Completed SOTK 1,2,3, data inside (Wyrmbane, 7/14/03)

Timeless Saves - Herein is stored lore, tips, and sage advice for the Paladin.

Moderator: Paladin Mods

Completed SOTK 1,2,3, data inside (Wyrmbane, 7/14/03)

Postby Valya Wanderfoot » Sun Sep 24, 2006 12:53 pm

wyrmbane
Dwarven Smith
Posts: 699
(7/14/03 12:25 am)
| Edit | Del All Completed - SoTK 1, 2, 3 - information inside

*****
EDITED 7-17-2003, 11:15 PM

Added additional SOTK 1 data

*****

Ok. I had looked at the data and thought about it. The rough numbers showed a % gain in DPS that seemed extraordinary for 3 aa points. I thought about Caladel's comment on weapon delays too short, etc.

After I got home, I pulled out my 16 MB log file prior to buying SoTK level 1. It is all standard fighting, various weapons, and various hastes. The slowest weapon I use is Windblade, Delay 44. The best haste I would have had, given the people I hunt with, is 115%. The effective delay is 2 seconds. Not perfect, but the best I can do now.

So. I used "eqstrip" from Caladel, and ripped out all SLASH attacks in this log. I then ran it through my double counter macro. I removed all lag effects, any time stamp with more than 3 events in a row. I also removed anythign that was not my attacks....misc information.

Based on this Data, the following Numbers showed up:

Single hits - 13,187 - or 41.72%
Double hits - 18,425 - or 58.28%
Triple hits - 414 - These are in error due to latency, deleted
Total hits - 31,612

So. For the Baseline, based on this large sample, I was incorrect in my statement that I knew I doubled 55% of the time. I thought I knew, but I was flat wrong.

Knowing this for a baseline, and assuming with 31,000+ hits, it should be fairly representative...I think we can go forward.

Ok, At some point during the day, I crashed to server select. I logged in and got some more parsing done. Here are the results on mr Harnoff Splitrock.

With SOTK Level 1
Singles -- 5092 - 39.83%
Doubles - 7542 - 58.99%
Triples -- 151 - 1.18%
Total attacks - 12785

More data would refine the numbers more than they are now.

****
edit from 7/17/2003 here
I parsed on Harnoff for another night and most of the day last night. Here is the new composite of the previous + new data collected.

SOTK level 1
Singles -- 10382 -- 39.8419%
Doubles - 15362 -- 58.9531%
Triples -- 314 ---- 1.205%
Total --- 26058

I bought SOTK level 2 tonight. I am attacking our friend as I type. I should have prelim data tomorrow, but will add additional attacks to it nightly. Would like to get anothe 20-30k parses on level 2, before I buy level 3.

*******

The tricky part is going to be Knights advantage, since SoTK adds to doubles and triples, we need to understand how they work so we know how much KA does independent of SOTK so we can determine which is the best % increase per aa spent.

More later.

wyrmbane

Edited by: wyrmbane at: 7/30/03 6:07 pm

**************************************************************************************

Aonanana
Noble Squire
Posts: 139
(7/14/03 6:27 am)
| Edit | Del Re: Placeholder SoTK level 1 - What it does

All it does is Add a chance for you to have another attack.. doesnt have to be a triple or even a double.


Aonanana's Profile
Mythic Legion

**************************************************************************************

Romidar
Knight of the Truthbringer
and PoN Admin
Posts: 2037
(7/14/03 7:35 am)
| Edit | Del
ezSupporter
Re: Placeholder SoTK level 1 - What it does

Quote:All it does is Add a chance for you to have another attack.. doesnt have to be a triple or even a double.



If I'm reading your statement correctly, you're claiming that you can get an extra attack basically between delay rounds. No one has claimed this previously and there's certainly been no parse data to demonstrate it. Can you explain how you came up with this theory?

An "extra attack" is read by most of us to mean that "when you get an attack, you might also get another attack." Otherwise, they'd have to put in code to just insert a random swing here and there whenever you have autoattack turned on. While that's possible, it seems less likely since the melee engine is built around the delay timer.

Another alternative is that every time you swing your weapon, it checks to see if you will swing it again based on the skill. I'll be interested to see the results of the parse. Smile


**************************************************************************************

HemingwayTZ
Noble Squire
Posts: 65
(7/14/03 10:04 am)
| Edit | Del Re: Placeholder SoTK level 1 - What it does

Wow, very cool of you to parse that--very interesting--seems to be a very viable skill!

**************************************************************************************


Caladel
Lord Crusader
Posts: 2024
(7/14/03 1:03 pm)
| Edit | Del
ezSupporter
Re: Placeholder SoTK level 1 - What it does

Just curious whats the setup of your parses? Normal fighting?

Fast weapons or normal fighting with ripostes are gonna skew any data ya get.

There are 2 invuln vendors that are perfect for parsing a swing rate increase against. One of these is in kaladim and the other is in qeynos, they are the high quality ore vendors that sell a limited quantity of ore. Really need a slow 2hander as well--i reccomend 30 hasted delay or greater--so any internet lag doesnt cause rounds to run into eachother.

Since you are going through the trouble of doing this, just like to see clean data on it so we can really know. I dont consider normal combat data for this one clean enough to show whats really happening.

Can just go afk while you sleep with your character swinging away at 1 of those 2 npcs and get a nice large sample with no ripostes.


**************************************************************************************

wyrmbane
Dwarven Smith
Posts: 703
(7/14/03 1:11 pm)
| Edit | Del Re: Placeholder SoTK level 1 - What it does
All data was from standard fighting. I use a windblade normally, with 94% haste, so 22.68 delay for the tests.

I didn't think it would honestly matter if it was from normal fighting or specific test runs, since I am uninterested in stats, buffs, or any other fact.

Here is how I checked the data...I know you had a parser for this, but I don't have it...so I made my own.

My code does the following - Using excel, because I don't have a copy of VB, except through Office.

--Code removes everything but You hit, You miss.
--Code Checks for Glitches due to Lag - By this, I check the current line time stamp, and the next 4 lines. If the lines all have the same time stamp, there was a write error / lag, so I remove all of the offending lines. The only thing I cannot rid is if a time stamp error has 3 lines...rare, usually it is more
--Removes all Ripostes lines from the log

From there, I go through and check the time stamp of attacks. If the current line is the same for 2 attacks I count it, if it is the same for 3, I count as a triple.

