Completed)^(200) which comes from .005%
D=(Total Ranked Guilds)^(1/7)
V=Value Rank of kill or achievement
X=world order for the individual guild getting the achievement
This took me far less time than you spent on a spreadsheet, and does exactly what you are asking…I told you it would be easy. Understanding the math is not the problem, the amount of calculations and pulling which can go into this do make a lot of difference as you are now stacking two fully variable regression curves and multiplying them. Each value will have to be recalculated each time a new guild does something new.
July 14th, 2009 on 3:46 am
and…sorry..C=(# of total achievements and kills being considered)^200
July 14th, 2009 on 9:07 am
About to leave for work so not a whole lot of time to reply, but I think you’re grossly overestimating how computation-intensive applying this formula would be.
He already has the data that says how many guilds have gotten X achievement or boss kill.
All he would have to do is calculate the point value of each achievement prior to re-calculating rankings.
At <80 achievements and boss kills, this step of calculating rankings would be trivial. It would be quicker than calculating the rankings of a given server, and as Xxav pointed out would NOT affect traffic speeds.
Note that the site is constantly crawling or being forced into an update, but any new achievements discovered is flagged as “NEW” and no point value is assigned until Rankings are calculated every night at 3am.
I only used excel because I was trying to fit a curve to the numbers you provided and I happened to have it open at the time. It took all of 2mins to find a reasonable logarithmic expression to hit your data points.
July 14th, 2009 on 10:36 am
I am done.
I have given the equation to accomplish what you want, it can be applied to this stuff or not. It will make the query, resave data cycle longer for each guild and will make the calculations take significantly longer as you are querying for more data against the server…you can believe me or not, but the site already goes down for a decent period of time.
If there is a way to recode the database to where it is more efficient…that would do great.
Also, because of the way your crawler works Xxav, you should probably rank anyone that gets it the same as others that day. There is no way for you to tell who actually got it before who within that day, so just apply a simple bit to the ranking setup
if(rankdate=rankdate(guild-1) then rank=rank(guild-1) else rank=(actual rank)
I don’t have your actual variables, but you know what I mean…this would fix any issues caused by one guild that actually got the achievement first getting ranked behind a guild simply bc they were added to database later.
July 14th, 2009 on 5:35 pm
Yes, next thing on my list is kills on the same day being ranked the same. Should be done by tonight’s rankings.
July 15th, 2009 on 1:29 am
Greetings,
I have not been posting the progress of my guild for long, since my guild has existed for less than 2 months. I am posting here now, because the current ranking mechanic is dysfunctional and does not actually rank the progress of guilds, but rather ranks which guilds did what first. There also seems to be no mechanic in place for accurately removing guilds who have lost the ability to take down certain mobs. Sure they took down a mob before, but their guild may have gone through major changes and cannot currently take a mob down.
I am in a situation currently where two guilds are ranked higher than my own merely because they existed before my guild did. Therefore the system is not accurately representing the fact that my guild has actually progressed farther.
My suggestion to rectify this problem is to either reduce the decay mechanic to the point that it does not overpower the kill progression mechanic, or change the value of each mob so that it is accurately measuring the difficulty of the kill. For instance, a guild that killed the first two mobs in palace, tombs and ykesha’s should not be ranked higher than a guild that has only killed all five mobs in Tombs. To do so would be to indicate that you think the newb guild who can’t kill Absatalius has progress farther than the guild that is killing Gynok.
I believe a fair representation of the difficulty of the mobs might be to have the most difficult (last mob in Munzok’s) be ranked at 100 and the easiest mobs (first mob in tombs) be ranked at 5 or 10. This way, killing harder mobs awards more points, and more accurately depicts the progression of a guild.
A guild that has cleared all content should not be ranked lower than a guild that has not, but killed some of the content earlier.
July 15th, 2009 on 1:38 am
Thanks for the comment. Most of the focus lately has been on the WoW scoring system. I will take a closer look at EQ and EQ2 in the next few days. I do agree with what you said.
July 15th, 2009 on 3:42 am
I tinkered with the EQ and EQ2 scoring a bit and added an algorithm that will assign max values for a mob based on how many guilds have defeated it. It may need to be tweaked slightly but I think it does correct most of the issues.
July 15th, 2009 on 5:09 am
Will try to get kill dates on the same day to be ranked the same as soon as I can.
July 15th, 2009 on 8:10 pm
Upon closer inspection, it seems like Wasserjunge was trying to incorporate both my idea and his into one formula. Perhaps I wasn’t being very clear.
The formula I was giving was meant to be JUST for the TOTAL point value of a boss or achievement. IE, Currently you have all bosses at 100pts except Algalon, who is at something like 22000 points.
I was suggesting using a logarithmic formula to determine the point value of these bosses based on the number of kills. In this way, the total overhead on preprocessing should be about 80 “Select count(*) from dbname.guildtable where HodirDate != null” SQL statements (one for each achievement/boss), 80 iterations of the logarithmic formula to determine the point value, and about 80 “Update” statements to give the current max value for that boss/achievement. The rest of the current implementation would be unchanged (I’d venture to guess that this step would be 1% of the current 20min update downtime each night).
