I'm running Excel 2007 under Parallels on my Mac and 351 items took 43s at 121ms/per.
Another nice piece of data on the gear planner would be 'number of items checked'. And maybe even 'estimated time' based on the last run. Currently the number of items only updates after you run.
The gear re-gemming still doesn't take socket bonuses into account. Overall it's much much better than the 1st version though.
Getting socket bonuses considered is a bit tougher than I thought. Will have to try a few things out first.
Should now have the correct names for JC gems. You can set which gems to use on the Settings tab (including 3.2 Epic ones). The gemmer should hopefully no longer make mistakes when it comes to meta-gem requirements. Dragon's Eyes are no longer prismatic either (and just count as the color of their respective stat).
On the Gear Optimizer, would it be possible to put "None" at the top or bottom of the drop down list? It's also not clear to me what the sort order of items in those drop downs is. Alphabetical might be easiest.
On the Gear Optimizer, would it be possible to put "None" at the top or bottom of the drop down list? It's also not clear to me what the sort order of items in those drop downs is. Alphabetical might be easiest.
The drop-down menus are the same as the ones used on the Gear tab, which are sorted by the Gear Planner. You can sort alphabetically with the button on that page. But I agree, it can be probably be done better.
When i try to read my spec, this happens
<snip picture>
Any idea why? It does it in my older version of the spreadsheet which i previously loaded it in.
It's hard to tell, does it list which VBA code line is in error afterwards, when you press Debug? I just tried your Armory profile in my own copy and it works normally, so it might have just been the Armory messing up. Otherwise, I really need more information if you can provide it.
Set NodeList = GetSingleDataXLS(URL, "//talentGroups").ChildNodes
oh and i'm using v89g, i tried it in one of my earlier spreadsheets and it wasnt working properly either. I'm using office 2007, in windows 7. It was working fine a few weeks ago, then it messed up last week, so i assumed something was off software wise. Now on a fresh install it's doing the same thing.
Set NodeList = GetSingleDataXLS(URL, "//talentGroups").ChildNodes
oh and i'm using v89g, i tried it in one of my earlier spreadsheets and it wasnt working properly either. I'm using office 2007, in windows 7. It was working fine a few weeks ago, then it messed up last week, so i assumed something was off software wise. Now on a fresh install it's doing the same thing.
Ah, the pre-PTR version isn't downloading the Armory data correctly, since Blizzard added pet talents.
On the "Rotation Test" tab, would it be possible to allow us to directly input a number of times to run the test, instead of just the up/down arrows? I like running a lot of iterations to get a more accurate "real-world" average, and it takes a bit for it to count up to 1000+.
On the "Rotation Test" tab, would it be possible to allow us to directly input a number of times to run the test, instead of just the up/down arrows? I like running a lot of iterations to get a more accurate "real-world" average, and it takes a bit for it to count up to 1000+.
Good idea. I'd limited it first because the rotation test was quite slow, but with all the speed ups I've done it should be ok to run it 1000 times.
There appears to be a bug in the beta sheet, but I'm not sure what's causing it. Gear planner lists Boots of the Great Construct as a 60+ dps upgrade from my Boots of the Living Scale, just for my Marks spec. The optimizer disagrees (more on that in a second) and actually swapping them in also disagrees. The only thing I can think of is that my yellow meta gem requirement is currently satisfied by my boots, so maybe it's somehow deactivating the meta before it compares the new boots.
Also played around with the optimizer. It's quite slick. However, it does exhibit some of the failure modes of a greedy algorithm in the optimization space. If I check 'start with current gear,' it doesn't find anything from where I'm at -- clearly my gear is already optimized to at least a local maximum. If I uncheck that option, it finds a setup for my marks spec that's 50 DPS better, and what it comes up with for my survival spec is 2-3 DPS below my current gear (and 33 rating over the hit cap). How does it decide that it's converging on the maximum? Could the speed of convergence be a tweakable option? I wouldn't mind letting it run for a few hours while I'm at work, for example, to get a greater degree of certainty that it's tried all the reasonable options (just as an occasional thing, of course; in general the current speed is great for interactive use).
Does the order of the items in the lists on the optimizer change anything? A couple times it seemed to be preferring things higher in the list, but it wasn't clear to me whether that's true or not.
I think it should also treat the enchants just like it treats items (or maybe gems), and find the optimal one (possibly from a subset you explicitly allow) -- at least in the hand, feet, and weapon slots, where you can pick up additional hit in cases where gems aren't enough.
I'd recommend swapping the positions of the 'copy top three from gear planner' and 'optimize' buttons. It seems like you'd be more likely to press optimize multiple times than copy from gear planner. One time I made some tweak and clicked the big friendly button in the upper left one time, wiping out my selections.
The algorithm is a fairly simple one at the moment, I still have to tweak it. Or change with a more sophisticated one.
Basically it does the following at the moment. Excuse the pseudo-code.
Loop
For Each slot
-Try each item available for this slot, record DPS increase,
but skip items that failed 3 or more times.
-Note which items 'failed' by actually decreasing DPS
-Pick item with highest DPS increase
-Update the current set with this item, continue from this new set
Next slot
Until Dps from this loop = dps previous loop
Obviously, the amount of loops is capped by the option to set the amount of iterations.
