DPS Spreadsheet - Page 51

nosille · 04/02/08, 2:28 PM

i was recently playing around with the spreadsheet and noted that enabling and disabling murder resulted in a 3% dps change with my mutilate build (not the 2% dps change I expected). After some review of the UnBuffed DPS worksheet, I noted that it appears that murder is getting applied twice to critical damage. Once in the crit factor and a second time in the base DPS calculation.

Is this correct? Does murder really increase crit damage by 4%?

saedo · 04/02/08, 2:30 PM

Originally Posted by nosille

Is this correct? Does murder really increase crit damage by 4%?

Yes. Yes it does.

Dbakes · 04/02/08, 2:32 PM

Hi, I'm a bit new to the spreadsheet (like an hour) but I think I've got it figured out and most of it makes sense to me. Just one thing that I was wondering if people would comment on.

I added my gear in the Gear_Buffs sheet and hit the Save DPS button, so I could make changes and see how it would affect me.

I swapped out my Nightbane chest for the new Badge chest, and my dps went up about 1.5%.

Swapped out my Bladeangel's belt for the new badge belt, and my dps went up about 1%.

Swapped out my Skulker's Greaves for the new badge pants and my dps went *down* .25%.

Swapped out the Skulker's Greaves for Slayer's Legguards (not that I really will anytime soon) and my dps went *down* .5%.

Does this seem right? It's certainly not what I expected.

Thanks in advance.

saedo · 04/02/08, 2:34 PM

Originally Posted by Dbakes

Does this seem weird to anyone else? It's certainly not what I expected.

Meta gem requirements?

Dbakes · 04/02/08, 2:38 PM

Originally Posted by saedo

Meta gem requirements?

Thank you. With the right gems, dps goes *up* 1.36%.

Dbakes · 04/02/08, 3:03 PM

While I'm here...

The spreadsheet runs fine in Excel on Windows, but using OpenOffice 2.2 on Kubuntu 7.04, OpenOffice crashes after the "Enable Macros" dialog without providing any information about why.

Not a big deal to me, but I thought someone might want to know.

Left · 04/02/08, 3:18 PM

Originally Posted by nosille

i was recently playing around with the spreadsheet and noted that enabling and disabling murder resulted in a 3% dps change with my mutilate build (not the 2% dps change I expected). After some review of the UnBuffed DPS worksheet, I noted that it appears that murder is getting applied twice to critical damage. Once in the crit factor and a second time in the base DPS calculation.

Is this correct? Does murder really increase crit damage by 4%?

Read through the Roguecraft 101 thread, or either this thread or the Gear Spreadsheet thread, for multiple discussions of this very issue. It is counterintuitive, sure, but the sheet models it as Blizzard actually implements it.

Originally Posted by Dbakes

While I'm here...

The spreadsheet runs fine in Excel on Windows, but using OpenOffice 2.2 on Kubuntu 7.04, OpenOffice crashes after the "Enable Macros" dialog without providing any information about why.

Not a big deal to me, but I thought someone might want to know.

Macros from Excel, as a general rule, don't run in OpenOffice. You should use the "OO" version of the sheet, which has macro functionality stripped out. (It's unfortunate, but that's kinda how it is. Most functionality will still work.)

Dbakes · 04/02/08, 4:10 PM

Originally Posted by Left

Macros from Excel, as a general rule, don't run in OpenOffice. You should use the "OO" version of the sheet, which has macro functionality stripped out. (It's unfortunate, but that's kinda how it is. Most functionality will still work.)

Same result with both "OO" versions, whether I choose to enable or disable macros when prompted.

Dontmindme · 04/02/08, 7:41 PM

Well, I have no problems opening it up in Open Office 2.3. In terms of Macros, given the large number of things to maintain on the spreadsheet, I really do not expect that to be done for OO before we see the expansion. Too many things need to be retooled to make that major transition easier. Besides that, I'm not very familiar with the intricacies of OO Macros and spending the time to get up to speed on what is or isn't compatible doesn't see like the best use of my time.

The only way I imagine we'd see OO Macros is if someone familiar with them sent me conversions.

Agkelos · 04/02/08, 11:09 PM

Hi,

When attempting to open up this spreadsheet I have been recieving the following message:

"This workbook has lost its VBA project, ActiveX controls and any other programmability-related features."

