Talk:Honor system (pre-2.0 formulas)

The plot linked here is based on ranking data from Weds Dec 14 and Weds Dec 21 2005 across 20 or so servers. I captured and saved the Rank Point scores for the top 1000 players on each server/faction. For each player, I calculated the difference betwwen the two week scores, and factored out the 20% point decay for the week, leaving me with a good idea of how many RP's were earned for the week. I then sorted the data by points earned, from highest to lowest. As was previously found in the references, the top scorer on the server/side always earned 13000 points.

Next I made a few assumptions:
 * I expect the data to follow a percentile ranking. That is, I need to know how many players were ranked in a given week, because I expect the top X percent of those players will earn roughly the same amount of RP for the week on all servers.
 * The honor guide says approximately 1.4% of the PVP ranked players will ever reach Rank 11. Rank 11 requires 45000 RP, or an earning rate of at least 9000 RP/week. So, I normalized the standing (1st, 2nd 3rd...) into a percentile score such that the 9000 RP crossover was at 98.6 percentile.
 * This in turn assumes I have the scores for all the top points earners listed in the overall top 1000 for the server. This is most likely not quite true. (On my server, my weekly rank was 310, but my scoring rank in this data was 277.) So the X-axis scale is skewed where there are missing data points.
 * I also don't consider DK deductions for any new DK's in the week. That will also throw the data off.
 * I also don't consider the level limits of the players. Most of this top-end data is level 60 characters, though.

So basically I can get a feel for how things go, but no hard numbers. There's just too many unknowns in here. Roughly though, it looks like:


 * The top 5% will earn at least 7000 RP/week and reach Rank 9
 * The top 3% will earn at least 8000 RP/week and reach Rank 10
 * The top 1.4% will earn at least 9000 RP/week and reach Rank 11 (given).
 * The top 0.7% will earn at least 10000 RP/week and reach Rank 12
 * The top 0.4% will earn at least 11000 RP/week and reach Rank 13

It's also interesting to note that the score curve for any one server/side is not at all smooth. See that one curve that trails off to 6000 RP for an example. That suggests to me that the weekly score may actually play a hand in how many RP you earn compared to your nearest neighbors in the standings.


 * Here is an ongoing discussion and explanation of how the system works on the VNBoards. A similar analysis to Greenman's later work was done and shows a few additional pieces of the puzzle. http://vnboards.ign.com/wow_general_board/b19789/94440078/p1


 * I'm trying to get registered on the IGN board to comment on your thread. -- Greenman

-- Greenman 10:24, 21 Dec 2005 (EST)

Sample Standings for Dec 20 2005 on 25 servers
This chart shows the actual standing numbers for the week ending Dec 20, 2005. I choose 25 servers, some at random, some because they were new, and some because they were full enough to be selected for character transfers.

This first four columns are the number of players who were at the given Rank at the end of the week.

The next four columns show the number of players earning at least the given number of RP for the week, as compared to the previous week's standings. Note that this data is cumulative, that is, the number earning more than 11k includes the number earning more than 12k.

Server/Fac :  Num. at Rank  Num Earn More Than : 14  13  12  11  12k 11k 10k  9k -   Blackrock/a :   1   3  14  40    7  26  62 127 Blackrock/h :  0   9  20  22    7  25  63 131 Bloodhoof/a :  1   4  11  13    3  12  28  60 Bloodhoof/h :  1   3   7  10    3  11  27  57 Detheroc/a :  0   4   4  10    4  13  29  60 Detheroc/h :  0   4   5   7    3  11  30  70 Draenor/a :  0   5   9  24    5  16  39  87 Draenor/h :  0   1   8  18    3  11  27  63 Eonar/a :  0   4   9  18    4  12  31  68 Eonar/h :  1   0   8  11    2   9  22  46 Frostwolf/a :  0   4  12  20    5  15  39  85 Frostwolf/h :  0   5   9  15    4  13  33  71 Hellscream/a :  2   5   6  10    3  11  27  56 Hellscream/h :  1   2   5   7    3  11  23  51 Icecrown/a :  1   4  10  14    3  10  25  53 Icecrown/h :  0   2   4   8    2   8  19  42 Illidan/a :  0   8  15  26    6  19  48 108 Illidan/h :  1   7  18  22    6  19  47 106 Kalecgos/a :  0   1   6  17    4  12  31  65 Kalecgos/h :  1   4   7  14    4  13  34  71 Kirin Tor/a :  0   0   0   0    2   8  19  34 Kirin Tor/h :  0   0   0   0    2   7  17  33 Lothar/a :  0   1   7  19    4  13  29  64 Lothar/h :  1   2   5   8    3  10  24  50 Maelstorm/a :  0   1   6  12    3  12  29  65 Maelstorm/h :  0   1   5   9    4  14  33  70 Medivh/a :  1   3   6  14    3  11  26  59 Medivh/h :  0   2   3   8    2   8  19  42 Sargeras/a :  1   8  10  21    5  16  40  81 Sargeras/h :  0   6   8  16    4  15  36  70 Skullcrusher/a :  0   8  10  20    5  18  42  92 Skullcrusher/h :  0   6   9  28    5  17  41  87 Spirestone/a :  0   5   9  28    5  16  43  90 Spirestone/h :  1   4  11  20    4  15  35  71 Staghelm/a :  0   1   2  12    3  12  25  56 Staghelm/h :  0   2   2   8    1   7  16  38 Stonemaul/a :  0   5  13  21    5  18  44  94 Stonemaul/h :  1   5   8  15    4  15  37  79 Suramar/a :  0   2  12  16    4  14  36  75 Suramar/h :  1   2   9  12    3  10  24  55 Tichondrius/a :  0   9   9  26    5  19  46  92 Tichondrius/h :  1   5  14  20    6  21  49  96 Uldum/a :  0   6  10  21    4  15  38  83 Uldum/h :  0   2   3  13    3  11  28  62 Ursin/a :  0   4   7  11    4  14  29  63 Ursin/h :  0   3   9  17    4  13  35  74 Uther/a :  0   6   7   9    2  10  25  53 Uther/h :  0   1   6   9    2   8  19  42 Warsong/a :  1   7  15  26    7  24  59 119 Warsong/h :  1   5   9  32    7  21  52 113