The big downside is that it always starts with the same slot. It leaves a lot of possible paths unexplored.
The advantage is that it's very fast, since it starts skipping a lot of items after the 3rd loop.
A better way would be to start with a random slot. Or to calculate DPS upgrades for ALL slots/items first, then pick the highest. This would require a great deal more loops though.
EDIT:
There's a bug in the automatic re-gemmer used by the Gear Planner (and the Gear Optimizer) that's making it fail at your boots. Since your boots are the only source of a yellow gem on your gear it fails to find a spot to satisfy the meta-gem requirements. The Gear Planner unequips your current item first, regems. Then starts trying all the boots in the list. Each time it equips a pair of boots it regems afterwards and checks the dps.
The Boots of the Living don't have a yellow socket so the gemmer fails at that point, leading to no meta-gem activation. The Boots of the Great Construct however do have one, which gives them the superior DPS.
Mostly changes to the regemmer, which should help the Gear Planner/Optimizer. The optimizer now picks the highest dps increase item across all slots, rather than pick the one of the current slot each time. Depending on your starting set this can take longer or shorter compared to the '1-slot each time' method.
The regemmer should also now consider socket bonuses. It tries to pick the best one for your token 'blue' prismatic gem first and then tries to determine whether the default gem for that slot is worth it, considering the socket bonuses, compared to the 'red' filler gem.
As an aside, I've switched to Office 2007 as a development platform, so if any 2003 users experience problems, let me know.
I'm getting some strange errors with the spreadsheet. I'm using the most recent live version of the spreadsheet, and using the "Read from Armory" buttons on the Talents sheet produces a runtime error.
Run-time error '91':
Object variable of With block variable not set
Its not an internet connectivity issue as the Read from Amory function on the gear page still works. Also any hand changes on talents, gear, or buffs page do not change DPS at all, but changing these things through use of the Clear and Load buttons result in the expected DPS change.
I've used the spreadsheet for some time and haven't had this issue before. It occurs both with the version posted in this thread and in the Best DPS thread.
Ahh, I thought of that but dismissed it as being a different issue when I was unable to hand change talents myself. If I close the spreadsheet and hand change without trying to load off the armory first it will work however. Thanks for your quick response.
another thing i was wondering was, in v89g, is there any reason why changing talents on the talent sheet have no effect on the dps whatsoever? I was going to see how my spec worked in the older spreadsheet by just inputting it manually, but it isnt working properly. Any thoughts?
Does the latest iteration of the spreadsheet have the buff to Rake?
Rake in the sheet has always been dealing the full damage per tick, since that's how it was when I last tested Rake in game, shortly after beta. It seems like this is just a tooltip fix.
91beta4 Gear:Regem button doesn't seem to optimize very well. Load my gear, and push it a few times. The dps will drop, and the value will oscillate between two values.
I still want to be able to push a button in the Gear Optimizer and have it load the gear that is currently in the Gear tab into a selected slot, say, slot #1. The problem with the "Start with currently equipped gear" button apparently serving the same function is that the gear it references changes over time.
The gear optimizer was able to find an item configuration I failed to find manually that was a few dps higher. For some reason, though, it chose a +16 hit gem for my gun, even though I was already over the hit cap. Unfortunately, I didn't save the starting gear setup to allow you to duplicate the problem.
Just using "get top 3 from gear planner" and hitting optimize ends up with Dark Matter being used twice.
Another bug: load my gear in beta4, turn off the Draenei Racial buff on the Buffs tab, then hit Recalculate DPS on the Gear Planner tab. The planner will show that Cloak of Mastery is -2.83 dps under the Hammerhead cloak. Substituting the Cloak of Mastery in the Gear page slot, however, drops dps by -14+. I'm assuming that Recalculate DPS actually substitutes the item, so the 2 numbers don't make sense.
The Gear Planner's calculations assume you re-gem after equipping. So hit the 'regem' button after making a gear change, to see what gemming the gemmer thought was 'better'.
As you may have realised the accuracy of the gear planner is heavily dependant on how it gems. It's difficult to make it consider socket bonuses and, above all, act in a predictable manner.
EDIT:
New beta, version 6, here
Re-wrote the gemmer again, hopefully for good now (barring bugs). Better logic design. Also many changes to the sheet in preparation for 'heroic' versions of gear, adding itemIDs to all items in the sheet, etc..
EDIT#2:
Re-uploaded version 6, since I broke something unintended (sorting on Gear Planner page).
EDIT#3:
And again. Not a fix, but a change so sorting works as it did before on the Gear Planner again. I didn't really like the new method after all.
The gem optimizer now does a better job than I have been able to do by hand; congratulations!
In beta6, the value of "Ranged Critical Hit Chance vs. Level 83" in the upper left of the Gear tab seems to have changed meaning, and dwarves no longer get a 46 racial crit bonus. Specifically, beta4 said 59.47% crit with 1023 rating and beta6 says 51.71% crit with 978 rating (exact same gear, gems, enchants, and buffs.) (1023-978 != 46 presumably due to rounding to nearest integer in the display.)
Darkmoon Card: Greatness is ilvl 200, not 226.
The Gear Planner is choosing Darkweb Bindings as the #1 choice for wrist at -271 dps. Pressing Sort has no effect. Seems like an off-by-one error.
07/12/09, 6:01 PM
Similar Threads
