Archived

This topic is now archived and is closed to further replies.

Yellowsix

Team Robot Simulator and Gear Comparison Tool

72 posts in this topic

Feral Spirits should get 5% from LoTP, 3% from Heart of the Crusader and then they get SoE as well. Turning off all raid buffs shows the wolves with 1.35%. Adding SoE brings them to 3.45% and then add the 8% from raid buffs and you get the 11.45% shown.

11.45% correlates pretty well with the 12.16% crit rate I saw from them on my last Festergut kill so seems EnhSim is working as intended.

Shock and Awe should turn on your pet panel for the wolves when they are summoned so you can easily confirm their base stats and their stats when buffed with SoE. They can also be buffed by things like scrolls/battleshout/blessings but besides battleshout it generally isn't worth the effort.

There seems to be a difference in real world observed uptime and the simulated uptime (in both TRS and enhsim). This is either through some unnacounted for human reaction or a difference in the actual mechanic and what we see in on the tooltip.

Which is exactly what I was trying to talk about. Thing is that the 4T10 mechanic conforms exactly to it's tooltip, I've confirmed that with dummy testing so that leaves Human Reaction time and Maelstrom Weapon efficiency to explain the discrepancy. There is a large variance in Maelstrom Power uptime even at similar efficiency levels but that seems more likely to be due to the small sample size inherent in looking through parses.

Still efficiency is something easily measured from the outputs of EnhSim or TeamRobot and by varying latency settings and moving MW5-LBs position in the priority list I can get the uptimes to correlate better. EnhSim only lets me adjust latency as a solid block though so it effects everything which was why I was hoping TeamRobot's additional settings options would let me fine-tune that more for my own individual performance.

The 200ms spacing after the 5th charge is reached but before it enters the priority list is a great start but it also needs to be applied to more reactive things and should have latency added to it as well since the report has to reach your computer from the server before it can be responded to as well. The additional reactive things would be LS and Flame Shock and then anything with a conditional such as SS_0 or LL_QE but I don't think those types of conditional priorities are incorporated into TRS yet anyway (and we should probably stop using them in EnhSim until we get the latency/reaction time mechanics nailed down better as well).

Flame Shock is also an interesting one since it would only be reactive if you are actively trying to prevent clipping DoT ticks. If you use a set ratio of FS to ES then it is just a cooldown ability (Shocks) without any reactive component. I'd love to be able to simulate differences on that level to see if it's just better to stick to a set ratio or stay with trying to be reactive in when to Flame Shock.

Mind filling us in a bit more on how exactly TRS simulates the client/server interaction and how it relates to ability use from the priority list? Also what your plans are regarding Reaction Time and if it will be able to simulate things to the level that I could use it to define whether a set ratio is better for Shocks or if I should stick to reactive Flame Shocking.

Share this post


Link to post
Share on other sites

Also it looks like there may be something wrong with the way Windfury is being modeled.

5 Minute Fight

Festergut HM kill: 94 Windfury Attacks

World of Logs - Real Time Raid Analysis

Mr Robot says: 38 Windfury Attacks

http://i50.tinypic.com/i3fp7k.png

3 Minute 50 Fight

Festergut normal kill: 62 Windfury Attacks

World of Logs - Real Time Raid Analysis

Mr Robot says: 30 Windfury Attacks

http://i45.tinypic.com/2llksgx.png

Simulations were done with the same gear I have on in my armory, aside from the fact that I switched black bruise heroic out for black bruise normal since that is what I was wearing for both kills.

Share this post


Link to post
Share on other sites

Hmm, we'll have to look into this. On first glance, I am thinking that maybe our reporting is treating both WF attacks as "one" count? The % of total damage seems to be in a plausible range. It looks like maybe it is being reported as one hit that is twice as hard, instead of two hits...

Share this post


Link to post
Share on other sites
Hmm, we'll have to look into this. On first glance, I am thinking that maybe our reporting is treating both WF attacks as "one" count? The % of total damage seems to be in a plausible range. It looks like maybe it is being reported as one hit that is twice as hard, instead of two hits...

If that is the case you are going to have to rework it out to separate hits since each is individually capable of critting and proccing additional effects like Maelstrom Weapon, Static Shock, Berserking and trinkets like Tiny Abom.

Share this post


Link to post
Share on other sites

Yes, I was careful to say it is a reporting error. We have modeled each WF hit, as far as I am aware.

Share this post


Link to post
Share on other sites
Yes, I was careful to say it is a reporting error. We have modeled each WF hit, as far as I am aware.

Might want to check on that too. I just simulated out the gear I used for my last Festergut (with the exception of swapping my Seducer offhand to another Slicer because of the Orc racial missing fist weapons). It shows 42 Lightning Orb attacks and then another 5 from the 2T9 bonus (47 total) but I had 57 from my parse. If you add up all the melee attacks (doubling up for Windfury) there would be 556 attacks which should equal 50 procs (2T9). Without doubling up it would be 524 melee attacks which would be 47 procs.