--Greenman 10:59, 22 Dec 2005 (EST)

Data from week ending Dec 27, 2005


Same plot as last time, only the ranking page is now listing the top 2000 players instead of the top 1000. That seems to give me better data convergence. This time I tweaked my fit points to match the graphed data, and illustrated the fit with a green line on the plot.

The data looks nearly linear from around 83 to 92 percentile. Exponential above that. Below that it changes slope again and may go back to a very shallow exponential curve? Not really sure so I didn't try to fit it. --Greenman 14:06, 28 Dec 2005 (EST)

DK cost
Look at this data for Whitewidow on Ursin server: http://www.wowguru.com/db/chars/whitewidow-id202808/

Between the weeks of Dec 27 and Jan 4, he earned 19 HK's and 47 DK's. Now 19 HK's is not enough to make the weekly RP award, so we can expect that his honor for that week dropped by the decay limit value of 10%, in addition to the DK discredits that he earned during the week.

Week of Dec 27, RP = 17504 Week of Jan 04, RP = 11524 Assuming 11524 is 90% of his week ending score: 11524 / 0.9 = 12804 So during the week he lost: 17504 - 12804 = 4700 RP Therefore, the penaly for a DK was: 4700 / 47 DKs = 100 RP/DK

-- Greenman 12:57, 11 Jan 2006 (EST)

Another data point, Tabithea of Frostwolf, weeks of Nov 30 and Dec 07: http://www.wowguru.com/db/chars/tabithea-id87425/

Week of Nov 30, RP = 16861 Week of Dec 07, RP = 14635, HK = 1, DK = 6 14635 / 0.9 = 16261   16861 - 16261 = 600 RP    600 / 6 DKs = 100 RP/DK

-- Greenman 13:47, 11 Jan 2006 (EST)

DKs - Alas Poor Yoricks
I have a level 31 paladin with a minimal amount of honor. He collected 41 HK's one week just to make private, then stopped.

I took him out to Silverpine Forest to find some civilians to pick on, to see how fast his RP bar would go down. I took screenshots after each kill. I picked on Rane Yorick (Lev 15) and Quinn Yorick (Lev 14), killing them 4 times each. These are my results:

0 DK  RP bar = 12 pixels 1 DK  RP bar = 11 pixels 2 DK  RP bar =  8 pixels 3 DK  RP bar =  6 pixels 4 DK  RP bar =  4 pixels 5 DK  RP bar =  2 pixels 6+DK  RP bar =  0 pixels

The total RP bar appeared to be 322 pixels wide in my shots. That suggests that:

12 pixels / 322 pixels * 2000 RP = 75 RP at start each kill cost about 1/6 of 75 RP = 12.5 RP per DK it is not possible to be demoted below Private, even with DKs.