If you want to send me your double / triple parser I will use that instead. Mine works, just takes longer since i wasn't interested in optimizing it for speed.

As for the Invul PC test. I can do that tonight. I will swing by kaladim and find the high quality Ore vendor, and go at him with windblade with no haste item on. Didn't know it was possible prior to SoTK1, or I would have done a night of parsing him without the skill.

wyrmbane


**************************************************************************************

Rendael Brigh
Blessed Cavalier
Posts: 946
(7/14/03 1:38 pm)
| Edit | Del Re: Placeholder SoTK level 1 - What it does

You can't attack them.

Rendael Brigh
Crusader of Truth, 65th rank
Avengers Federation, Tholuxe Paells
Master Smith 250
Master Brewer 200
My Magelo character profile
"I was a young man with unformed ideas. I threw out queries, suggestions, wondering all the time over everything; and to my astonishment the ideas took like wildfire. People made a religion of them." - Charles Darwin

**************************************************************************************

Caladel
Lord Crusader
Posts: 2027
(7/14/03 1:40 pm)
| Edit | Del
ezSupporter
Re: Placeholder SoTK level 1 - What it does

You could 6 months ago when i ran tests, all swings get 'your target is invulnurable' but it doesnt matter, you still get doubles and tripples etc, you can very easily just count swings on them without anything else screwing up the data, or having to worry if your character is gonna die from an npc hitting you.

**************************************************************************************


Abbul Spiritblade
Noble Squire
Posts: 30
(7/14/03 9:50 pm)
| Edit | Del Re: Placeholder SoTK level 1 - What it does

It still works...in fact I just got done from him.
*Edit--The invlunerable npc in the mines of Kaladim.

Edited by: Abbul Spiritblade at: 7/15/03 1:38 am

**************************************************************************************

Amanensia
The Purple Paladin
Posts: 309
(7/16/03 2:06 am)
| Edit | Del Re: SoTK level 1 - Clean Data - 7/15

Apparent rate of triples due to latency without SotK: 1.31%

Apparent rate of triples with SotK1: 1.18%

The percentages you give for your two sets of numbers are a little misleading perhaps, as one excludes all triples whereas the other excludes none. Assuming your latency issues are pretty constant, you should probably exclude a similar proportion from the SotK1 sample. If you do this, there is no evidence, based on these samples, to suggest that SotK1 provides a chance to triple attack, or even provides an increased chance to double attack. I guess this is probably down a combination of the relatively small sample size and the likelihood that level 1 of SotK is of minimal benefit. Will be very interesting to see either larger sample sizes or parses of SotK2 and 3.

Regards,

Amanensia


**************************************************************************************

wyrmbane
Dwarven Smith
Posts: 706
(7/16/03 10:03 am)
| Edit | Del Re: SoTK level 1 - Clean Data - 7/15

I don't disagree with what you state Amanesia. There is a notable, and argueably significant, difference between the baseline and the SoTK level 1 parses.

It goes back to the suggestion Caladel made. Namely, that one was done under a controlled situation to reduce those effects, and the other was not.

Specifically, the baseline was compiled from the only data I had available pre SOTK 1 purchase. Fighting in normal groups in zones against experience stuff. Largely it was done in the Plane of Valor. There are 20-60 people in the zone on a typical night. My haste ranges from 34% (worn belt) to 115% with enchanter haste + bard song. The effective delay with windblade varies in these cases from 20 to 31.

Additionally, when fighting in normal groups, I have Spell effects turned on, am in a zone with more people, and a lot more is happening with OOC's, auction, tells, group chat.

The parse of just SOTK 1 was in a relatively controlled situation. I was in North Kaladim in the mines. Typically there will be 1-10 people in the zone. Most using the bank, tradeskilling, or working quests. Rarely will anyone go into the mines of North Kaladim. Only reasons are to buy supplies, or do 1 of 2 quests.

Also, when I did the test, I removed all information that was unnecessary, including, Guildchat, OOC, Auction, Group, spell effects, and sumarily removed all buffs so my weapon delay was steady at 31.

The result is that while in kaladim there is less opportunity for there to be latency issues. It can happen. It will happen. By and large it will occur on far less occasions that it does in the first case. In the first case, 2 seconds of lag could cause the problem. In the second, it has to be 3 seconds long. It also has less opportunity as I removed a majority of the causes.

One way to get a better baseline is for another lvl 65 paladin to go and attack the same mob in the same zone, under similar conditions (everything turned off, delay 30 weapon or slower).

The only other option I have to additionally mitigate it from the SOTK1 parse is to pick up a weapon with a delay around 50, and do another parse. This will increase the time between time stamps and further reduce those same chances.

If I could un-buy my SOTK level 1, I would do another parse under the same conditions to obtain a baseline. It just isn't an option, and I found out about the test case mob too late.

I can also do further testing to increase the data in the SoTK 1 parse. If I am to do this, and it seems valuable to the discussion, I have to do it tonight. I will be able to buy SOTK level 2 tonight, so I would hold off that purchase until I had more data.

wyrmbane


**************************************************************************************

Romidar
Knight of the Truthbringer
and PoN Admin
Posts: 2060
(7/16/03 12:57 pm)
| Edit | Del
ezSupporter
Re: SoTK level 1 - Clean Data - 7/15

Would be nice to have a 65 paladin with the skills do the same test. Unfortunately, I also have the first level of both SotK and KA. :\

In any case, thanks for doing the work - it's not always easy. I look forward to seeing your parse of the 2nd level against that mob. Smile


**************************************************************************************

Amanensia
The Purple Paladin
Posts: 311
(7/16/03 1:00 pm)
| Edit | Del Re: SoTK level 1 - Clean Data - 7/15

I have no SotK. I would be happy to provide some data, but it will be some time before I buy SotK I think so probably better if I wait until just before I buy it to eliminate gear differences, minor as they may be.


