Rawr (v2.3.23 released on 9/12) - Page 32

hypetech · 04/15/08, 5:09 PM

Originally Posted by MyrddinE

Is there a way to eliminate all the T6 gems from the item database? I deleted the gems themselves from the database, but the database repopulates because there's hundreds of individual items socketed with them... removing hundreds of items manually sounds pretty damn tedious.

Alternately, could I remove all gems from all equipment, and then have the game make several 'likely gemmings' for the equipment based on my needs? A set of gems that ignore socket color and a set that gets the socket bonus, using the highest-rated gems that fit (and are available)? It's just tedious re-gemming equipment over and over using only gems that I am likely to get.

Don't select T6 gems in the optimizer?

MyrddinE · 04/15/08, 5:25 PM

Originally Posted by hypetech

Don't select T6 gems in the optimizer?

That's fine for when I'm using the optimizer... but trying to compare individual items in the equipment list is rather annoying when some items are gemmed, some not, some poorly, some well, some with T6 gems, some with blues... it's difficult and annoying to see how much better a particular piece of equipment would be for me without going in and deleting the versions of the equipment gemmed with T6 epics, making a new set gemmed with similar quality gems to what the other piece is, etc...

It's very time consuming.

grayrest · 04/15/08, 5:50 PM

Originally Posted by MyrddinE

Is there a way to eliminate all the T6 gems from the item database?

There is a socket gems button that isn't working in 13.1--I reported the bug, it's been fixed but no new release--which will fill all gems according to color. Fill with rares and delete the epics from the db and they'll be gone.

Astrylian · 04/15/08, 5:50 PM

MyrddinE, the Fill Sockets feature is designed to exactly solve your problem, but was broken in b13.1. It'll be fixed in b14. (Or if you're more technically inclined, everyone's always welcome to build their own copy from our latest development source code)

Kavan · 04/15/08, 5:53 PM

For purposes of evaluating upgrade value of items there will be a new tool in the next release. It is a combination of optimizer with gear ratings. While the usual comparison charts give you the direct comparison of swapping the item, this will tell you how much you can gain by allowing it to swap some other gear with items that you have available. This way for example if you're at hit cap an item with hit will be properly evaluated as it will consider swapping some other pieces with non-hit gear that you have available. Why I mention this is because for socketed items it will show you just the best gemming.

Larisroth · 04/15/08, 10:53 PM

Thanks for the latest beta Astrylian. I should play around with the source one day, but in the meantime here are a few comments and ideas.

The optimiser is a lot of fun, and pretty useful, but pretty dumb when it comes down to micro effects such as which gems to pick. Fortunately it's pretty easy for a human to do that by hand.

One of the things I was playing around with, was including some constraints in bear form. Following the descriptions of the mitigation and survival ratings, a somewhat sensible system to set up would be to maximise my mitigation rating with the constaints of at least a certain (I picked 100K) survival rating, and no chance to be crit. It certainly worked this out, but there were some obvious issues (like using a darkmoon trinket over the commendation because it had enough stamina on it), and doing some rather stupid gemming to get to the crit cap.

Anyway it motivated me to think about some ways that would improve the speed I could optimise by hand, and also help setup systems for the optimiser.

The first would be to allow tweaking of the total rating coefficients. As the survival and mitigation ratings measure different things, the scaling for the mitigation rating is arbitrary. Selecting sunwell radiance drops more-or-less halves the mitigation rating for the range I'm interested in. This would also allow users to tweak the ratio they like the two components in, and I could then optimise for mitigation plus a small bit of survival to resolve the stamina issue.