Share this post


Link to post
Share on other sites

It looks like the two windfury hits are getting lumped together as a single event... I'll fix that.

I'll make sure to adjust the feral spirit crit rate. If I'm reading this correctly, feral spirits probably do not suffer from the 4.8% melee crit suppression against bosses that normal players do?

Regarding latency and reaction time:

Wouldn't flame shock never be a "reactive" ability? You can use some kind of mod to track the DoT duration, and anticipate when it will end, thus making it not a purely reactive choice? I haven't played enhancement myself, so it's possible that I'm missing something.

Right now, latency is handled in a pretty simple way: if you say "do flameshock" at time 2.0, and your latency is 200ms, it actually executes flameshock at time 2.2. I think that is the only use of latency right now.

With regards to latency... I basically assume that the people who coded the game are a lot smarter than me. For instance, if I have a 200ms ping, and I press a button at time 2.0... I highly doubt that the game does this:

2.0: client executes command

2.2: server receives command and executes it

2.4: client receives response that command was executed, and shows its effects

That would be some pretty crappy network code. I don't even pretend to know the details, but I would hope that the game does some kind of more sophisticated synchronization that can lessen the effects of latency, and deal with the fact that multiple clients with variable latency are interacting in the same environment.

The other piece in the Team Robot code is asynchronous decision logic. It goes something like this:

2.00: player casts stormstrike, triggers a GCD

2.20: stormstrike actually executes (after 200ms latency)

3.25: 250ms before the GCD is up, use the current conditions of the simulator to decide the next action

3.50: execute the next action

That 250ms is hard-coded right now... it's basically the player's "anticipation" time -- how far in advance does the player need to decide in order to time up his next key press with the end of the previous GCD? For casters, a similar mechanic is used: you can chain-cast with the server spell queue, which necessitates deciding which action to use prior to the previous cast completing. 250ms is definitely closer to robot-efficiency, so I might make that a configurable parameter.

Using this current mechanic, let's examine the MW5 scenario:

If the 5th stack of MW is applied at or before time 3.25, the simulator will recognize it, and thus choose LB as the action to perform at time 3.50. Note that as long as your human reaction time is less than 250ms, you don't need a separate reaction time constant to get a decent simulation: you see MW5 proc, make the decision, then you have 250ms to press the button in response to that stimulus. The timing is definitely tight right now, but I think that it's a pretty realistic model overall.

The not-yet-implemented reaction time constant only comes into play in the following scenario: if at time 3.25 the simulator can't find any action to perform, it will wait for the next "significant" action. If the next significant action is the 5th proc of MW, and it happens at time 3.60, the player could not realistically respond to that until time 3.60+[human reaction time].

Note that if the next significant action were one of the following, human reaction time is not relevant: a cooldown finishing, a DoT running its duration and expiring. A human can anticipate these events, and if your timing is impeccable, completely negate reaction time. It's very hard to do sometimes... but it's humanly possible.

One aspect of latency that I haven't tested or modeled, is if there is a latency-related delay between when an auto attack occurs, and procs from that attack happen. e.g. if I auto-attack at time 2.0, does windfury happen at time 2.0 or 2.2? I don't know the answer to this, so I don't model a delay right now.

Share this post


Link to post
Share on other sites
I'm not bad-mouthing EnhSim, I'm simply reporting what I see. This is a pretty minor thing anyway -- every program has its bugs, Team Robot included, EnhSim included. My ultimate goal is to simply model Searing Totem correctly, and comparing to another simulator is a great way to confirm my results.

Here's your proof:

I put in the following priority in EnhSim:

FE

ST

That's it: just use fire elemental first, then searing totem. I did a 5-minute fight:

If only those two abilities are being used, searing totem should be near 60%, with the only reduction coming from the player's reaction time after the fire elemental despawns. Somewhere, 38 seconds of searing totem up-time are disappearing. This indicates either a bug in the up-time report calculation, or a flaw in the priority logic.

Can you confirm from the snipped data that the EnhSim DAMAGE is right but the UPTIME is wrong? ie: is it a problem that its showing the wrong damage or is it that the uptime calc is displaying wrongly?

Re: network lag issues.

It is my understanding that the game implements a Spell Queue for a SINGLE spell in the buffer. This queue is active 250ms before the GCD finishes. ie: if within 250ms of the GCD expiring you press a key that spell goes into the buffer and is executed on the server the instant the GCD finishes. Now of course it takes time for the data about the key press to leave the client and reach the server and if your network lag is greater than 250 its going to arrive after the GCD has finished thus giving you "idle" time. This is why addons like Quartz show red lag bars so you have an idea of when to send the next spell.