**************************************************************************************

Romidar
Knight of the Truthbringer
and PoN Admin
Posts: 2062
(7/16/03 2:59 pm)
| Edit | Del
ezSupporter
Re: SoTK level 1 - Clean Data - 7/15

Quote:I have no SotK. I would be happy to provide some data, but it will be some time before I buy SotK I think so probably better if I wait until just before I buy it to eliminate gear differences, minor as they may be.



While the absolute best test would be within subjects, if you can run the exact same test we should be able to get a baseline. (In fact, I'd recommend removing one's haste item for this test to exaggerate the delay as much as possible.)

It's possible that haste, weapon delay, AC, HP, whatever could effect double attack rate, but we've no reason to believe it to be the case that anything other than Double Attack skill (and AA's, of course) would have a consistent effect on it. With a large sample, they should be randomized out as "error."

If I remember correctly, the old formula we used for double attack was:

(Double Attack Skill + Level)/500

It may very well be the case that they tweaked it, though, over the last 3 years. At 50th level, that gave us a 50% chance of double attacking. It seems like that was verified way back then.

At 65th level, that would be:

(235 + 65)/500 = 300/500 = 60%

That looks VERY close to numbers shown by Wyrmbane in the post here. Even more interesting, is that if one assumes that the skills allow you to score triple attacks, then you might see double attacks dropping a bit but single attacks staying about the same - also similar to the parse shown above. (Of course it could also be just noice.)

In any case, I'd love to see that parse Amanensia - would be very nice and clean if the double attack rate came out to 60%. Smile


**************************************************************************************

Merts
Wandering Knight
Posts: 324
(7/16/03 3:27 pm)
| Edit | Del Re: SoTK level 1 - Clean Data - 7/15

Just wanted to add, you can also attack the chest in Nathyn's house in katta, it moves around but you can attack it into a corner or something and try then.

Just in case that helps anyone... it's an invulnerable NPC. Doesnt riposte or attack back or dodge or parry or anything.

Duke Mertine Soulcleanser
Level 65 Human Lord Protector of Mithaniel Marr
Unspoken Legacy
Drinal


**************************************************************************************

wyrmbane
Dwarven Smith
Posts: 708
(7/16/03 8:35 pm)
| Edit | Del Re: SoTK level 1 - Clean Data - 7/15

I decided to look at the data I parsed for my Baseline tonight. As described above, I use an excel macro to parse the information. Latency will correct the data, so I have a check for this, and delete any data where more than 4 lines have the same timestamp. This is the best I can do.

Excel has a limitation of 64,000 lines. Because my data file was 16 MB and 225,000 lines long, I wanted to cut out the junk. To do this, I used EQstrip as I mentioned. The command line I used was:
eqstrip eqlog_wyrmbane_preSOTK slash

The "diff.txt" file output from this contained all my attacks that were slashes. Including, unfortunately the ripostes on mobs. Oops, this is a different way to get the same "anomoly" I had when I first wrote my macro.

So, I created a wordlist.txt file with two lines
Slash
riposte

When I ran this:
eqstrip eqlog_wymbane_presotk.txt -w wordlist.txt

This now gave me all of my slash attacks and let me know when I would riposte the mob.

I took this, and ran it through my Macro, the results were:
singles - 13,527 -- 41.73%
doubles - 18,878 - 58.23%
triples - 12 ------- 0.04%

So, I still have latency issues I cannot remove. However, the magnitude of them went from 1.31% of the total down to a 0.04%.

Human, the greatest failing in everything.

For those interested. My Excel macro code is blow. Just cut and paste into Excel (I use 2000).
On sheet 1, Cell A-1, this should be "filename"
on sheet 1, cell A-2, this should be "path"
On sheet 1, Cell B1, input the filename with .txt to import
on sheet 2, cell B2, input the full path, C:\ etc

Tools -> macros -> import_file to import
Tools -> macros -> clean data

That will run the clean_data, Latency, and count macros in succession. It is best to have your import file under 64,000 lines to make it all fit. But this will clean up whatever is in the file and give you a list when complete of the %'s for singles, doubles, triples.

Wyrmbane

Cut Below the Asterisks, and past into Excel to use.
*******************
Sub Import_file()
'
' Import_file Macro
' Macro recorded 7/16/2003 by Comp1
'

'
Dim path As String, filenm As String, importStr As String

filenm = Cells(1, 2).Value
path = Cells(2, 2).Value
ChDir path
importStr = path & "\" & filenm
Workbooks.OpenText Filename:= _
importStr, Origin:=xlWindows, StartRow:=1, DataType:=xlFixedWidth, FieldInfo:= _
Array(Array(0, 9), Array(5, 3), Array(11, 9), Array(12, 2), Array(20, 9), Array(26, 2))

End Sub

Sub Clean_Data()

Dim i As Long, rng As String
Dim spaces As Integer

i = 3
Cells(i, 1).Activate
Do
' Any riposte done by you is recorded on the line above the riposte message
If Right(Cells(i, 3), Cool = "riposte!" Then
rng = i & ":" & i
Rows(i).Select
Application.CutCopyMode = False
Selection.Delete Shift:=xlUp
i = i - 1
' The above 5 lines delete the riposte message, and moves up 1
rng = i & ":" & i
Rows(i).Select
Application.CutCopyMode = False
Selection.Delete Shift:=xlUp
i = i - 1
' The above 5 lines delete the actual riposte line, and moves up 1
Else
Select Case Left(Cells(i, 3), 10)
Case Is = "You slash ":
spaces = 0
Case Is = "You crush ":
spaces = 0
Case Is = "You pierce":
spaces = 0
Case Is = "You try to":
spaces = 0
Case Is = "":
spaces = spaces + 1
rng = i & ":" & i
Rows(i).Select
Application.CutCopyMode = False
Selection.Delete Shift:=xlUp
i = i - 1
' This deletes the Blank line
Case Else:
rng = i & ":" & i
Rows(i).Select
Application.CutCopyMode = False
Selection.Delete Shift:=xlUp
i = i - 1
' This deletes everything except Slash, crush, pierce, misses
End Select
End If
i = i + 1
Loop Until spaces > 3
Latency
End Sub

Sub Latency()

