Hunter Spreadsheets -- In development - Page 65

Indora · 03/01/08, 6:37 PM

130 - 50 = 80

100 - 80 = 20

Midnight · 03/02/08, 6:52 AM

Well you´re right in that neithet Lactoses (at least what I can see of it) or my testing can prove that this calculation is wrong, but I think it´s highly unlikely. There should be an easy way of finding out though. Strip naked so both your chance to crit is very low and your chance to miss is very high. If you can get a lower crit than miss chance you shouldn´t be able to get any crits at all with that system.

Indora · 03/02/08, 12:30 PM

The problem we have is 'Does misschance decrease my critchance?', am I right?

Tests like Lactose's one can only proof that hunters doesn't use a 1-roll-system like meeles. But they do not necessarily proof that misschance reduces critchance!
There can be a lot of other system, which allow misses with 100% critchance but doesn't decrease the critchance if critchance + misschance < 1.

The only way we can (nearly) surely say that crits can be misses are tests with 10.000 and more attacks.
Of course there's a chance that the result of the test is wrong. But it is very very low, definitely litte enough.

Glaurong · 03/02/08, 7:31 PM

Miss does have priority on the melee table. When glancing blows were 40% it used to act as a "crit cap."

25% Miss
40% Glance
5% Block
5% Dodge
5% Parry

Only left 20% for crit since it has the lowest priority. Any crit higher than 20% did absolutely nothing.

I'd need to read over the tests again to see if they proved ranged attacks behaved differently but for my first read through I agreed.

ohrion · 03/03/08, 12:06 PM

Originally Posted by Kathucka

Um, why not?

Imagine some total BS numbers: 130% crit rate and 50% miss rate, with everything left over as hits.

The misses (in my theory) then push some crits off of the table, to make room for themselves.

That leaves 80% crits, 20% misses, no normal hits, on a single-roll table So, there you go: Misses with over 100% crit rate on a single roll.

That might be different than melee attack table behavior, but we already know that the ranged attack table is different, as it doesn't have parries or dodges.

With respect to the testing thats been done your theory would produce the following numbers:

107.2% crit rating - (5% hit, whatever chance to miss is on a 68 mob) > 100 % chance to crit. So in the event that miss pushes crit off the table in a single roll system you should still see 100% criticals on Dr. Boom for Midnight's testing.

Unless of course the system caps any single rating at 100% in which case we're back to either one being possible.

PiousFlea · 03/03/08, 12:53 PM

In a 1-roll system, with 100% (or more) crits you shouldn't have a single shot that is a miss.
The test being done the way it has been, and misses appearing, it can not possibly be a 1-roll system.

Originally Posted by Glaurong

Someone did this with over 100% crit rate on the test realm with the stat stacking bug. With 0 hit rating they were still able to get misses and blocks.

The melee attack one-roll system puts priority on avoidance, with hit having the lowest priority. A warrior using Recklessness gets 100% crit chance, but you can still get misses, blocks, glancings etc. (in fact, against a warrior with shield block, you won't get any white crits!)

Seeing misses and blocks with 100% crit chance should not be any surprise whatsoever. Avoidance cancels crits under both a one-roll and a two-roll system; in the former you are pushing crits off the hit table, in the latter if you fail the first roll (you miss) you never get the chance to roll for crit.

Kathucka · 03/03/08, 5:50 PM

Ok, I'm going to go out on a limb here and declare that we can't yet be absolutely sure what the mechanics of the miss/hit/crit table are in terms of one-roll versus two-roll. Even if the tests under weird and/or buggy situation seems to point to one particular result, we can't be sure that result really applies to normal conditions, or whether it is the result of being in a weird and/or buggy situation.

Ideally, someone at Blizzard would come clean and explain the whole mess. Since that's probably not going to happen, I assert that this is the best test we can hope for:

Fire thousands of autoshots at some target with a high miss rate and high crit rate. Analyze the numbers.

Perhaps, take a level 64 hunter out to meet Dr. Boom. That's a 12% miss rate, right? Gear the hunter with lots of crit gear, but no hit gear. If the hunter has a 20% paper doll crit rate, that's a 16% crit rate against a target 4 levels above him. (Apologies if these numbers are wrong. They're from memory, and I'm only using them as an example).

Fire a lot of shots. 5,000 would be enough. More would be better. Fewer would probably be OK. Feel free to calculate intervals of confidence.

Look at the numbers. If you get 16% crit, 12% miss, and 72% hit, it's a single-roll system. If you get 14% crit, 12% miss, and 74% hit, it's a two-roll system. A higher miss rate would separate the two possibilities more and therefore require fewer shots to get a reliable result. Dueling someone who will put Scorpid Sting and/or Insect Swarm on the test subject might help, although that sure would be tedious.