-- Greenman 17:27, 11 Jan 2006 (EST)

DK by Level Solved


This graph shows DK penalties amassed by a large number of players, as calculated from the PVP scores posted for the weeks ending Jan 10 and Jan 17, 2006. The data points are plotted as red crosses, and have not been rounded off from the calculation. Note that the data uses the character level at the end of the week, which may be higher than the level at which the DK's were committed. The green line shows a fit of the highest RP penalty by the level of the character. Each segment is linear, with the end points of the segments labelled. This data will be used to complete the DK section on the parent page. -- Greenman 12:50, 19 Jan 2006 (EST)

Decay correction
I have modified the Inspector mod slightly so that it will record the last weeks CP and WS for everyone that I inspect. This lets me run around IF and SW and inspect as many honor-titled players as I can find, and thus record their real WS and CP for the previous week. This information has proven to be very useful.

I found and checked at least 5 cases of people who had not earned enough RP for the week to beat the 20% decay. I've discovered that Blizzard is kinder with the Decay than was previously assumed.

This old formula is wrong: Decay = 0.2 * RPold Delta = Earn - Decay if (Delta < -0.1 * RPold) then Delta = -0.1*RPold if (Delta < -2500) then Delta = -2500

This is the correct formula: Decay = 0.2 * RPold Delta = Earn - Decay if (Delta < 0) then Delta = Delta/2 if (Delta < -2500) then Delta = -2500

For example, consider a player right at rank 7 with 25000 RP. He earns a small amount of RP for the week, not enough to overcome the decay. This table shows the difference between the two formulas for various earning amounts.

Earn  RP(wrong)  RP(right) 0    22500      22500 1000     22500      23000 2000     22500      23500 3000     23000      24000 4000     24000      24500 5000     25000      25000

So the system does reward you for making an effort, even if you are going down.

real decay example
This is the situation that tipped me off to this correction. Using the old formula, I found this discrepancy:

Name            CPs     WS    RP earned (calculated) Shortandmad    83186    88   8661 Mbedo          83028    89   8807 Kaloc          82726    90   8650

So why was Mbedo's apparent earning so much higher than those ranked on either side of him? I looked up his RP totals for this and last weeks: Mbedo (A, Lothar) - Jan 11 to Jan 18 RPold = 44785  RPnew = 44635   Delta = -150  Decay =  8957

So assuming his real RP earning was bracketed by the two on either side of him, I guessed: Earn = 8657 Delta = 8657 - 8957 = -300 and thus Delta = Delta/2 when Delta < 0

--Greenman 11:03, 24 Jan 2006 (EST)

CP to RP formula approximated


It appears that the CP's earned in a week actually do directly determine the RP's earned for that week. I was able to collect enough CP data from the last two weeks to graph up some points, and I noticed that the relationship between CP and RP was a fairly smooth curve. After a bit of fiddling I came up with an approximation function that fits the observations fairly well.

The two graphs linked here show my observations, using data collected on the Alliance side of the Lothar server. The red pluses are the actual data points. The CP were collected in game by inspecting the honor of as many rank-titled people as I could find. The RP earned for the week was calculated from the postings on the Blizzard PVP Ranking pages. Data points for people with DK's earned durning the week were dropped, since those can throw the RP calculation off.

So why is this better than the percentile approximations used before?
 * Exact WS and CP values for each player, instead of guessed WS and no CP in the previous data. Guessed WS were often wildly off for anyone with a WS more than 40 or so.
 * Explains the frequent case where 3 players ranked one after the other (say 12,13,14) can have two very close RP scores and one very different RP score (because it's based directly on how closely together they scored in CPs).
 * Shows that my NR assumptions are way off. Previously I was calculating a NR around 5200 for Lothar/A. With this data, I can see that number is closer to NR=2000. (Highest WS I could find for the two weeks was 1920 and 1979 respectively).
 * This gives us an idea of how the second cutoff works. You get no RP if you have less than 25 HK, and at least on Lothar, you get no RP if your CP is less than around 1000. - This is not true, see below.

Of course, this is just a fit. There is a noticable amount of deviation from the fit that is not explained. This error is on the order of 100 RP so it's still a very good approximation. --Greenman 13:28, 25 Jan 2006 (EST)

Great work. Keep it up. I just had one question. Before when the honor system came out it said that About 5% of the players participating in PvP are expected to have a Rank of 11 or above when the system settles itself after some months. Now they have revised the page to say 1.4%. Does your data show that the 1.4% is accurate? --JaredPWagner 12:05, 01 Feb 2006 (EST)

Hard to say. The big variable there is how many players are active in PVP on each server/side. I can't even estimate that very well without actually being on the server to inspect players for their WS each week. That said, I am on Lothar/A and this week we have 27 players at or above rank 11, out of a pool of at least 2155 ranked players for the week. 27/2155 = 1.25%, looks pretty close.