Dim i As Long, rng As String
Dim endloop As Boolean
Dim loopcnt, j As Integer

' Since we are looking for any time the computer records the same time stamp, but must
' preserve any triples, we search for any occurences with the time stamp on 4 consecutive
' lines. In those cases, the code counts the number of occurences, and deletes all of the
' consecutive lines of the same stamp.
'
' Note, a latency where 1 single + 1 double occures on the same time stamp will be treated
' as a double. There is no viable way to eliminate these rare occurances
i = 3
Cells(i, 1).Activate
Do Until Cells(i + 1, 2) = ""
If Cells(i, 2) = Cells(i + 4, 2) Then
endloop = False
j = 0
Do
j = j + 1
If Cells(j + i, 2) = Cells(i, 2) Then
Else
endloop = True
End If
Loop Until endloop
For loopcnt = 1 To j
rng = i & ":" & i
Rows(i).Select
Application.CutCopyMode = False
Selection.Delete Shift:=xlUp
Next
End If
i = i + 1
Loop
count_types
End Sub

Sub count_types()
Dim i As Long, rng As String
Dim Sgl As Long, Dbl As Long, Tpl As Long

' The code below goes through the log file line by line. It compares the current date stamp
' to the date stamp of the next line, based on the seconds. If the seconds are identical
' the code knows it is a triple or double. It looks at current + 2 and checks the time stamp.
' If current + 2 = true, then it adds to the triples
' If current + 1 = true, then it adds to the doubles
' Otherwise, it is a Single
i = 3
Sgl = 0
Dbl = 0
Tpl = 0
Cells(i, 1).Activate
Do Until Cells(i + 1, 2) = ""
If Cells(i + 1, 2) = Cells(i, 2) Then
If Cells(i + 2, 2) = "" Then
Else
If Cells(i + 2, 2) = Cells(i, 2) Then
Tpl = Tpl + 1
i = i + 2
Else
Dbl = Dbl + 1
i = i + 1
End If
End If
Else
Sgl = Sgl + 1
End If
i = i + 1
Loop
Cells(1, 5).Activate
Cells(1, 5) = "Singles"
Cells(1, 6) = Sgl
Cells(2, 5) = "Doubles"
Cells(2, 6) = Dbl
Cells(3, 5) = "Triples"
Cells(3, 6) = Tpl
End Sub

Edited by: Romidar at: 9/25/03 5:50 pm

**************************************************************************************

Medion Palemoon
Noble Squire
Posts: 41
(7/22/03 10:17 am)
| Edit | Del Re: SoTK level 1 - Clean Data - 7/15

Romidar - SotK has a chance to add an attack per round. You read that statement wrong, and this is clearly stated in the AA description.

It doesn't add an attack between rounds it just has a chance of adding an additional slash during your round, thus, turning a single into a double or a double into a triple.

Also, if you have SotK1 and KA1, you can still parse the difference between that and not having it. Just parse the double/triple rate with a 1H, as they are not affected by these AAs, and compare to parses with a 2H.

wyrmbane
Dwarven Smith
Posts: 720
(7/22/03 1:03 pm)
| Edit | Del Re: SoTK level 1 - More data on SOTK 1 added - Sotk 2 soon
Medion,

Thanks for the reminder that SOTK is 2 handed weapons only. I had totally missed that simple fact when attempting to improve the quality of the baseline parse.

Once I get my monitor fixed, I can get my old Wurmslayer from an alt, and use that to help establish a consistent baseline against the same mob I am using for the comparative improvements.

wyrmbane


**************************************************************************************

Romidar
Knight of the Truthbringer
and PoN Admin
Posts: 2104
(7/22/03 2:55 pm)
| Edit | Del
ezSupporter
Re: SoTK level 1 - Clean Data - 7/15

Quote:Romidar - SotK has a chance to add an attack per round. You read that statement wrong, and this is clearly stated in the AA description.



I think you didn't read what I wrote correctly, actually. I was responding the following statement:

Quote:All it does is Add a chance for you to have another attack.. doesnt have to be a triple or even a double.



My response was:

Quote:If I'm reading your statement correctly, you're claiming that you can get an extra attack basically between delay rounds. No one has claimed this previously and there's certainly been no parse data to demonstrate it. Can you explain how you came up with this theory?



The statement that it "doesn't have to be a triple or even a double" logically implies that it's a separate attack you wouldn't otherwise have gotten, which is why I took issue with the statement. That is, if it "has a chance to add an attack per round" then it must, by definition, result in a double or triple attack.

Also, keep in mind that not all of the AA descriptions are correct. For example:

-The Spell Casting Mastery skill indiciates that it increases the chances of making the specialization check when casting a spell. There is official word AND parsing to show that there IS NO SPECIALIZATION CHECK.
-Improved Lay Hands indicates the skill causes Lay Hands to become a Complete Heal. Technically that is not correct. It neither matches the Complete Heal spell nor does it necessarily heal a target to its full hitpoints (e.g., a charmed pet with 20k hitpoints won't be complete healed by lay hands).

There are other skills that mistate what they do - the statement was either always ambiguious/incorrect or the skill was modified without fixing the description. It's not a good idea to take the skill descriptions at face value when it's quite possible they could be interpreted in a different way.


**************************************************************************************

wyrmbane
Dwarven Smith
Posts: 725
(7/23/03 11:02 pm)
| Edit | Del Re: SoTK level 1 - More data on SOTK 1 added - Sotk 2 soon

SOTK level 2 parse.

Singles - 4648 - 39.72%
Doubles - 6718 - 57.40%
Triples - 337 - 2.88%


THis parse was done on Harnoff again. No haste item equiped, using a windblade, delay 44.

The result is interesting because the singles have not changed. Based on a comparison of SOTK 2 parse to SoTK1 parse, only the doubles decreased and the triples increased. No change to singles. Given that, it appears the original theory of only adding triples is correct.

The baseline is the problem. Based on Medion's comment / reminder that SoTK works only on 2 handed weapons, I am going to get a new baseline on Harnoff tonight.

I pulled my old wurmslayer, delay 40 off of an alt. I will be running another test, roughly 12k or so attacks. Tomorrow I should have the data, provided i don't go Linkdead for some reason. My expectation is similar to Romidar's post earlier that I will single 40%, double 60%. That would be consistent with the change from SoTK 1 to SoTK 2. Will have to wait until tomorrow to determine if that is true.