I have no idea what to do about blocks, except conclude that mob target attack tables are not the same as player target attack tables (which we already know). Also, I remember from my youth that ranged attacks against targets much higher in level were sometimes deflected. I have no idea if that's still in there.... Really, you'd have to run the test over with a different %MISS but the same %CRIT, and see if the count changed on crits and blocks.

Then, there's the possibility that auto shot and special shots use different attack tables. Auto and special attacks certainly use different tables for melee.....

Of course, this requires finding a level 64 hunter, getting him to cooperate, and firing away for around three hours. I'm not sure who is going to bell that particular cat, but a fast, low-power ranged weapon would sure be a good idea. Maybe dropping a whole lot of gold on the test subject would be an appropriate thing to do....

Kathucka · 03/03/08, 6:05 PM

Originally Posted by ohrion

With respect to the testing thats been done your theory would produce the following numbers:

107.2% crit rating - (5% hit, whatever chance to miss is on a 68 mob) > 100 % chance to crit. So in the event that miss pushes crit off the table in a single roll system you should still see 100% criticals on Dr. Boom for Midnight's testing.

Ah, I understand the reasoning. Thank you.

Still, there is the possibility that, with full table where %MISS + %CRIT > 100%, misses are put on the actual roll table before crits are, with the extra crit falling off the table.

Now, if you got %CRIT < 100%, but %MISS + %CRIT > 100%, and then managed to get all three of crits, hits, and misses, that would be solid evidence of a two-roll system.

From what I know of Blizzard and of programming, my wild guess is that the correct answer is "none of the above". There's probably some weird exception in the case of debuffs, and another weird exception in the case of the table going over 100%, with even weirder exceptions when both of those happen at the same time.

ohrion · 03/04/08, 9:56 AM

So far from testing we have:
if you have a higher crit rate than hit rate you can still get hits. (Lactose i believe)
if you have a > 100% crit rate you can still get misses. (Midnight)

Those two test suggest that a two roll system is being used.

I believe Midnight saw something like 3% misses which is probably about what you would expect on a level 68 mob, some blocks, and the rest were crits which would make sense if either miss could not be pushed off the table *or* there is a two roll system.

Lactose saw that you could still get hits if your crit rate was greater than your hit rate. This tends more towards the two roll system unless there is some other reason for the hits, defense/resilience/level difference on the target, which I don't believe was the case.

Vito · 03/04/08, 10:14 AM

I would like to ask your help about Cheeky's spreadsheet. Until today i used an older version and downloaded version 44 today. I loaded my gear and my talents into both version and started to compare ranged weapons, so i could choose which one to use. After every choice i re-generated the shot rotations. I'm 0/21/40 specced, so i went for the 1:1.x rotation in all case.
My problem is the following:
The 2 versions of the spreadsheet shows totally different numbers on dps and i'm confused which weapon to choose. My prefered choice was the Gladiator's crossbow, it won the competition in the older spreadsheet, so i started to farm honor

But in the new version, it's the weakest. Anyone knows why is that? As MM/SV hunter i think i should go for a hard hitting slower weapon and that crossbow is the slowest tbc epic weapon atm. But numbers are confusing, here is the list of weapons and the results in dps with the old and with the new spreadsheet:

Wrathtide old: 745
Wrathtide new: 768

Steelhawk old: 733
Steelhawk new: 802

Gyro old: 722
Gyro new: 789

Sunfury old: 759
Sunfury new: 804

Gladiator's old: 761
Gladiator's new: 766

Wolfslayer's old: 728
Wolfslayer's new: 797

2.4 badge reward old: 767
2.4 badge reward new: 824

I checked gear, talents, rotations, everything more than 5 times. I appreciate every suggestions, replys. Thanks.

Tomcool · 03/04/08, 1:47 PM

I just started using this spreadsheet about a week ago and love it. Question I have is after I calculated my shot rotation into my DPS, the highest one seems to be the 1:1.X. Is there a macro to spam for this rotation and do ppl who use it have mana issues on mana intensive encounters? BM hunter btw.

Thanks

Indora · 03/04/08, 4:28 PM

Originally Posted by Kathucka

Now, if you got %CRIT < 100%, but %MISS + %CRIT > 100%, and then managed to get all three of crits, hits, and misses, that would be solid evidence of a two-roll system.

I still think all tests of this kind can by manipulated by blizzard's weird calculation, but this at least make sure that the typical 1-roll system isn't used.

Sadly, it doesn't necessarily mean that crits can be misses... as long as we doesn't know anything about blizzards calculations we can just imagine systems and hope that they reflect the truth.