Let me add that this week I found one player that appears to disprove the statement that there is a second cutoff after they drop everyone with less the 25 HKs. S/he is: So unless the second cutoff dropped everyone with less than 26 HKs, there was no second cutoff, at least on my server (Lothar/A). That WS=2155 is the highest WS I found in my inspections, and thus is the number that I assumed was the size of the PVP player pool for the week. -- Greenman 09:10, 2 Feb 2006 (EST)
 * Frontstab  HK=26,  CP=178,  WS=2155

CP to RP Solved!
This is part of a post on another discussion board, but it explains breifly how I came up with the published CP to RP calculation.

Last night I was playing with more graphs, and thinking about slope between points on (RP vs CP) graphed for the various WS. The graph looked like a bunch of horizontal lines, with a fair bit of scattering from rounding errors. Aha! The slope was constant for clusters of WS values. I played with a few fits, and found that the neighboring lines appear to intersect at 1000 point increments, which conveniently correspond to 1/5th of the total RP needed for each Rank level. I cross checked my other data sets, wrote a new script to work out the fits and published everything this morning.

A few more comments on side topics...

What if there's an agreement such that everyone on the server scores almost exactly 40000 CP for the week?

If my formulas are correct, then they'd still drop the breakpoints in at the expected percentage points. You might end up with something like this: WS=1,  CP=40100,  RP=13000 WS=4,  CP=40074,  break point for 12000 WS=14, CP=40015,  break point for 11000 WS=35, CP=39974,  break point for 10000 and so that number 35 person would find himself still earning about 10000 RP in spite of having 99.68% of the CP of the leader.

Where's the power curve come from?

It looks like it's a natural result of the honor system. There is no way to predict it since you cannot predict the actions of the 2000+ people that define the curve over a week's worth of play time. It would be very interesting to see the RP vs WS curve on a server that follows some sort of voluntary honor cap system.

-- Greenman 14:25, 9 Feb 2006 (EST)

Wonderful work!
I have a question, Greenman.

If there is only 499 plays join the caculation, that means 0.2*499<1, Can the ws=1 player earn 13000?

Most likely. The position is rounded. If there are less than 250 people on a side, then I'm not sure what happens. Presumably the #1 spot always earns 13000, but maybe not? --Greenman 21:55, 6 April 2006 (EDT)

Lower level honor
I plan to use this space to record level-specific honor bonuses in battlegrounds. If you have any data to add, please do. -- Greenman

As you know they are based on honor for 1 kill of rank 1 player of highest level in group.

Actually, I didn't know that. Thanks! -- Greenman

I asked for and got screenshots from players on the official PVP forum. Here's the threads and the results:

http://forums.worldofwarcraft.com/thread.aspx?fn=wow-pvp&t=492020&p=1&tmp=1 http://forums.worldofwarcraft.com/thread.aspx?fn=wow-pvp&t=491234&tmp=1

I'll go make a big pretty table out of this for the main page. PS. Thanks Dru for the bonus honor multiples in AV.

They belong to good old times where 4h AV was worth 10k honor as opposed to current 4k honor for 15h AV... It's only a good idea to monitor all listed things as honor sources both as enemy ones are killed and at game completion. Drundia

Honor for killing rank X
The table was inserted in this document by another author. I personally have no way to test the numbers since I do not play PVP to make 1-on-1 kills. (Read: I suck at duels) A post on the WoW PVP Forum claimed that...

Your contribution points data is incorrect or outdated. Killing a rank 14 yields more than 332 honor points... killing a rank *13* yields more than 332 honor points. -- Xtremist of Mug'thol, 21-Apr-2006

Can someone better than me at duels shed any light on this claim?

--Greenman 07:15, 21 April 2006 (EDT)

The numbers changed in patch 1.8. Updated numbers are coming in. Thanks! --Greenman

Honor for Faction Leaders
Saving this report for future reference:

"Tonight my guild and I killed Thrall with 36 people for a server first, and he gave 488 honor, 1 gold, and no rep. " http://forums.worldofwarcraft.com/thread.aspx?fn=wow-pvp&t=489825&p=1&tmp=1#post489825

Weekly RP Earnings, Breakpoints
I have a strange feeling that the breakpoints aren't really at the percentage of players, but instead using some more complicated formula like BRK[i] = Sum(WS=1,NR)(p[i][WS]*CP[WS]) / Sum(WS=1,NR)(p[i][WS])

which would mean that honor of all ranked players is accounted for when determining honor at breakpoints.

However the p[i][WS] is highest when WS is closest to breakpoint. Since function Honor(Standing) is rather smooth (or whatever it's called in English...) the breakpoints indeed happen to be close to percentages you have provided in table, however they aren't exactly determined this way, or at least I assume so checking some of my records.