The crit cap is a bit more of an issue. I think we're always going to have to work to get to it, and with that in mind I'm wondering if some sort of crit chance reduction rating based on that sort of idea would be useful. If you've got no chance to be crit, then equiping crit reduction items are pointless so all stats would have a weighting of zero, below that a negative weighting might be quite useful. Obviously it only applies to resilience and defense rating, with resilience being basically 1.5 times as good. I think the rating at 2% should be quite a bit higher than at 1% because you're going to have to do a lot more work (regem, swap items/enchants etc) to get to the cap. You could then incorporate that rating into the total rating used by the optimiser and see how it goes, finishing it off by hand of course.

A related issue I have when trying to work out which upgrades to target in a scarce loot environment, is the limitation that each piece can only have one enchant (and gemming), but competing pieces may have different enchants. Is it possibly to incorporate the enchants into the item as well? (Are my t6 pants with the set bonus and a tanking enchant better than this offset rogue piece with the dps enchant.

The one major thing the feral community doesn't really model is threat (maybe dps) in bear form. That could be another possible component of the total rating if someone could come up with a suitable calculation.

Maax · 04/16/08, 12:29 PM

Regarding Rawr.Moonkin:

1) Is spell haste even being valued? It seems to have 0 effect on a gear/gem rating.
2) Why is INT rated so high? It tells me +10 INT gems are better than +12 damage gems, and I don't even have Dreamstate!

Astrylian · 04/16/08, 12:46 PM

Larisroth, thanks for the excellent feedback!

Optimizer being dumb... Yeah, we're working on improving it, just takes time.

Tweaking the Rating Coefficients... This won't be customizable for Survival and Mitigation, because they're a standard. Survival has no coefficient, it's a distinct, defined value. Mitigation's coefficient is just a multiplier to get it to the same scale as Survival. Threat will be included in b14 though, which will have a customizable coefficient.

AntiCrit... I've often thought about this, but it's such a specific and relatively small amount needed, that fine tweaking of it is necessary, so a rating isn't really useful as far as I can see. I'm certainly open to suggestions on this.

Specific item+gem+enchant combinations in the optimizer: I'd like to do this, but it's not an easy thing to implement, both technically and usability-wise. I'm thinking about some sort of 'Advanced Optimizer' that would let you more specifically define possible item+gem+enchant combinations for each slot, but that's a couple versions out, at least.

thedopefishlives · 04/16/08, 1:10 PM

Originally Posted by Maax

Regarding Rawr.Moonkin:

1) Is spell haste even being valued? It seems to have 0 effect on a gear/gem rating.
2) Why is INT rated so high? It tells me +10 INT gems are better than +12 damage gems, and I don't even have Dreamstate!

1) In both the old and new Rawr.Moonkin calculators, spell haste can only apply in integral amounts. That is to say, if adding a certain amount of spell haste doesn't give you enough haste to fit another Starfire cast into the DoT duration, then haste has no effective value for you. Note that this does include the reduction in cast time of Moonfire and Insect Swarm in the new calculator (not so sure about the old).
2) If your Time To OOM is anything other than "Not during fight", then the math will most likely show mana regen gear as having heavier weight than spell damage gear. This is especially true if the predicted time to OOM is much less than the fight duration. For most accurate results, please make sure you configure your buffs, debuffs, and options to match the fight as closely as possible. Also, I can't vouch for 100% accuracy of the old (beta 13) spell calculation engine, because a variety of mistakes were pointed out to me by some people after the release. Beta 14 will integrate the new calculation engine, which uses Adoriele's mathematical equations. I consider this to be significantly more reliable than the old one, although that will be seen after release when it gets into the hands of the other theorycrafters.

JettJaguar · 04/16/08, 6:12 PM

Hola. I just started using this tool for my ret pally and am very impressed. I did notice a couple of issues:

Cloak of Darkness is missing its crit rating
As a Draenei, rawr adds 1% hit for Heroic Presence. When I select the buff, it adds another 1% hit. I am pretty sure that they do not stack.

A handy feature might also be a check-box to automatically hit-cap the character so you don't need to worry about the values for items changing and it is a little easier to browse upgrades.

