Mew -- The 4.0.1/Cataclysm Feral DPS Simulator - Page 3

Deathwing · 10/17/10, 8:10 PM

Originally Posted by Leafkiller

I have posted a Sim script on Alaron's forum here: The Fluid Druid - View topic - Mew: The new Cataclysm Feral DPS simulator

It is very close to the script that is currently the default script - but there are a few interesting differences detailed in the comments at the top of the script including prioritizing Rip over SR and how OOC procs are handled.

The script was developed in conjunction with the Ovale script posted here: The Fluid Druid - View topic - New Feral DPS Ovale Script

This allows you to see what the rotation actually looks like in game - which can be surprising - especially with the Rake clipping. The Ovale script is fully functional for those who like to use Ovale when raiding.

One concern I have with the current default script is the inclusion of Rip clipping at 2 seconds. Rips cannot always be clipped ("A more powerful spell is already active"). A question for Yawning - is there anything in the simulator that takes this into account?

You mentioned that there were comments at the top about OOC procs, but I don't see anything. Can you give some details as to why you didn't prioritize shred for OOC?

kalbear · 10/17/10, 11:06 PM

Shred being prioritized for OoC procs was already somewhat flawed and way too dependent on other factors to be a hard and fast rule; that was when shred was a significant damage portion, only less than melee. Now that bleeds are more important it's even less of a sure thing to do OoC for it. Also given the energy differences are so similar now there's not nearly as much a gain. It's better to not miss rip/rake ticks if you would otherwise, and likely better to make sure SR is up. With the way FB works, it is probably a significant win to FB with OoC if you don't have the glyph.

This will obviously need to be revisited at 85 with lower levels of crit and haste; my guess is that making sure you do a CP-generating move with OoC if you're less than 5 CPs is going to be the most critical thing, as you can't waste CPs like you can now.

tangedyn · 10/17/10, 11:41 PM

Originally Posted by korkut

Well, nothing changed. I only see gray color in online/offline applet. ( i downloaded Mew-20101017.zip and tried, result is same). If i am the only one that cant get it work, just nvm. Probably weird problem.

Strange. Is anyone else using Windows 7 with Mew?
There's going to be a rewrite of the UI soon, probably using Swing instead of Pivot, hope that will work better for you.

Leafkiller · 10/18/10, 1:50 AM

Originally Posted by Deathwing

You mentioned that there were comments at the top about OOC procs, but I don't see anything. Can you give some details as to why you didn't prioritize shred for OOC?

Very simple - if I put in code to Shred on an OOC proc (right after the Blood in the water code) it calculates as a 33dps loss.

If I add a check to not cast Rip on an OOC proc , there is no dps difference. This may change if I put Rip clipping back in.

As a side note, putting Rip clipping back into my sim script puts it at 28dps above the current default script (using Yawning.mew). Also, if I move the Shred on OOC to the bottom of the default script, it's dps goes up by 26dps. So the two rotations are dps-wise very close (2dps) and the Shred on OOC should be moved to the bottom of the default script. (I tested that while typing this post...go figure).

korkut · 10/18/10, 9:56 AM

Originally Posted by tangedyn

Strange. Is anyone else using Windows 7 with Mew?
There's going to be a rewrite of the UI soon, probably using Swing instead of Pivot, hope that will work better for you.

I created a Virtual XP for Mew :P
Its working on this virtual xp, prolly have problem @ win7 or its just me. Anyway, now i can reach it

Mihir · 10/19/10, 6:31 PM

I noticed you're using the "crit suppression" in the mew calculations, wasn't it confirmed in Crit depression and combat table that crit suppression doesn't exist and was a result of incorrectly parsing partial blocks as hits by recount? Or am I mixing things up now?

in CharacterModel.java:

    public double getMeleeCritChance() {
        double pAgiCrit = getAgility() / mLevelDep.getAgiConv() / 100;
        double pCritRCrit = (mToon.critRating + bonusCritRating) / mLevelDep.getCritConv() / 100;
        double pCritSuppression = mLevelDep.getPCritSuppression();
        double pRawCrit = bonusMeleeCrit + pAgiCrit + pCritRCrit;
        return Math.max(pRawCrit - pCritSuppression, 0);
    }

In LevelDep.java:

    protected double pCritSuppression = 0.048;

Yawning · 10/19/10, 6:51 PM

Originally Posted by Mihir

I noticed you're using the "crit suppression" in the mew calculations, wasn't it confirmed in Crit depression and combat table that crit suppression doesn't exist and was a result of incorrectly parsing partial blocks as hits by recount? Or am I mixing things up now?

There (as far as I know, and according to the rogues Cataclysm Mechanics Testing) still is a penalty that applies to crit rate because of level difference. The partial block thing addressed if it was a penalty or if it was something that forced 4.8% hit onto the attack table.