The better players learn to press the button up to 0.25 sec before the GCD finishes and thus get off their next ability with zero idle time. If you press the key BEFORE the 0.25 sec before the GCD finishes (eg: 0.3 sec before) then you get the "That ability is not ready yet" error.

The idea behind this 250 ms spell "queue" is to compensate for the effects of network lag. Note that you cannot queue up more than one ability.

Share this post


Link to post
Share on other sites

I did some extensive in-game tests around lag and spell queuing when I was writing the moonkin module. My tests showed the following:

For spells with a cast time, the server-side spell queue eliminates any effects of latency. As long as you send your next spell close to the end of your current cast (people say there's about a 300ms window), the server will perfectly chain your casts with no gaps between them.

For instant abilities that are GCD limited (I use Earth Shock, wait for GCD to end, then use Stormstrike), I didn't notice any queuing mechanism. As far as I could tell, the game would not accept input until the GCD was complete. I tried my hardest to precisely time up my next ability to hit the edges of such a queue... and I simply wasn't seeing it, or it is smaller than 250ms.

That said, a player with good timing can nearly perfectly time up GCD-limited abilities. I was able to chain-cast moonfire for 2 minutes without accumulating any significant delay: over 2 minutes, I cast the theoretical maximum number of spells that was possible at my level of haste.

It can be pretty tough to get timing that tight though... you have to really focus on it. The Team Robot simulator will let you put a little lag after a GCD before the next command is input, which defaults to 50ms. Whether there is a queue around the GCD or not shouldn't matter too much -- the current code wouldn't need to be changed, since it already does "anticipated" decision logic.

Share this post


Link to post
Share on other sites
Can you confirm from the snipped data that the EnhSim DAMAGE is right but the UPTIME is wrong? ie: is it a problem that its showing the wrong damage or is it that the uptime calc is displaying wrongly?

I took a quick look... it might just be an error in the up-time report. The report gives these numbers for searing totem (this is a slightly different run: it is using the exact same priority as the default in the Team Robot simulator, but with searing totem instead of magma, using the character Annhilate of Spinebreaker):


Searing Totem:

miss                0              0.00%

hit                 453642         56.24%

crit                352994         43.76%


...

searing totem uptime           39.10%

...

...

                    DPS                 PPM       MPS

...

searing totem       265.63    2.89%

...

I guess those are numbers of casts? If I take the sim duration (1000 hours * 60 min * 60 sec) and divide by the fight duration (5 min * 60 sec) to get 12,000 runs. Then if I do 453642+352994 / 12000, I get 67.2.

With the same setup (character, priority order), the Team Robot summary produces these numbers for searing totem:

Searing Totem

3.0% total damage

1162 avg damage

787 avg hit damage

1647 avg crit damage

43.6% crit rate

69 casts

58.5% up time

So it looks like the number of casts and percent damage is pretty close, which would indicate that it's just the report that is showing the wrong number. The 1 or 2 extra casts that Team Robot gets is probably a minor artifact of the simulators using different decision logic implementations.

Share this post


Link to post
Share on other sites

A 3.3.3 beta is now available:

Team Robot Wow Simulator Beta

Read the patch notes for the full details: WoW Simulator Change Log

Please post in the following thread if you would like to see a change or bug fix incorporated into the beta -- you will get a much faster turn-around from us: 3.3.3.

Quick notes for shamans: glyph of flame shock was implemented to give a 60% crit damage bonus, even though the tooltip still has the old 3.3.2 text. Let me know if that is not what it actually does on the PTR... I haven't had time to get on the PTR myself, I just used wowhead's ptr section. I was also not clear on how the 2T9 flame shock extension works with haste now. I made it add exactly 3 ticks instead of a static 9 seconds. I also made 4T10 always add exactly 2 ticks when lava burst hits. Let me know if any of these assumptions are not correct.

Also, gear rankings/estimates have not been updated for 3.3.3, just the simulator! Press the simulate button on the beta site!

Share this post


Link to post
Share on other sites

Did a couple sims and found this quite pratical.

A few things I missed.

A way to limit the lists of gear and gems like a thru a filter: item lvl stats etc. Making it faster and easier to pick stuff.

Warnings would also a good thing to implement. For when under expertise cap, hit cap ( you could include the crit cap formula here), warning if meta gem is not activated ( no blue gem equipped).

Also noticed that it says im not expertise capped on stats. I see it uses the blood fury in the sims and it its working like I am capped but doesnt show it on char stats.

Thank you for your effort and keep up the good work,

P.S. Pardon any english error, I´m a bit rusty.

Share this post


Link to post
Share on other sites

Thanks for checking it out and taking time to post suggestions.