Wyrmbane


**************************************************************************************

Wulfhere Hausakliufr
PvP Forum Moderator
Posts: 1431
(7/24/03 12:22 am)
| Edit | Del Re: SoTK level 1 - More data on SOTK 1 added - Sotk 2 soon So it sounds like 3% additional attacks then?

Wulfhere Hausakliufr
Hammer o Brell, Rallos Zek
Back to top
View user's profile Send private message Send e-mail




Author Message
Valya Wanderfoot
PoN Admin Staff


Joined: 06 Dec 2004
Posts: 840


PostPosted: Thu Dec 23, 2004 4:39 pm Post subject: Reply with quote Edit/Delete this post Delete this post View IP address of poster
Qualtar Shylok
Blessed Cavalier
Posts: 514
(7/24/03 5:08 am)
| Edit | Del Re: SoTK level 1 - More data on SOTK 1 added - Sotk 2 soon

What would the disadvantages of determining DPS prior to the purchase of these AA's and comparing to DPS after obtaining each level of these AA's and noting the difference for determining the amount of increase in DPS be?

Seems to me that would be the easiest way to get the data, assuming you continue to take parses from the same mobs, and you have the same equipment etc....then things like latency, etc would become irrelevant.

*******************************************************************

Romidar
Knight of the Truthbringer
and PoN Admin
Posts: 2124
(7/24/03 7:34 am)
| Edit | Del
ezSupporter
Re: SoTK level 1 - More data on SOTK 1 added - Sotk 2 soon

Quote:What would the disadvantages of determining DPS prior to the purchase of these AA's and comparing to DPS after obtaining each level of these AA's and noting the difference for determining the amount of increase in DPS be?

Seems to me that would be the easiest way to get the data, assuming you continue to take parses from the same mobs, and you have the same equipment etc....then things like latency, etc would become irrelevant.



There's nothing really wrong with doing that, and that's the type of parse we've already seen.

However, there are two "issues" with doing it:

1. You need a huge amount of data in very similar situations to get a halfway decent estimate of DPS.
2. It doesn't give the fidelity to address exactly what the skill is doing, which is what Wyrmbane is trying to establish. That is, does it give you a chance on every swing or only a chance at a third swing you get a double attack?

In practical DPS terms, the latter question doesn't really matter - it's going to give a particular number of extra swings in the long run.

However, consider whether you should get SotK or KA first. If SotK gives you the chance for a third attack WHEN you get a triple but not otherwise, then it makes sense to get KA first. That way, when you do get SotK, you have more double attacks for the skill to work on, so to speak. (Since pretty much everyone will get both skills, the order in which you should get them is just nice to know. Smile )

The problem is really that the wording of the skill description is ambiguious (what is an "extra attack?") and Wyrmbane's "study" will disambiguate it - the DPS comparison wouldn't. As I said, though, if what you really want to know is "how much will it increase my DPS?" then you could use the method you propose with a sufficiently large sample size.

For estimating DPS, though, knowing the rate of doubles and triples is EXTREMELY helpful. That, combined with an estimate of your average hit will provide an extremely good estimate of how much damage you can expect to do. I'm often very impressed with Caladel's parser in terms of how close his computation of DPS comes to estimate DPS based on the real average, the delay of the weapon, haste, double attack rate and miss rate.

That is, I can take those disparate elements and predict DPS (take the miss rate from the log, my estimated double attack rate - I've been using 60% - my average hit, etc.) and calculate DPS:

DPS = (averge * hit rate * (1+ DA rate))/(adjusted delay)

That comes out very close to what EQParser MEASURES as DPS in a real fight; that's particularly surprising given the stuff that happens in a real fight that isn't in the equation (being stunned, being out of range, casting, etc.). Over a sufficiently large sample, those things get smoothed out, though.

What should happen as I get higher levels of SotK and KA is that that equation should become worse and worse at estimating DPS. I need to add in a triple attack factor (could add in a riposte factor too) and adjust the estimate of Double Attack rate.

Measuring change in swing rate vs. change in DPS is really a matter of granularity.


*******************************************************************

wyrmbane
Dwarven Smith
Posts: 726
(7/24/03 9:11 am)
| Edit | Del

BASLINE DATA
This is short, but shows the trend, which is per the old formula.

Using a WURMSLAYER, Delay 40, NO haste item equiped. Using our friend Harnoff as the target, Everything filtered except my hits, my misses, and the attacker missing me, hitting me.

Singles -- 2733 - 39.53%
Doubles - 4181 - 60.47%
Triples -- 0
Total - 6914

First off, Yes, this is small. However, if I extend this another 10-20k attacks, is should merely refine the numbers as they trend closer to the final number. As posted by romidar earlier in this thread, the expected theoretical double is 60% at level 65, based on the formula:
Doubles = ( double attack skill + level ) / 500
Doubles = ( 235 + 65 ) / 500 == 0.6

If we list the skills in order, Base - SoTK 1 - SoTK 2
40%(39.53%) -- 39.84% -- 39.72% - singles
60%(60.47%) -- 58.95% -- 57.40% - doubles
0% ------------ 1.205% -- 2.88% --- Triples

Based on the numbers, it appears that the singles are not significantly changing. The doubles decrease as the skill increases as do the triples.

Without significantly larger samples, 100K+ attacks to refine the %'s to whole numbers, we won't know the exact increases. However, we can make reasonable estimations.

All triples are simple doubles + the extra attack provided by the SoTK skill. To normalize this we divide by the Number of aa's spent for the skill.

Sotk 1 = 1.2% chance / 3 aa = 0.4 % increase per aa
Sotk 2 = 2.88% - 1.2% (delta for the new skill) = 1.68%
Sotk 2 = 1.68% increase / 3 aas = 0.49 % increase per aa

My computer monitor has been problimatic, so I don't have SoTK 3. I think I am halfway there. Once done, I will parse that as well.


wyrmbane

*******************************************************************


wyrmbane
Dwarven Smith
Posts: 729
(7/24/03 10:37 am)
| Edit | Del re: SoTK theory
here is my working theory....won't know if it is correct until i finish SOTK 3

Sotk theory is it gives a 2%, 5%, 10% chance to make a double attack a triple attack.