Some records from first and second weeks after patch 1.8 from Horde on EU-Sylvanas:

Honor system week 26 (yes, half year of the Honor system) had such standings at brekpoints:

3, 12, 30, 65, 128

(3 got 12k+, 12 got 11k+ (i.e. 9 got 11000-11999) etc)

The next week got a slight change:

4, 13, 30, 67, 135

Did we have some DK farmers in our top? I didn't calculate DK impact back then so that could be the case, but still 67*128/65=131.9, and one percentage can't shange this much really.

The earlier week had a high advantage of WS=1 over WS=2 (WS=2 earned only 12107). So it appears that very high honor of WS=1 actually raised honor at breakpoints.

I removed my own theory "CP location at breakpoint never seems to be exactly in the middle. I would assume WS at breakpoint is a fractional number so if WS at breakpoint would be e.g. 4.5 it would be in the middle, if it would be e.g. 4.2 it would be closer to CP(4) and if it would be e.g. 4.9 it would be closer to CP(5)" since that good old week also failed it quite good (not sure why I put it there in the first place)

Here are breakpoints for both Horde and Alliance on EU-Sylvanas for the maintenance on 19th of April, 2006

Horde:

6, 20, 48, 104, 209, 355, 510, 672

Alliance:

5, 20, 48, 102, 209, 362, 532, 690

Again doesn't appear to be simply at percentages, but still I didn't account for DK, but can account for them as well. There shouldn't be many DK farmers anyway.

But then again percentages seem accurate enough for real situations (i.e. when not "everyone on the server scores almost exactly 40000 CP for the week")

So what I am about. I'm about possibily to calculate highest p[i][WS]'s as used in formula above to refine the breakpoints even though I don't think it will be of any real usefulness anyway.

Drundia 18:32, 25 April 2006 (EDT)

breakpoint reply
Correct me if I am wrong. It appears that you are counting number in each bracket based upon the advancement in the posted weekly standings. (The Top 2000 per server list on the US WoW site.) If that is the case, then it is common for those numbers to be missing people for at least two reasons:


 * DKs, as you suggest, will appear to move people into the wrong brackets.
 * Level limits will also cut-short the advancement of a good PVP player at a lower level.

The only way to be sure of the number of people in the brackets is to inspect everyone in the bracket while in game during the following week. Then you will know their true game-assigned weekly standing. I've been doing this for the last 3 months or so on various US servers. My datasets are available on the web at http://mosa.unity.ncsu.edu/WoW/datasets/

My theory (as posted) is:


 * If we know the total number of ranked players (NR) on a server/side for the current week, then...
 * there exists a table of breakpoint fractions brk(i) such that...
 * the highest WS of a player in bracket 14 is WS = floor( brk(14) * NR + 0.5), in other words, rounded to the nearest integer value.

I can take all of my datasets so far and make a plot, with RP earned on the Y axis, and (WS/NR) on the X-axis. If my theory is correct, then all of the datasets should converge at each 1000 point multiple of RP, and the X-axis point of those convergences should be the breakpoint fractions brk(i).

This link contains a PDF file of the plot I just described. http://mosa.unity.ncsu.edu/WoW/datasets/all.pdf Red plusses are observed datapoints. Green X's are the estimated breakpoints for each dataset.

The behavior is as I expected, some variation inside the brackets due to differences in CP within the brackets, but the breakpoints at the ends of each bracket do converge well at specific fractions of WS/NR.

--Greenman 13:04, 27 April 2006 (EDT)

update accounted for DKs
I'll check all stuff related to DKs and see the actual values for maintenance on 19th April accounting for dishonorable kills. Also note that the top part of table is accurate, but for some reasons it's not equal in top part. Horde should be 100% precise up to WS=60 (unless someone under level 33 made there), or even to standing 63 because that was my own standing and it matches, Alliance should be 100% precise up to WS=101 (this time unless someone under level 32 made there). I'll update my tables accounting for DK and see breakpoints...

Bah, cursed DK farmers, they play hard to make top 10 and also farm DK...

6, 20, 48, 104, 211, 358, 513, ...

6, 20, 48, 105, 210, 366, 533, ...

I still recon some issues here 104-211 is opposed to 105-210. Apparently it's time to start whispering people about their standings...

I think it's also worth noticing that lower level players have harder time making to the top because they earn lower honor for everything they do. Though I admit their presence.

Alterac Valley notes
Going to move the commented AV stuff to this section. Makes more sense to leave it here. Thank you for the work so far on the AV bonuses. Good stuff! -- Greenman 17:40, 2 May 2006 (EDT)


 * * Your team earns 198 CPs each for killing the opponents' Captain (Galvangar or Balinda), and for killing each of the various opponents' Lieutenants found guarding points in enemy territory.
 * * Your team earns 198 CPs each for destroying each tower/bunker.