But now that I'm looking at the code it does need to be capped off at 100%, though that's an edge case that shouldn't occur in realistic uses of the app apart from "Hey, what happens if I have over 9000 Agility?". (And fixed in svn, thanks for getting me to look over this.)

Mihir · 10/19/10, 7:19 PM

Alright, thanks for the quick reply

Updated it in my mmo-champ cat guide, got this formula now to calculate how much crit rating you would need to hit the crit cap, given your unbuffed agility in casterform, and your % hit and % expertise: (51.24 + hit + exp - (agi / 75.45)) * 45.93.

Edit: forgot to factor in agi food, 51.77 should be 51.24.

Yawning · 10/21/10, 8:46 AM

Mew-20101021 is now available.

Major changes since last release:

Relative stat value calculation uses a stat scale of 1 for Formulation.
Coefficients updated to exact values.
Level 85 support enabled again.
The previous versions missed the nerf to Hyperspeed Accelerators. (340 -> 240 haste rating)
A checkbox now exists that will allow the user to disable RSV generation. (Useful when evaluating rotations.)
Yellow crit is correctly capped at 100%.
Standard Deviation is reported for Simulator results.

Comments, feedback, bug reports welcome.

Achnom · 10/21/10, 2:56 PM

Mew rocks, I love it. Thanks for creating it.

Though I still for the life of me, can't get to your models DPS output when hitting a dummy, yes, I realize I'm not a computer.

I'm using an Ovale script that exactly mirrors your Mew script, and I get ~94% of the Mew DPS, very, very consistently over 5 minutes with just MoTW.

The percent of damage from each attack is almost spot on, with just two small differences, Bite and white damage.

My damage from Bite is usually double what the model's output is, and I'm not sacrificing uptime for extra Bites. No complaints there.

The average white attack from the model looks too high. Maybe it has to do with the white crit percentage? When the output says, Crit % for White: 67.97%, that's before misses, dodges and glancing blows, right? I calculate my actual white crit % to be 59.3% when it's all said and done. If you're calculating the white crit % correctly, then I think your normal white hits calculation is approximately 11-12% too high in the model.

The model says the average white hit of 1624, which given my crit of 68.8%, 24% glancing blows, chance to miss/dodge at 8% means a normal white hit would be 1093. However, I hit the dummy for roughly 980 per normal hit over a five minute period with 100% uptime on SR.

tangedyn · 10/21/10, 3:44 PM

I believe boss mobs have 4.8% crit suppression. If not for the crit suppresion you would be crit capped at 68% and will never see a normal hit. So your hit table should look like this:
8% miss/dodge
24% glance
64% crit
4% normal
980 * (0.04 + 0.64 * 2.266 + 0.24 * 0.75) = 1637, less than 1% off

59.3% actual crit is interesting. If this is actually accurate, it could be that Blizzard has also ninja-modified normal melee attacks to use the two hit tables... 0.92 * 0.64 = 0.588 May be worth doing further tests

Not sure about the FB damage discrepancy... maybe Yawning has an idea

Achnom · 10/21/10, 5:35 PM

Well, it seems that over the course of five minutes, I didn't have enough melee hits to get to the true averages, plus the white hits themselves can be skewed by DBW procs.

I downloaded the latest version and am not seeing the standard deviation results. That would be very telling.

edit:

Quick question, is the model output for DPS the mean of all the simulations?

tangedyn · 10/21/10, 11:42 PM

Originally Posted by Achnom

Well, it seems that over the course of five minutes, I didn't have enough melee hits to get to the true averages, plus the white hits themselves can be skewed by DBW procs.

I downloaded the latest version and am not seeing the standard deviation results. That would be very telling.

edit:

Quick question, is the model output for DPS the mean of all the simulations?

If you are running Simulations, you should now see the standard deviation (in the line right after mean DPS +/- 95% confidence interval) with the latest version of Mew.

The reported mean is the mean of all (default) 10k iterations of the baseDPS. Doesn't include the additional sims run to calculate RSVs, of course.

P.S. don't suppose you're familiar with Java?

Achnom · 10/22/10, 5:02 PM

Well, Master Shapeshifter was checked inadvertently and causing the large discrepancy between the model results and my actual dps. I'm within one standard deviation now (unfortunately to the lower side) of the simulation results. I didn't think my play was that much less than perfect, hehe.

Time to start writing custom scripts!

No, Tangedyn, I am not familiar with Java, although the code isn't too hard to follow. My coding experience ended after a few semesters of C++ about ten years ago. I just stick to excel spreadsheets now, mostly, financial models.

Mjoedgaard · 10/22/10, 8:22 PM

I have a problem with both the online and offline version. nothing happens when I press run calculations.