So, level 1:
Triples = Sotk1% (2) * Base Doubles% (60%)
Triples = 1.2% of the total

Level 2:
Triples = Sotk2% (5) * BaseDoubles% (60%)
Triples = 3% of the total

Level 3:
Triples = Sotk3% (10) * BaseDoubles% (60%)
Triples = 6% of the total

I do not know if this is correct, this is a Theory only at this time.

wyrmbane


*******************************************************************

gowrygabriev
Noble Squire
Posts: 65
(7/24/03 11:06 am)
| Edit | Del Perhaps look at it differently

Thanks for doing these parses! This
info is great.

After seeing the trend of decreasing single
and double attacks but increasing triples. I
began to wonder if we aren't looking at the
data wrong.

Could you look at your raw data and just count
raw # of attacks in a fixed period of time. That
would give us an idea of overall effectiveness.

I'd be interested to know in a 3 or 4 hour time
period, how many swings were made at each
level, provided the weapon had same delay and
haste and all were the same.

Gowry


*******************************************************************

Pankrat
Noble Squire
Posts: 208
(7/24/03 7:23 pm)
| Edit | Del Re: Perhaps look at it differently

OK went with Wyrm's theory and came up with my own about Knights Advantage and putting the 2 together seem to have come up to the answer of the effect of both skills:

KA 1,2,3 adds a 2%,5% and 10% increase to your double percentage per level, ie if you normally double 60% of the time you will now double 61.2% at level 1, 63% at level 2, and 66% percent of the time at level 3.

If you connect with an opponent 1000 times for an average dmg of 100 this would mean you would do:

160,000 pts of dmg with Double Attack and no KA (100K base,60K doubles)
161,200 pts of dmg with KA1 (61.2% double attack) -1.2K extra dmg - .75% dps increase
163,000 pts of dmg with KA2 - 3K extra dmg - 1.875% dps increase
166,000 pts of dmg with KA3 - 6K extra dmg - 3.75% dps increase

Now do the same thing with SoTK and wyrms theory and you get (1000 connects avg dmg 100):
SoTK1 = 2% of 60K in doubles become triples = 1200 extra dmg - .75% dps increase
SoTK2 - 5% of 60K in doubles become triples =3K extra dmg - 1.875% dps increase
SoTK3 - 10% of 60K in doubles become triples = 6k extra dmg - 3.75% dps increase

Both are showing equal point for point value per level in dps increase, however, based on these 2 theories, KA will increase your double percentage hence automatically slightly increase your triples as well. But SoTK will have no effect on KA.

Lets do the math on KA3 and SoTK3 together.

KA3 = 166K dmg on 1000 connects, 66K in double (66%). 10% of those doubles get a 3rd attack or 6.6K. 166K+6.6K=172.6K

172.6K vs 160K = a 7.9% increase in DPS.... which happens to be exactly what Caladel parsed KA3 and SoTK3 to come out at...

By George I think we have it =).

Conclusion (assuming the theory is correct): Up until level 2 KA and SoTK are of equal AA value - ie 6 pts for 1.875% dps increase in each respectively. However at level 3, SoTK becomes the much better buy (half the cost).

Also KA slightly increases DPS for SoTK but the reverse is not true. Because of the high cost of KA 3, one might consider going with SoTK3 and KA2 for a 5.8% DPS increase for 15 AA, vs 7.9% increase in DPS for 21AA (both skills maxed) at least in the shorter term.

Pankrat Lord Protector of Tarew Marr

LLTS!
Edited by: Pankrat at: 7/24/03 9:12 pm

*******************************************************************

wyrmbane
Dwarven Smith
Posts: 733
(7/28/03 10:10 pm)
| Edit | Del Updates
More information.

I did some more parsing of Pre-SOTK with my wurmslayer. Also, I got SOTK 3 last night, and did more parsing with it.

PRE- SOTK - Baseline - Wurmslayer, 40 dly, no haste

Singles - 7513 - 39.899%
doubles - 11316 - 60.096%
triples - 1 ------ 0.005%

Note: I included the triple for one reason. This is a latency value. three timestamps were the same, but I cannot remove an exact 3 occurance because I use the same code for evaluation the effects of SOTK. In short, assume all other numbers of SOTK 1 - 3 are off by +/- 0.005%


SOTK Level 3 - Windblade, dly 44, no haste

Singles - 6757 -- 39.810%
Doubles - 9420 -- 55.50%
triples - 796 ---- 4.690%

I am going to be running more tests on SOTK3. I don't think I have enough attacks to bring in the correct number of triples at this point. More data will help the triple rate occur more often and normalize to the final value. My expectation is I will need 30-60k attacks to determine just how much it helps. Just takes time with a 4 second delay.

wyrmbane

*******************************************************************


Romidar
Knight of the Truthbringer
and PoN Admin
Posts: 2158
(7/29/03 9:54 am)
| Edit | Del
ezSupporter
Re: Updates

Well, if we go out on a limb and assume that the numbers are actually:

Single: 40%
Double: 55%
Triple: 5%

We can compute how many additional swings you get.

No SotK
If 100 "delay intervals" pass:

40 will be Singles
60 will be Doubles

Total swings = 1*40 + 2*60 = 160

SotK3

40 will be Singles
55 will be Doubles
5 will be Triples

Total swings = 1*40 + 2*55 + 3*5 = 165

So, an extra 5 swings you wouldn't have gotten over 100 delay intervals. With an unhasted Windblade, that would mean 5 extra swings every 440 seconds or 7.3 minutes. Hasted at 100% hasted windblade would mean 5 extra swings every 220 seconds or 3.7 minutes.

Another way to put it is an increase of 5/160 or 3.125% increase in swings/damage output.

Just to look at the math in an example, I've been computed my DPS by looking at my average damage, adjusted delay and miss rate, then estimating double attack rate (correctly, as it turns out). To see the change I can expect with SotK3....