 * Taken from http://www.wow-europe.com/en/info/basics/battlegrounds/info-alteracvalley.html
 * * Killing the enemy General - 6 Bonus Kills
 * * Killing the enemy Captain - 3 Bonus Kills
 * * Killing enemy Lieutenants/Commanders - 1 Bonus Kill each
 * * Destroying an enemy tower - 3 Bonus Kills each
 * * Keeping your captain alive (until game end) - 3 Bonus Kills
 * * Owning an intact tower (at game end) - 3 Bonus Kills each
 * * Owning a graveyard (at game end) - 1 Bonus Kill each
 * * Owning a Mine (at game end) - 1 Bonus Kill each
 * 1 bonus kill equals 198 CPs.


 * I'm pretty sure this was lowered after written, so this should be all checked, corrected and listed. --Drundia 23:13, 25 April 2006 (EDT) -->


 * I believe the captain bonuses are 1 BHU, or 198 CP each. The rest I am not sure of. I only have one screenshot from an old AV match on an AV weekend. It shows these bonuses awarded at the end: 198, 792, 396, 198, 198, 1584. We won, but at this point I have no clue what graveyards, towers, or bosses were left. I agree these needs to be re-checked and then reformatted to match the new tables below. There's only one level range for AV (51-60) so the table will be simpler. -- Greenman 14:41, 27 April 2006 (EDT)

Also, I'm going to wade through Thott for a bit and see if I can get a count or even a list of all the for-honor NPCs on each side. It would be nice to have that to figure out holiday/non-holiday totals. -- Greenman 17:43, 2 May 2006 (EDT)

There appears to be 1 BHU for summoning ice/firest lord, but I need to check it. Drundia

My end-game info on 3 games:

7 GY, 3 towers, 3 commanders and 2 lieutenants at victory: 4+1+1+1+1 BHU

6 GY, 4 towers, 4 commanders and 4 lieutenants as well as captain at victory: 4+1+1+1+1+1+1

6 GY, 4 towers, 3 commanders and 2 lieutenants at victory: 4+1+1+1+1+1 BHU

Now that I have 1 more game and 3rd one is nearly the same as 1st one I can assume there is 1 BHU coming per undestroyed tower. Comparing 3rd and 2nd I can also estimate there is 1 BHU for alive captain, while graveyards, commanders and lieuetenants yield nothing.

Drundia 05:27, 9 May 2006 (EDT)

I played a losing AV from start to end last night. We lost with 18 BHU total bonus for the match. There were a few events of note.
 * Someone posted to chat that one of our wing commanders made it back alive, we received 1 BHU bonus within a few seconds of that comment. I had no way to verify if that was the cause.
 * Even though we lost, we got an extra 1 BHU bonus when the match ended. The Horde did not assault or capture South Bunker, so I'm guessing we got the bonus for it staying up to the end of the match.
 * Every bonus came in as 1 BHU at a time. We achieved no objectives that were worth more than 1 BHU (but we lost, so we didn't kill the general.)

18 BHU total = --Greenman 08:16, 9 May 2006 (EDT)
 * 10 for all the Lieut's and Commanders (pretty sure we got them all)
 * 4 for all 4 towers (I know we got them all)
 * 1 for Captain (I think we got him, but didn't get a chance to check)
 * 1 for the escaped wing commander
 * 1 for the survived South Bunker
 * 1 unaccounted (maybe other wing commander escaped?)

I'm confirming 1 for rescuing wing commander. One unaccounted BHU seems to be simply for completion. Granted to both teams. Updating article. --Drundia 19:51, 10 May 2006 (EDT)


 * If that is true, then we did not get 1 BHU for the surviving south tower. There was only 1 extra BHU given at the close of the match, not 2 (1 for tower + 1 for finish). -- Greenman 23:20, 10 May 2006 (EDT)

I'll need to check it more. Bonus honor for surviving things may only be granted to winning team? Did you summon Ivus maybe? I'm still missing 1 BHU for winning team. --Drundia 16:31, 11 May 2006 (EDT)

Killing a wing rider = 0 bonus honor. Just watched one die with no bonus. --Greenman 17:39, 28 June 2006 (EDT)

25 HK rule changed
I noticed during my weekly inspections this afternoon that I found at least 2 different characters who received a weekly standing and earned RP for less than 25 HK in the previous week. I checked my inspection logs and found many more cases of people getting WS>0 for 15 or more HK, and none who got WS=0 for 15-25 HK. This appears to have started the week after the 1.10 patch came out (after 28-Mar-2006). I will update the main page with this new information. -- Greenman 18:12, 3 May 2006 (EDT)

Diminishing returns/Weird issue
There are some specualtions that diminishing returns for CP/kill within 24h period aren't present in battlegrounds.