Improved/Expanded filters on the item lists is definitely a feature we plan to work in.

I like the idea of having some warnings that could show on the UI if you are under certain important numbers, such as the hit or expertise cap. We'll put it on the list as a UI enhancement to work in once we have all the classes implemented.

We'll look into the character stats not displaying expertise from blood fury.

Share this post


Link to post
Share on other sites

There is something wierd in the enhancement model. So it always shows me that the [iTEM]40125[/iTEM] is the best especially for yellow sockets and the [iTEM]40128[/iTEM] is always 23th orso.

I think there is a bug with haste-rating cause the tool shows me that [iTEM]50458[/iTEM] has less dps than [iTEM]50463[/iTEM] in the list of totems. But when i simulating the haste totem is just the better one as indented. Same with the Gems when i have enough hit for my crit rating and of course hit cap, the haste gems simulating a better dps than the hit gems, but the tool always shows hit gems are the best dps.

Sry for my english i hope you can comprehend.

Share this post


Link to post
Share on other sites

It's possible that hit rating is being over-valued by the estimates once you go over the 8% "soft cap" -- I'll take a look at them and see if anything needs tweaking.

As for the totems... this probably ties into some of the other estimates: it's possible that the estimates are either slightly undervaluing haste, or slightly overvaluing attack power. I'll see if I can determine what's going on, and make some tweaks.

Share this post


Link to post
Share on other sites

We'll look into the character stats not displaying expertise from blood fury.

My phrasing was not very good and led you into error. When I meant was that the sim used blood fury (ap buff), it was for proving that the sim detected my char as an orc. Expertise comes from Axe Specialization racial. Which is accounted for in the sim but not displayed on char stats.

About the the ranking of gear and especially gems, its quite off, I simply ignored it, until the stat ranking is fixed or we can filter them.

Share this post


Link to post
Share on other sites

Weapon-specific expertise buffs like axe specialization will not show up in your expertise total, as they are specific to the hand that is holding the weapon, and not to your "generic" expertise total. As you saw, when the simulation actually runs, it applies the expertise bonus to the hand that is holding the weapon of the proper type.

Share this post


Link to post
Share on other sites
I've posted an update with some tweaks and bug fixes. The full list of changes is at: WoW Simulator Change Log.

- I noticed in some of the threads that people rate [iTEM]Tiny Abomination in a Jar[/iTEM] very highly for enhancement. I must not have some of my information correct, because my simulations are not showing it doing very good DPS at all. For the normal version, my simulations show it charging up and attacking roughly every 8-10 seconds (charges build off of any melee attack [auto, stormstrike, lava lash, etc.] with a 45% proc rate) or about 33 procs in a 5-minute fight. It uses a 50% chance to hit with MH or OH, and I'm using normalized weapon damage. I'm also assuming that it cannot trigger windfury or flametongue, or any other proc for that matter.

With those assumptions... it's not a very good trinket. Are some of those assumptions incorrect?

I didn't see any response to this so -

Bitty Bom Jar *does proc* FT and can proc WF when it is triggered by the appropriate hands.

Share this post


Link to post
Share on other sites

Thanks miril -- I more or less assumed as much, otherwise there would be no way that it could be ranked very highly as a good DPS trinket. I think that FT and WF are triggering off of tiny abom on the live version, but I'd have to double check. I know for certain that it is in my latest test version, which will be posted shortly after 3.3.3 goes live.

Share this post


Link to post
Share on other sites

3.3.3 changes have been posted! A full list of changes is at:

WoW Simulator Change Log

Please note that gear estimates for enhancement are still slightly favoring AP too highly (and/or haste not enough), though it's a tad closer now. Following Mr. Robot's gem suggestions will give you pretty good DPS, but it could be slightly less than optimal for certain gear levels (my tests showed it to be about 0.2-0.5% less than an "optimal" gem setup for certain characters, or about 20-50 DPS for a well-geared player -- so definitely not a huge deal, but worth noting). Make sure to double-check gem choices with an actual run of the simulation until we can determine where the inaccuracy is creeping into the estimates.

Share this post


Link to post
Share on other sites

[iTEM]Royal Scepter of Terenas II [/iTEM] is still showing as an mp5 item for Ele. It was switched to crit/haste shortly after Lich King went live.

Also, are there any plans to support Demonic Pact? Using the ToW values for +spell power is extremely inaccurate for modeling, as most raids maximizing their buffs will be using a Demo Lock at this point, giving drastically different buffing numbers.

EDIT:

Also, adding in Fire Elemental to the custom Searing totem rotation (before Searing Totem) gives a decrease/very minor increase in DPS. It appears to not be using his melee damage or his fire shield in the damage totals.

Share this post


Link to post
Share on other sites

  • Recently Browsing   0 members

    No registered users viewing this page.