When I click the log file to view repairs this is what appears:

<?xml version="1.0" encoding="UTF-8" standalone="yes" ?>
- <recoveryLog xmlns="http://schemas.openxmlformats.org/spreadsheetml/2006/main">
<logFileName>error029840_01.xml</logFileName>
<summary>Errors were detected in file 'C:\Documents and Settings\Josh's\Desktop\RogueDPS_2_4_0_6.xls'</summary>
- <additionalInfo>
<info>This workbook has lost its VBA project, ActiveX controls and any other programmability-related features.</info>
</additionalInfo>
</recoveryLog>

Now I have no idea what any of this means. All I know is that I'm unable to use this spreadsheet unfortunatly. I am running Microsoft Office 2007.

Anyone know what I may be doing wrong?

I have enabled all macros, activeX, and everything else I could completly. Help?

Acyrith · 04/02/08, 11:19 PM

Originally Posted by Agkelos

Hi,

When attempting to open up this spreadsheet I have been recieving the following message:

"This workbook has lost its VBA project, ActiveX controls and any other programmability-related features."

When I click the log file to view repairs this is what appears:

<?xml version="1.0" encoding="UTF-8" standalone="yes" ?>
- <recoveryLog xmlns="http://schemas.openxmlformats.org/spreadsheetml/2006/main">
<logFileName>error029840_01.xml</logFileName>
<summary>Errors were detected in file 'C:\Documents and Settings\Josh's\Desktop\RogueDPS_2_4_0_6.xls'</summary>
- <additionalInfo>
<info>This workbook has lost its VBA project, ActiveX controls and any other programmability-related features.</info>
</additionalInfo>
</recoveryLog>

Now I have no idea what any of this means. All I know is that I'm unable to use this spreadsheet unfortunatly. I am running Microsoft Office 2007.

Anyone know what I may be doing wrong?

I have enabled all macros, activeX, and everything else I could completly. Help?

You may want to check the posts on the previous page of this thread covering some possible solutions to your problem.

Littlefinger · 04/03/08, 8:27 AM

I don't know if this has been reported in any way but it seems that my unbuffed AP always shows to high. I checked by swapping around a few items and it seems the total unbuffed AP number is around 10% too high. I don't have deadliness and Kings shouldn't effect unbuffed AP so there seems to be something messing with the summed up total AP. Is this a know issue or is there some setting I have that's effecting the numbers?

Left · 04/03/08, 8:40 AM

Possibly the numbers are including normalized contributions from your trinket or Mongoose procs. If you disable Mongoose and unequip trinkets, do the base numbers make sense?

Left · 04/03/08, 1:25 PM

Well, as you know, I have been working on improving the cycles calculations in the DPS spreadsheet. (By the way, I am almost ready to post a proof-of-concept of the new cycles system for people to check out. I hope to have it up later today.)

In the process of making my own cycles, I got to the point where I was comparing the outcomes of my cycles with the outcomes of the existing Mutilate cycles. Here's what I found out: armor reduction is not taken into account anywhere in the existing Mutilate cycles when calculating the damage contribution of Instant attacks and SnD contributed damage. This is a fairly major bug, in my opinion.

Now, what I don't know is how exactly the information in the cycles is translated back to the actual DPS calculations from which the totals are drawn. However, I do know that the other cycle calculations (the non-Mutilate ones) take armor into account when figuring out the best cycle. Thus, the Mutilate cycles as they stand now are over-valuing instant attacks and SnD damage versus Rupture and Eviscerate (since Evis does actually include the armor reduction). This is another major reason to find the current Mutilate cycle selection process highly suspect.

Whether this throws off the actual numbers or not depends on whether the Mutilate cycle analysis module's output is used as actual damage output, or only used for cycle selection. If the latter is the case, then the bug only messes up cycle selection but armor is (hopefully) included in the actual damage calculations.

Left · 04/03/08, 2:36 PM

Well, I have finally been able to implement the new cycles model that I have been discussing. Here is a modified spreadsheet with the changes implemented:

DPS Spreadsheet (Mod) at Savefile.com

CAUTION: This is not a beta, or even an alpha. It is a proof-of-concept, and therefore not suitable for any real use. It is not recommended for download and use unless you are a spreadsheet developer.

What I Did
All cell references in the following list are with respect to the modified version of the spreadsheet, not the original.