Adjusted haste (self buffed): 44/1.61 = 27.3
Average hit (from log): 168
Hit rate (from log): 0.504
Double attack (estimate): 0.6
Average Single = 168
Average Double = 336 (2*168)

DPS = (0.4*168 + 0.6*336)*0.504)/2.73 = 49.62

If my double rate drops to 0.55 and I introduce a triple of 0.05:

Average Triple = 504 (3*168)

DPS = (0.4*168 + 0.55*336 + 0.05*504)*0.504/2.73 = 51.18

That's a DPS increase of:

(51.18 - 49.62)/49.62 = 3.13%

Now, here's something interesting, I think.

Assume that KA3 increased double attacks by 10%. Assume that SotK3 changes 10% of the double attacks into triple attacks.

Using the "100 delay" system I described earlier, we'd start with:

40 Singles
60 Doubles
Total swings = 160

If we increase the doubles by 10%, that results in 66 instead of 60:

34 Singles
66 Doubles

Now if we turn 10% of those doubles we get:

34 Singles
59.4 Doubles
6.6 Triples
Total swings = 34 + 2*59.4 + 3*6.6 = 172.6

That's an extra 12.6 swings that weren't there before, for a total increase of 12.6/160 = 7.9% in swing rate. If I remember correctly, the estimate for the total effect on DPS of the two skills combined was around 8%.

Now, to add another IF... if the above assumptions are correct (i.e., KA3 = 10% increase to doubles), then KA without SotK causes you to go from 160 swings to 166 (3.75% increase). Based on the same assumptions (10% of doubles become triples), SotK without KA increases your swings from 160 to 166 (3.75% increase). The two numbers do add up to only 7.5% because they are not combined in a simple additive fashion - SotK is boosted by having KA.

What this means, though, is that SotK is probably the better buy, if I remember the cost of the skills. Total cost for SotK = 9 and total cost for KA = 12. KA is higher cost for the same increase in DPS (in isolation) but has the added benefit of improving another AA if you have it.

I actually bought the first level of both skills when I levelled to 61, but I'm pretty sure I'll be getting level 3 of SotK before I get level 2 of KA. Doesn't make a huge difference either way, really, but I can get to level 3 of SotK for only 6 more points whereas KA would take another 10.

I'd be very interested to see an analysis of KA in isolation, but I'm surprised at how well the numbers fit together on a theoretical analysis if you assume a 10% increase in doubles and a 10% double to triple conversion rate.

EDIT: Death to Emoticons!

Edited by: Romidar at: 7/29/03 10:29 am

*******************************************************************

wyrmbane
Dwarven Smith
Posts: 734
(7/29/03 11:38 am)
| Edit | Del Re: Updates
Actually Romidar, I have done tentative analysis in the exact same fashion you did above. I am also working on 2 assumptions, that SoTk is a 2, 5, 10% increase, and that KA is a 2, 5, 10%, and subsequently that KA and SOTK stack.

I am parsing more SOTK3, to see if it does get closer to 10%, currently it is not that high. However, if the above holds true, the ideal path for KA / SOTK is:

Skill - Percent increase
KA1------0.75%
sotk1----1.52%
sotk2----2.66%
sotk3----4.58%
ka2------5.68%
ka3------7.88%

If you do it with KA first, then sotk
ka1------0.75%
ka2------1.75%
sotk1----2.53%
sotk2----3.71%
sotk3----3.71%
ka3------7.88%

Based on number of aa's for the skills and gains over the run, the first option is the best one for the aa's spent. The second gets you a better step 2, but you suffer less total increase in step 3, 4.

Anyhow, I am still working SOTK3, and hope to have KA1 done and parsed before I leave the country, for a week, on sunday.

wyrmbane


*******************************************************************

Banvaren
Noble Squire
Posts: 199
(7/29/03 11:42 am)
| Edit | Del Re: SoTK level 1, SoTK 2 data added as a reply

Excellent data. Are you going to continue parsing KA 1-3? It should be possible if you take your SotK3 parse as a baseline for KA0.

I think it would be very interesting to see how many more double attacks you get. Also, if those added double attacks can, in fact, be converted into extra triple attacks.

Banvaren Holybane
Officer of Stoic

*******************************************************************

wyrmbane
Dwarven Smith
Posts: 735
(7/29/03 12:07 pm)
| Edit | Del Re: SoTK level 1, SoTK 2 data added as a reply
Yes, I plan on getting a very large sample of SOTK3, so I know the final effect of the skill, and also because it will be the baseline for my KA1-3 parses.

So yes, I am planning to continue this until we have good numbers on the skills to know what they do.

wyrmbane


*******************************************************************

wyrmbane
Dwarven Smith
Posts: 739
(7/30/03 6:07 pm)
| Edit | Del Final SOTK3

Final number for SOTK 3 - 3 nights of parsing, all with Windblade, all with no haste belt, all on harnoff, maximum filtering to reduce latency

Singles - 19048 -- 39.985%
doubles - 26305 -- 55.221%
triples -- 2284 ---- 4.794%

Based on the numbers, It is very safe to conclude the following:

Speed of the Knight:
This skill gives the Knight a 2%, 5%, 8% chance to score and additional attack on any Double attack while wielding a 2-handed weapon.

Effects?
Given the base of 60% doubles, we see the following amount of triples:
SoTK1 - 2% of doubles = 1.2% triples
SoTK2 - 5% of doubles = 3% triples
SOTK3 - 8% of doubles = 4.8% triples.

The DPS increment is not the value above though.
For 1000 Attacks.
Sotk level - Singles - doubles - triples - % increase
Baseline --- 400 ---- 600 ----- 0 -- 0 / 1600 = 0
SoTK1 ----- 400 --- 588 ----- 12 -- 12/1600 = 0.75%
SoTK2 ----- 400 --- 570 ----- 30 -- 30/1600 = 1.875%
SoTK3 ----- 400 --- 552 ----- 48 -- 48/1600 = 3.0%

That is what you get for SOTK 1 - 3 only. Once I get Knights advantage, I can see how much double attacks go up, and whether the two skills are cumulative. After that, we can figure out what the ideal path of Skill / aa spent is.

wyrmbane


*******************************************************************

Kysharn
Noble Squire
Posts: 95
(7/31/03 10:41 am)
| Edit | Del Re: Final SOTK3

wyrmbane,

With your permision, when this is finalized and you and Romidar have said "This is it" I'll add this to the Comprehensive AA list information.