However, my experience shows that honor acts weird still.

My main (60) still tends to get more honor than ever reported to combat log, which suggests that there are some hidden bonus honor given.

On the other hand having some pvp fun at level 19 showed that I got less honor than total reported to combat log, which suggests that diminishing returns are actually in place.

This seems to require a good research.

Any nice information anyone can provide will be appreciated.

Drundia 18:20, 4 May 2006 (EDT)

I enabled my combat logging and ran WSG on two of my characters last night. The logs and results can be found here... http://mosa.unity.ncsu.edu/WoW/datasets/honorscores.txt Neither character had played PVP in the last 2 days before that, so I know there were no other BGs or HKs clouding the totals.

The honor for my first alt, Gair, agreed exactly with the expected total, assuming diminishing returns is working as stated by Blizzard.

The honor for my main, Greenman, fell between the DR and Non-DR totals. With DR, I expected to get 909-910 honor (depending on rounding). Without DR, I expected to get 1061 honor. My honor tab said 967 honor total.

So I agree we're missing part of the picture here. I don't believe that DR was removed from BGs. (I've changed your comment on the main page to reflect that.) It looks like it's either not working as reported, or there is some other bonus we've missed.

--Greenman 21:38, 7 May 2006 (EDT)

Another alt, Whill, with 41 HKs in a single WSG. (logs added to same page linked in previous comment) --Greenman 08:10, 8 May 2006 (EDT)
 * expected (No DR) = 1491
 * expected (DR) = 807
 * reported honor = 860

I'm still wondering how comes I had one day with extremely high honor, which was like:

HonorReported = 2 * BonusHonor > BonusHonor + 2*HKHonor

Too bad I don't have log for that day any more.

I'll generate some data with greater share of bonus honor, we'll see what we can dig from there.

Drundia 22:02, 8 May 2006 (EDT)

Another result for Greenman, losing AV match. My total showed 14 HKs, but my Combat log only recorded 13 of them. Total bonus honor was 18 BHU = 3564. Now I'll never know how much that missing HK was worth, but it still looks like my 10% of HK honor theory is on track. Copy of logs added to my page linked earlier.
 * Honor (NoDR) = 3992 (plus missing HK)
 * Honor (DR) = 3955 (plus missing HK)
 * Honor (Reported) = 3977

--Greenman 12:56, 9 May 2006 (EDT)

Since we had 100% DR accuracy when noone was killed more than twice a better theory is that kills past 2nd yield more honor than advertised.

With my main (who among other things tends to PVP around daily honor update) my yesterday's records (I had to guess when yesterday started for honor system, there is 30 minute uncertainty), so here are my calculations with both possibilities of start. Luckily over this 30 minute period I didn't get a single HK (but did get some honor), so I managed to determine end of day by matching number of kills.

earliest start:

DR=0: 12096 + 8153 = 20249; count = 123

DR=1: 12096 + 6135 = 18231; count = 123

latest start:

DR=0: 11500 + 8153 = 19653; count = 123

DR=1: 11500 + 6135 = 17635; count = 123

Honor tab: 19524

Doesn't appear to be 10%. The start occured sometime in-between, but even with earliest start we have (19524-18231)/6135 = 21%. The problem is that they possibly don't update tab at the same time they update honor, so it can cause some troubles and I'll need to do more checks which wouldn't have any honor gains close to daily update.

So overall 10%, but not 10%. Guess we need to figure correct coefficients for kills past 2nd. I have enough of math knoledge to do it, but don't have enough data yet.

It was suggested on the PVP Forum by Aiku of Zuluhed that DR's might be applied in order of descending kill value, rather than in time order. I took this suggestion and re-ran the totals for my first Greenman/WSG run in the log linked above. This is my result: 1061.0 = total (no DR) 909.5 = total (DR, time order) 976.0 = total (DR, best kill order) 967   = total (reported on honor tab)

So it's a better fit, but still not quite right. Even if you round all the fractions down to 0, you still get 971 not 967. I'll probably re-check my other runs to see how they work out.

--Greenman 09:58, 1 June 2006 (EDT)

Overpopulated servers
Any chance that this kind of servers can have Rank 15 bracket? What was the highest recorded estimated NR so far? --Drundia 17:19, 26 May 2006 (EDT)

http://mosa.unity.ncsu.edu/WoW/datasets/Report-BleedingHollowA-2006-02-22.txt NR=3113, Bleeding Hollow/A, Feb 22, 2006.