The table located in cell range A21:E56 is a finisher table
This table replaces the old finisher table, which is located in A3:H18. This table also replaces the finisher table used by the old Mutilate cycles, located at A204:H221. (I didn't delete the old tables, because the old tables are still in use by the old cycles and I was running comparisons of the results.) This new finisher table uses, by and large, the same formulas as the old tables.
One thing which I did change is how Hemo is calculated. In the old table, Hemo was just tacked on at the end. However, we know that Hemo can crit and is multiplied by constant and percentage-based abilities, such as backstab. Thus, I relocated the additional damage from Hemo so that it would be properly multiplied by crit and other damage multipliers. Grayed out items determined to be inaccurate; I have removed/reverted them in my working version. See also additional discussion in the following posts.
The table located in cell range G21:M25 is a combo point probability table
It replaces the previous tables located at K14:L18, N9:P14, and K216:N220. The new table is valid for both Mutilate and Non-Mutilate cycles, as it will automatically select between the two based on the instant attack being employed.
The tables located at G27:M54 and O27:U47 are non-Mutilate and Mutilate probability tables, respectively
These tables predict the number of combo points reached and the number of instant attacks used based on various target combo point totals (1-5), talents (such as Seal Fate), and crit chance. They are based directly off of the state/path diagrams I posted earlier in this thread. These tables are combined in the probability table at G21:M25.
Beginning at cell A58 is a new cycle analysis module format
You will notice that the new format is much expanded from the old format (it has quite a few additional columns). In this new format I tried to break out many different calculations individually, both to make the sheet easier to understand and to allow more items to be compared/retrieved directly for use elsewhere. Columns A and AG list the cycle name, with column AG being the version of the name returned to the other sheets. Columns B-G total the number of finishers used of various types. Column H calculates the number of instants required to achieve the finishers used. It does this via the new probability table. Columns I-K calculate energy used over the course of the cycle, based on the probability weighting and the finisher table's energy cost column. Column L reports the average number of combo points generated in the cycle. Columns M-S are time and uptime calculations for finishers and events with duration, such as SnD time, Rupture time, and the actual cycle time itself. Uptimes are achieved by comparing finisher time to cycle time, and are capped at 100%. (This prevents overestimating of SnD or Rupture damage in a cycle where clipping occurs.) Finally, columns U-AC are damage calculations, weighting by the computed find weakness factor in column T. Column AC is DPS, which is the point of comparison. Columns AD-AF are there for reporting purposes. They show the slack factors on various finishers. (In other words, this is the amount of cut time on SnD.)

You will note that I only implemented a few cycles. I did this for testing; I'm not going further until we go over what I've done already. (I don't want to put more work in if we find a bug.) You will note, though, that I highlighted (and in some cases commented) the lines in the old cycles which correspond to the new cycles I created. This can be helpful for comparisons.

Known issues

Find Weakness is modeled continuously, instead of discretely applied to each ability. This is a generalization that doesn't work in practice, but is much easier for spreadsheet purposes. Actual FW damage contributions will likely be higher, as people's playstyles put as many abilities into the FW window as possible.
Non-SnD (or low SnD uptime) cycles are over-estimated for Combat builds. This is because Combat Potency and SnD feed the Energy Expansion Coefficient (eec), which is used to calculate cycle times. The eec will be lower when SnD is not used, but this isn't currently accounted for.
SnD uptimes for Seal Fate cycles may be overestimated. This is because the modeled cycle uses an average, but in actuality the cycles have a lot of variability. The variability can cause SnD to drop early a lot, where as extra SnD time is wasted. Thus, actual SnD uptimes for cycles with little SnD slack will likely be lower.
For cycles with Rupture clipping (not many), Rupture is modeled (and clipped) as continuous damage, not tick-based damage. So, 1% Rupture clipping results in 1% lost damage instead of 1 tick lost damage. This causes a bit of inaccuracy. However, most cycles don't have Rupture clipping.
I made spots for Expose Armor, but it is not modelled in any way currently.
Several items included in the old cycles (in a somewhat esoteric manner), such as ghostly strike, are not yet modeled in the new cycles in any way.

Observations
The new cycles are coming out in the ballpark range of the old cycles, but are often different by a significant amount. I feel that this is because the new cycles are a more accurate model, but I could be wrong. In particular, I'm seeing differences where SnD has either a lot of downtime or a lot of slack in the cycle, as these cases present the largest differences from the olds cycles.

For the reason stated in the above post (lack of armor reduction being included in the old cycles), all the new Mutilate cycles come out well below the old ones. However, if you modify the old Mutilate cycles to include armor reduction properly, the two start to look closer. It remains to be seen how this will affect DPS estimates.

Next Steps
Assuming that this looks good, or can be made good with a bit of modification, the next steps are:

Implement many more cycles
Migrate the red block which returns info to the DPS sheet over to the new cycle analysis module
Implement the whole thing in the Unbuffed sheet
Address remaining issues, such as Ghostly Strike, etc

Please give me your feedback! (DMM and Todemax in particular here.)