Larisroth · 04/16/08, 7:28 PM

Okay I should perhaps be more clear. What I guess I mean would be it would be great to be able to construct custom ratings as linear combinations of the other ratings (or in fact measurable things). The is certainly true for the bear, but considering your hp could be useful if you're dpsing. I guess the constraints do handle that a bit, but being able to display those other components side by side like we do with the inbuild total tanking rating would be nifty.

I played around with tanking gear a bit more, and it seems the balance between mitigation and survival ratings on items becomes a bit more normal with the late sunwell itemisation. Certainly the value of stamina vs agility got much closer to the easy to work with 1.5:1 ratio :P (ideal from an item value perspective).

After running the optimiser a few times with different design targets, I get a lot of superfluous gemmings floating around, perhaps I should just write a python script to clean up the item cache :P

Ivorthemage · 04/16/08, 7:45 PM

"EDIT: Ivor, look at Kavan. Also, remember that int will never be as good for you as it is for the Gnomes, since they get an extra 5.5% int."

Ok, I downloaded Kavan's profile and I see where I am getting different results. Its not the gnome thing.

Rawr is showing 1254 dps for his toon (un raid buffed). Select the 40/21 frost spec for his toon instead, with the same gear, and his dps jumps 100, and runed crimson spinels jump way ahead of brilliant lionseyes.

I suspect the difference has to do with spell rotations, but frankly I can't make heads or tails of it.

What it looks like to me is that Int values are extremely sensitive to minor changes in spec and spell rotation for arcane specs.

But it also looks to me like int>damage is only true for an "off" arc/frost spec.

MyrddinE · 04/16/08, 8:01 PM

Originally Posted by Kavan

For purposes of evaluating upgrade value of items there will be a new tool in the next release. It is a combination of optimizer with gear ratings. While the usual comparison charts give you the direct comparison of swapping the item, this will tell you how much you can gain by allowing it to swap some other gear with items that you have available. This way for example if you're at hit cap an item with hit will be properly evaluated as it will consider swapping some other pieces with non-hit gear that you have available. Why I mention this is because for socketed items it will show you just the best gemming.

That sounds wonderful. Thank you Kavan and Astrylian. :-) I'm looking forward to it.

Tuhalu · 04/16/08, 10:17 PM

Originally Posted by Astrylian

Threat will be included in b14 though, which will have a customizable coefficient.

Does this mean b14 will include hit and expertise ratings and the ability to add dps enchants in the Bear model? Outside of rewriting the way gems are done (which you know about), it's pretty much my only complaint with that model. It has definitely been difficult to see what the best gear is for many roles with the addition of so much Hit Rating and Expertise tanking gear.

Kavan · 04/16/08, 10:55 PM

Originally Posted by Ivorthemage

"EDIT: Ivor, look at Kavan. Also, remember that int will never be as good for you as it is for the Gnomes, since they get an extra 5.5% int."

Ok, I downloaded Kavan's profile and I see where I am getting different results. Its not the gnome thing.

Rawr is showing 1254 dps for his toon (un raid buffed). Select the 40/21 frost spec for his toon instead, with the same gear, and his dps jumps 100, and runed crimson spinels jump way ahead of brilliant lionseyes.

I suspect the difference has to do with spell rotations, but frankly I can't make heads or tails of it.

What it looks like to me is that Int values are extremely sensitive to minor changes in spec and spell rotation for arcane specs.

But it also looks to me like int>damage is only true for an "off" arc/frost spec.

Well you won't get an accurate picture unless you select the buffs available. In the setting I usually have (CoS, JoW, shadow priest, el. shaman, no CoE) the spec I'm using comes on top, the difference between it and 40/0/21 almost non-existent. If I change to 40/0/21 I still have int gems over dmg. Main influencing factor I think is whether you have CoE and/or CoS. The main point is that both dmg and int gems are very close in value in arcane and as a result matching socket colors makes a lot more sense than just stacking one everywhere.