The problem is: As long as we can't surely say the one-roll or two-roll system exists, we maybe base all our tests on wrong assumptions!
Because of this, I think the best way to solve the problem is a very huge test with a lot of shots.

Glaurong · 03/04/08, 5:03 PM

Originally Posted by Lactose

My stats:
35.83% Crit
48 Hit Rating (~3.04% Hit)
Banshee Curse: -66% Hit
Adjusted Hit Chance: 95 + 3.04 - 66 = 32.04
Target: Level 70 (Shadowy Necromancer) - no debuffs on mob affecting Hit

In this scenario, my crit chance is almost 4% higher than my hit chance.
Using a 1-roll system, there is no way I would see anything other than MISS and CRIT on a Level 70 target.

Stripped combat log:

1/30 21:28:24.412  You are afflicted by Banshee Curse.
1/30 21:28:25.583  Your Steady Shot missed Shadowy Necromancer.
1/30 21:28:25.802  Your Auto Shot missed Shadowy Necromancer.
1/30 21:28:27.240  Your Steady Shot missed Shadowy Necromancer.
1/30 21:28:27.552  Your Auto Shot hits Shadowy Necromancer for 843.
1/30 21:28:28.833  Your Steady Shot missed Shadowy Necromancer.
1/30 21:28:30.099  Your Auto Shot missed Shadowy Necromancer.
1/30 21:28:31.068  Your Auto Shot missed Shadowy Necromancer.
1/30 21:28:31.896  Your Steady Shot missed Shadowy Necromancer.
1/30 21:28:32.365  Your Auto Shot hits Shadowy Necromancer for 827.
1/30 21:28:33.568  Your Steady Shot missed Shadowy Necromancer.
1/30 21:28:34.240  Your Auto Shot missed Shadowy Necromancer.
1/30 21:28:34.427  Your Multi-Shot hits Shadowy Necromancer for 989.

I stopped the combat log shortly after this. Banshee Curse was still on me at the time, which is why a 'Banshee Curse fades from you' message is not displayed here.

Hunter shots are 2-roll, both Auto Shots and specials.

I went back and read Lactose's post again. I remain convinced hit and crit are on separate rolls for all of our attacks.

Kope · 03/04/08, 5:48 PM

Deathblow X11 Goggles - Items - World of Warcraft (available now)

and

Quad Deathblow X44 Goggles - Items - World of Warcraft (available in sunwell)

both need to be added to the head options.

ghungadihn · 03/04/08, 5:53 PM

Originally Posted by Kope

Deathblow X11 Goggles - Items - World of Warcraft (available now)

and

Quad Deathblow X44 Goggles - Items - World of Warcraft (available in sunwell)

both need to be added to the head options.

I don't believe hunters can make these, they default to the mail ones I'm pretty sure.

Kope · 03/04/08, 5:57 PM

Wow, I must be mistaken, I assumed there was a way for us to get them as well.

Lame imo :\. Sorry for the confusion.

Kathucka · 03/04/08, 6:05 PM

Originally Posted by Glaurong

I went back and read Lactose's post again. I remain convinced hit and crit are on separate rolls for all of our attacks.

Notice that there are no crits at all, though! Using a pure two-roll system, there should be both hits and crits here. We don't have the whole story, yet. Blizzard is up to something funny. My guess is that there is something weird about the way debuffs that make you miss are calculated into the system, such that they eliminate crits. That could happen under both a two-roll and one-roll system.

To make it worse, Blizzard could still do something under a two-roll system to give you a consistent number of crits, regardless of your %hit.

Here's one weird way they could do it: Imagine that you have a hit% of 60% and a crit% of 30%.
Roll 1: 1-60: Go to roll 2. 61-100: Miss.
Roll 2: 1-50: Crit. 51-100: Hit.

End result: 30% crit, 30% hit, 40% miss on a two-roll system, with exactly the same percentages as we'd get in a one-roll system.

What is meaningful is that we have an example of a blocked crit on a special. That strongly indicates that blocks and crits are on separate rolls for ranged specials.

We need more data. Blizzard does so much weird stuff that two extreme examples don't give us the whole story.

Glaurong · 03/04/08, 7:21 PM

Originally Posted by Kathucka

Notice that there are no crits at all, though! Using a pure two-roll system, there should be both hits and crits here. We don't have the whole story, yet. Blizzard is up to something funny. My guess is that there is something weird about the way debuffs that make you miss are calculated into the system, such that they eliminate crits. That could happen under both a two-roll and one-roll system.

To make it worse, Blizzard could still do something under a two-roll system to give you a consistent number of crits, regardless of your %hit.