Blizzard supposedly used to reserve rank 15 for the highest scoring player of the week. They were named City Protector. That rank was removed.

--Greenman 23:22, 26 May 2006 (EDT)

That is supposedly not so much of a rank but instead a special status. The question I meant is if by any chance there is a possibility for a standing 1 to earn 14k in one week. We may never know it though until Blizzard raises population caps. --Drundia 18:56, 27 May 2006 (EDT)

I doubt it. Given the quote I gave above (highest NR = 3113), the breakpoint for rank 15 would have to be less than 1/3113th or 0.032% of the population.

--Greenman 14:16, 30 May 2006 (EDT)

I've now recorded 3 server/sides with 8 people in the #14 bracket. (so NR near 4000). None of them had over 13000 RP earned.

-- Greenman 10:08, 1 June 2006 (EDT)

I'm wondering if you've talked with Kuroma of the Daggerspine Server Alliance Side. He has been rank 14 sence October 2005 and currently has 64,432 total RP. If anyone has data on earning RP at the highest rank it would be him. --Bleatingsheep 11:25, 05 July 2006 (MST)

Talked to him? no. Watched him? yes. His advancement has never exceeded 13000 RP earned per week. -- Greenman 12:42, 25 August 2006 (EDT)

Patch 1.12 breakpoints
As promised in the patch notes, the breakpoint curve was changed in the 1.12 patch. Based on the assumption that Lothar/A has around NR=2200 after the patch, I found these rough estimates:


 * 11 in bracket 14, 11/2200 = 0.005
 * 28 in brackets 13-14, 28/2200 = 0.013
 * 65 in brackets 12-14, 65/2200 = 0.030
 * 110 in brackets 11-14, 110/2200 = 0.050
 * 178 in brackets 10-14, 178/2200 = 0.081

Best I can do for now. I've lost interest in PVP so I don't know if/when I'll get around to collecting all the data I had before when I did the first derivations. -- Greenman 13:34, 31 August 2006 (EDT)

Turns out the NR for Lothar/A jumped to at least 3500 this week. So those estimates are way off. I just updated the main page with a better set of data. --Greenman 10:00, 1 September 2006 (EDT)

Does anyone have any updated information on the new 1.12 break points? --Shurafa 19:00, 11 September 2006 (EDT) Well looks like you were pretty close on the break points for the upper ranks. However to my suprise the BIG changes were made at the LOWER break points. Im showing that break 8 went from about 7% to 20%!!!!

Can anyone else verify this?

Update... I have verified over the past few weeks that the upper ranks have been changed very slighly. I have been saving the data to an excel spreadsheet but do not know how to post it to this site. Anyone who can host this let me know as I think it will be a great asset to the community. Well till the xpac comes out. --Shurafa 11:00, 09 October 2006 (GMT)



Has anyone done a reevaluation of this graph since the changes in path 1.11? Blizzard stated that they were 'changing the honor curve' to allow some players to rank more easily, so I'm wondering if this changed the graph much, and if someone could check some more recent data and see if it still matches up.* -- tensafefrogs 11 Oct 2006

Yes there has been some changes in the upper brackets. I have a ton of data confirming this. --Shurafa 30 Oct 2006

= Decay Formula Clarification =

I'd like to suggest an update to the formula for decay. This is in reference to where it states if your decay is greater then your earn, it cuts the decay in half. As a slight clarification, it actually seems to work like this:

If WeeklyDecay > WeeklyEarn then FinalWeeklyDecay = (WeeklyDecay - WeeklyEarn) / 2

So it actually subtracts any earnings from the decay and then cuts that remainder in half. My "real world" example as follows:

PreviousRP = 40940 NewRP     = 39410 WeeklyEarn = 5128
 * Facts:

PreviousRP = 40940 WeeklyEarn = 5128 WeeklyDecay = 40940 * 0.2 = 8188
 * So to get the final number of 39410 with those facts:

NewRP = 40940 - 8188 + 5128 = 37880
 * If we assume no halfing of decay:

NewRP = 40940 - 4094 + 5128 = 41974
 * If we assume decay is cut in half and then earn added:

NewRP = 40940 - (8188-5128)/2 = 39410 (!!)
 * So, neither of these methods match our 39410. So we try our new formula...

--Wickiwawa 15:15, 24 October 2006 (EDT)

There is no decay, there is only averaging.

The formulas listed there are mathematically the correct way to look at it.

Otherwise, you are correct... it is the net adjustment that is cut in half, not the decay alone. Which is what the article already says, if you'll go back and re-read it.

But, if your weekly adjustment (Earning - Decay) negative, that adjustment is cut in half.

-- Greenman 10:15, 21 November 2006 (EST)