*******************************************************************

wyrmbane
Dwarven Smith
Posts: 742
(7/31/03 10:46 am)
| Edit | Del Re: Final SOTK3
By all means, please add it to the list for AA's. Actually, that is the reason I included the quoted "correct" description in my last post. The intent was to let everyone know what it does.

Once I finish the parses of Knights advantage, I plan on making a single summary post of both skills, what they do, how they increase DPS, and what order to buy them in for maximum increase in DPS for each aa spent.

I won't be doing any more SoTK parses, so the information above showing 2%, 5%, 8% chance to turn a double into a triple is the best I know at this time. It is possible if knights advantage stacks with SoTK, that more doubles will refine that final SoTK numbers. However, with 50,000 attacks, I think it is safe to say we are within an acceptable range of error on the skill effect.

wyrmbane

*******************************************************************


Kysharn
Noble Squire
Posts: 96
(7/31/03 10:51 am)
| Edit | Del Re: Final SOTK3 Adding now.

Thank you VERY much.

*******************************************************************

Watlow
Noble Squire
Posts: 146
(8/2/03 8:29 am)
| Edit | Del Re: BASLINE DATA

Nice job!

One thing tho, i could've missed it, there's so much data in thread, but are you correcting for ripostes? Or are those not applicable at all? Im asking cause i noticed you mentioning in your baseline parse saying they were taken during normal fighting, which could mean tanking and hence ripostes. This would explain why you are seeing triples. Those are not due to latency, but due to a double-attack and riposte in the same second. This would also contaminate the single and double attack rate.


*******************************************************************

wyrmbane
Dwarven Smith
Posts: 748
(8/2/03 8:39 pm)
| Edit | Del Re: BASLINE DATA
Ripostes Are excluded from all data. THis was a learn as I go process.

All of the final data is based on ONLY fighting Harnoff in Kaladim. All data filtered to reduce latency.

Basically, I eliminated as many variable as possible by using the same weapon, no haste, and the same mob for all tests.

Wyrmbane


*******************************************************************

Bahcon
Noble Squire
Posts: 145
(8/3/03 9:55 am)
| Edit | Del Re: SoTK level 1 - More data on SOTK 1 added - Sotk 2 soon

believe improved LH is a "complete heal" based on what a "complete heal" spell did before they changed it to a 7500 heal. heals for just a lil over 10k with no focus, aa's etc... not to sidetrack the thread Smile . Just think the information on the aa's page was accurate when it was put there.

Conarch Azmiel
65th Paladin of Brell
My Gear

*******************************************************************

Luthair
Lord Crusader
Posts: 2443
(8/3/03 10:31 am)
| Edit | Del Re: Placeholder SoTK level 1 - What it does

All the new quest mobs in SS are like the old Soulbinders, they won't fight back and have a load of HP.

At 64 this was the consider:
Ocoenydd Fe`Dhar scowls at you, ready to attack -- You could probably win this fight.

Luthair [64 Crusader]
Alakazaam [57 Conjurer]
Gagbaz [41 Shaman]

*******************************************************************

Qualtar Shylok
Blessed Cavalier
Posts: 537
(8/3/03 1:52 pm)
| Edit | Del Re: Placeholder SoTK level 1 - What it does

The real question is, how much does this increase the amount of ripostes you take due to the extra swings...

Guessing its mob dependent as each mobs riposte skill maybe different.

Assuming a mob has a 5% riposte rate, this would add nearly .5% of the damage you take.
Back to top
View user's profile Send private message Send e-mail




Author Message
Valya Wanderfoot
PoN Admin Staff


Joined: 06 Dec 2004
Posts: 840


PostPosted: Thu Dec 23, 2004 4:40 pm Post subject: Reply with quote Edit/Delete this post Delete this post View IP address of poster
wyrmbane
Dwarven Smith
Posts: 749
(8/3/03 2:52 pm)
| Edit | Del Re: Placeholder SoTK level 1 - What it does

2.97% total increase in damage due to ripostes.

No more, no less.

wyrmbane

***************************************************************

Qualtar Shylok
Blessed Cavalier
Posts: 539
(8/3/03 10:12 pm)
| Edit | Del Re: Placeholder SoTK level 1 - What it does

How did ya figure that out? ~3% damage increase nearly negates 2 levels of LR or ID. If thats the case seems it would certainly not be worth getting. Especially since we don't have the numbers from KA yet.


***************************************************************

Romidar
Knight of the Truthbringer
and PoN Admin
Posts: 2205
(8/4/03 8:06 am)
| Edit | Del
ezSupporter
Re: Placeholder SoTK level 1 - What it does

It's only the riposte damage that increases by 3%, not overall damage.

Say you swing 10,000 times without SotK and KA. Most mobs I've fought riposte about 4% of the time. That means you'd be riposted 400 times.

After getting SotK3 and KA3, you get 8% more swings in the same period of time, so 10,800 instead of 10,000. Based on the same 4% riposte rate, that's 432 ripostes.

That's an increase of 32 ripostes over 400, or an 8% increase - the same increase in ripostes that you got in swings.

To put some other numbers to it, say a mob does 40 DPS to you and 2 DPS of that comes from ripostes (that's about 5% of total damage, which is probably a bit high). If you had both skills maxed, that'd go up by 8%, so instead of taking 2 DPS from riposte, you'd take 2.16 DPS from ripostes. That's an overall increase of 40 DPS to 40.16 DPS, or a 0.4% increase in damage taken exchanged for an 8% increase in damage output.

Ripostes suck for sure, but they only really matter when you're low on health OR the mob his hitting SO hard that getting an extra riposte will kill you before CH can land. In those cases, turn off attack - for a normal experience grind, though, ripostes account for only a small fraction of total damage taken, and that small fraction is only increased by a small fraction by getting these skills.
User avatar
Valya Wanderfoot
 
Posts: 102
Joined: Sat Aug 26, 2006 12:54 pm

Return to Tomes of Knowledge

Who is online

Users browsing this forum: No registered users and 51 guests

cron