Wasserjunge’s formula for the percentage (100% for first kill, 30% for 25000′th kill) is great as it is.
It would scale so much that the top guilds who have done hard-modes would still be getting plenty of points. Achievements that are “easy to get”, even ones that are skipped by the highest end guilds (like Unbroken) would be done by enough guilds who have bothered to do it that it won’t put those top guilds at a disadvantage. The top guilds are getting so many points from hard-modes (algalon is currently worth 220x more than Yogg on normal mode) that I can’t see this formula hurting those guilds, if the scaling is set up properly.
Currently about the same number of guilds who have killed Yogg have done Unbroken, which would put them at the same value. If you wanted to apply a multiplier on top of this for achievements that aren’t bosses, such as cutting it in half (which would make Unbroken worth about the same as General Vezax) or 1/4th (which would make Unbroken worth the same as Hodir), I support that.
I think this is necessary because there’s a danger in assigning numbers based on a perceived difficulty, such as saying Firefighter and Knock x3 are the same difficulty. If they were the same exact difficulty, they’d have the same number of guilds defeating them. If more guilds have done Knock x3 than Firefighter, then Firefighter should be worth more than Knock x1+Knock x2+Knock x3.
This also provides a built-in decay. As more guilds kill a given boss, the value of that boss goes down for everybody. The first guild will still get 100% of the points, and the 25000th guild will still get 30% of the points, so the relative difference between these two guilds will remain the same.
And to Wasserjunge, I’m sorry if I frustrated you in any way. You came up with a great system, and I agree with the scaling that you suggest at each tier of difficulty. My only suggestion is that the scaling be based on the actual number of guilds who have killed it at any point in time. Wanting it to be the most accurate for the top <1% of guilds is admirable and I agree completely that it should be done, and your work achieved that. However, it does start to fall apart outside that 1%, and that’s why I’ve been pressing for this.
July 16th, 2009 on 7:33 pm
again, I would ask that you show me an example that I can actually address from your server, where you actually feel that one guild is misranked over another. (Please make sure it is two legit guilds you are comparing, not one that has a few members that might have tripped up the crawler). If you can show me how you feel this system is misrepresenting somehow due to the achievements system, I might understand what your frustration is. However, I have looked over these results for my server, and am more than satisfied with the way it is ranked. Many of the tied guilds are offset by achievements. Because many of those lower guilds only get maybe 30-40 points per new boss anyway, those 5 points here and there and then 25 points for the harder but accessible achievements make a huge difference. Again, I need to see this issue from your perspective…and the only way for me to do that is to see evidence of your complaint.
In getting back to your question of stacking log functions, it is a necessary evil. If you don’t stack them, that means you give everyone the same points. You don’t want to just rank the fight, you have to rank the people taking the fight. Now, if you are suggesting that by simple virtue of the point values being adjusted that simply ranking the fight would add up, I would have to argue completely…because you have no way to differentiate between two guilds who have all the same content. It is as much about kill order as it is what has been killed.
July 18th, 2009 on 4:46 am
Ties should now award the same points.
July 19th, 2009 on 3:08 pm
I can’t give an example at this exact minute because, as I said, I agree with the general numbers you’ve given. The example of inaccuracies doesn’t exist yet. That’s why I tried to curve-fit over your numbers; right now, they’re great.
A month from now, however, I’m guessing we’ll realize that there should be a new tier for interim difficulty (ie, that lots of guilds are taking out knock x3 and heartbreaker, but that still nobody’s done Firefighter). I can’t predict what the exact problem will be, because I can’t see the future. (Note: I’m on a server that has absolutely zero hard modes down on 25man, across every guild. US-Galakrond)
What I can see, however, is that a month from now, the current numbers we have won’t accurately reflect the difficulty of fights. Consider what happened when people learned you could zerg Sarth3D 10man; numbers of guilds who completed it shot up, and its value as an achievement should have gone down, even to those who didn’t zerg it. Proclaiming that you did 3d becomes less valuable for everybody.
—–
Regarding the log function stacking:
I still think you’re not understanding my question. I want to rank the people who have taken the fight.
A*(1 – .1 * logc(c*x) )
A = maximum number of points for achieve or kill
x = world order of the kill
c = log base=4.2890620492
That formula is used for each and every guild, each night during rankings.
—
Now, we want to set A. Currently, I’d imagine that A is being pulled from the DB for each achievement, and we want to keep this as the case for performance reasons. So, before we start looking at each guild, we want to calculate A for every achievement, based on another linear formula. This is a pre-process calculation, based on the number of guilds total that have gotten said achievement.
A=1000*(2-LOGC(V*C))
C=(total # of guilds who completed the achievement)^(200) which comes from .005%
V=Value Rank of kill or achievement
Yes, the final calculation per guild per achievement winds up being:
1000*(2-LOGC(V*C))*(1-.1(LOGD(X*D)-1)
But the first log is only needed to be done once per achievement (simple optimization). The second part is done once per guild per achievement. Hence my argument that it’s not going to add much to the downtime.
I do have a question as far as what V (“Value Rank of Kill or Achievement”) means, though. Is that related to tiered hardmodes?
July 21st, 2009 on 3:52 am
I’d have to say that the decay of worth currently used is so high that early kills can cause a guild to be ranked higher than one which has progressed further. For instance, my guild’s 10-man runs started later than the most progressed on a server. Because of the huge point different from the earlier hard-mode kills, we are ranked behind a guild that is 8/10 hard modes even though we have completed 9/10, and have about the same “pointless” achievements. While I agree that killing order is important, it seems flawed that the points awarded could cause one guild to be ranked lower than a guild that is less progressed simply because some hard-mode bosses were killed earlier.
July 21st, 2009 on 12:38 pm
On the server Gnomeregan, my guild has downed the same bosses as a guild ranked higher than us….and we have downed 2 additional bosses…and we have either the same or more achievements in each category for Ulduar Normal. We have downed the bosses the same week as them or earlier…yet they are still higher. Why is this?
July 21st, 2009 on 3:47 pm
@Solo – What guild?
July 21st, 2009 on 7:35 pm
To answer these questions as easily as possible. Their earlier progress does increase their value of rank on the server. This is by design. When you complete major steps above them, you will begin to see a difference. However, if you have followed this all along you would have seen that it was clearly indicated that this would happen and why it should happen. It should not be drastic if they were not among the higher ranked guilds to take these down first. For example, if they were the 5000th guild and you were the 7000th guild, there would not be a huge difference. Hell, if they were the 5000th guild and you were the 20000th guild, then it is very well possible that getting ahead of them in total kills would put you ahead. However, if they were among the first 1000 and you are around the 20000th, their points SHOULD outweigh yours. Once you down Algalon, you should pull ahead, but you do have to progress to a tier above them…and if they do so close to the same time that you do, they will STILL likely be ranked higher than you and rightfully so.
July 26th, 2009 on 6:09 pm
Related to the values and age of the progression. I was comparing my guild to a few other “rivals” and have noticed that the date they achieved their kills seems a tad skew.
How are these other guilds able to get a kill register as 12+ days ago when they only downed the fight during this week?
Further to this, I have noticed guilds that have kills registering as back in 2008, when they must’ve only just completed the kills in the last week.
Understandably SOME of their members had completed a kill back in 2008 (or 12 days ago for Uld content), but as a guild they’ve only just gotten the kill in the last seven days. And as you mentioned this site is GUILDprogress.com.
Even though this is meant to be a fun comparison between different guilds, it is a little disheartening when another guild can just form up with a few very experienced raiders and come out on top of a guild that has worked hard and battled through content. It’s also a little rough to see your server rank for a kill/achievement slowly slip backwards when other guilds complete the kill/achieve after you, but they are placed ahead because some of their members have an earlier date.
(hope this makes sense)
July 31st, 2009 on 2:15 am
Should less emphasis be placed on kill dates? Maybe change 25,000 world kill to award 50% instead of 30%? Thoughts?
August 4th, 2009 on 1:24 am
Hmmm. I have to admit I haven’t followed how the whole scoring thing is calculated… Frankly it fried my mind…
For us the issue is the actual date that is being registered for a kill.
I notice that over the last fortnight my guild on the “Must deconstruct faster (10man)” achievement has slipped from 30th on the server to 33rd… previously we were 20th…
Some of this ranking slip will no doubt be from guilds updating their profiles after being stale.
This is part of how we measure our performance against other guilds in certain areas.
January 26th, 2010 on 3:36 am
I still can’t understand, how it possible to be in the 44th place when we have 9/12 of ICC10, and all above us have 6/12! Same deal on ICC25 too…
Don’t believe me? Just check EU\The Maelstrom rankings and guild “Usual Suspects”.
March 27th, 2010 on 10:29 am
Same, We have 10/12 down in ICC 10 man on Terenas US, and other people with fewer than 10 are higher ranked than we are…messed up
July 8th, 2010 on 10:45 am
The date a GUILD is credited with downing a boss should be no earlier than the date the 8th (for 10man) or 17th (for 25man) person in the guild downed that boss – not the date of the first person.
That’d prevent re-guilds (where a bunch of other guilds disband and reform one uber guild) from manipulating the ranking charts simply due to the people they’ve invited.
Also, if I happen to pug a raid early on in the tier and beat Boss_X, my guild shouldn’t be credited with the kill on that date after beating it two months later.
February 25th, 2011 on 5:23 am
hi
i need to speaking with a somebody who’s speak french please.
my english is very bad and i can t explain my problem in english.
it s about boss we’ ll down.
thank to answer me please
March 17th, 2011 on 12:05 pm
I certainly understand ranking guilds that get things done earlier higher on the list. But since the amount of credit is so varied the guild “progress” rank often puts guilds with fewer but earlier kills above guilds with more kills. It’s my opinion that any formula that doesn’t rank the quantity of kills first should no longer to claim progress.
So one solution would be to rank the guilds first on the raw number of kills then to break ties you can use the formula you have now.
Finally I’d like to say I have no problem with some kills counting more for everyone, but the progress chart has guilds with 5 kills in front of guilds with the same 5 kills plus an additional kill, which is what doesn’t seem right to me.