Here's one weird way they could do it: Imagine that you have a hit% of 60% and a crit% of 30%.
Roll 1: 1-60: Go to roll 2. 61-100: Miss.
Roll 2: 1-50: Crit. 51-100: Hit.

End result: 30% crit, 30% hit, 40% miss on a two-roll system, with exactly the same percentages as we'd get in a one-roll system.

What is meaningful is that we have an example of a blocked crit on a special. That strongly indicates that blocks and crits are on separate rolls for ranged specials.

We need more data. Blizzard does so much weird stuff that two extreme examples don't give us the whole story.

There isn't any precedent or evidence to suggest this though, you just came up with a "wierd way" to do things. Which doesn't make any sense, why would the second crit/hit roll be 50/50?.

For every other combat table miss chance makes up a component that can be reduced, hit chance by itself cannot be increased. This is an important distinction.

Look at Lactose's log snippet again.

He is firing at a lvl 70 monster so he has ~5% base miss rate. Miss has the highest priority in other attack table generation so, in a one roll system, before the curse he would have:

~5% Miss
~5% Block
~35.5% Crit

The rest is hit by default (hit chance is purely determined by the first three percentages).

After the curse is applied, in a one roll system, he would have:

~71% Miss
~5% Block
~24% Crit

He has a 0% hit chance since it is purely determined by the other percentages, even part of his crit is pushed off the table. However he still got hits, this is impossible if Crit and Miss are on the same table.

So, in a two roll system before the curse he would have:

~5% Miss
~5% Block
~90% Hit (of those hits, ~35.5% will Crit, effectively ~32%)

After the curse, in a two roll system, he would have:

~71% Miss
~5% Block
~24% Hit (of those hits, ~35.5% will Crit, effectively ~8.5%)

This is very similar to the way it works for spells and is the logical conclusion given the log snippet. You can make up all kinds of nonsensical ways to account for the misses but you can't explain why it would happen that way, or demonstrate a precedent.

Kathucka · 03/04/08, 7:42 PM

Originally Posted by Glaurong

There isn't any precedent or evidence to suggest this though, you just came up with a "wierd way" to do things. Which doesn't make any sense, why would the second crit/hit roll be 50/50?.

That would be the pre-calculated value to ensure that the end result would be 30% crits, consistent with the stat sheet and with the comment from Blizzard about 30% crit rate always getting you 30% crits. It's just something I made up as a hypothetical mechanic to make a point. Don't pay further attention to it.

~24% Hit (of those hits, ~35.5% will Crit, effectively ~8.5%)

But there are no crits. That could be an effect of the small sample size, or an effect of some unknown mechanic.

In the end, though, the real question we want answered is, "How does %hit affect the actual number of crits and blocks you get?" The point I'm making is that knowing whether we have a one-roll or two-roll system is not enough information to answer this question, since there could be other factors. The definitive answer can come only from a big test to see if changing the %hit changes the actual crit rate, or from an explanation from a Blizzard developer.

Glaurong · 03/05/08, 1:16 AM

Originally Posted by Kathucka

That would be the pre-calculated value to ensure that the end result would be 30% crits, consistent with the stat sheet and with the comment from Blizzard about 30% crit rate always getting you 30% crits. It's just something I made up as a hypothetical mechanic to make a point. Don't pay further attention to it.

But there are no crits. That could be an effect of the small sample size, or an effect of some unknown mechanic.

In the end, though, the real question we want answered is, "How does %hit affect the actual number of crits and blocks you get?" The point I'm making is that knowing whether we have a one-roll or two-roll system is not enough information to answer this question, since there could be other factors. The definitive answer can come only from a big test to see if changing the %hit changes the actual crit rate, or from an explanation from a Blizzard developer.

It did change the crit rate, in Lactose's test a 1 roll system would have only produced crits, the crit rate was reduced by the two roll mechanic, allowing hits.

Edit: The only thing garnered from the test was demonstrating crit and hit are on independent rolls. We structured those rolls based on how spell tables are generated. It could be some whacky other way, but that is highly unlikely.

◊ Lactose · 03/05/08, 9:32 AM

I sent this in a PM to Cheeky ages ago before he changed to a 2-roll system for the spreadsheet. I thought I had posted it here as well, but apparently I forgot.
HIT, CRIT and MISS all observed.
---
Just a bit more data I was able to grab out of my combat log, turns out I didn't disable it as quickly as I'd thought. I'm condensing the misses to save space.

1/30 21:28:24.412  You are afflicted by Banshee Curse.
Steady, Auto, Steady: MISS
1/30 21:28:27.552  Your Auto Shot hits Shadowy Necromancer for 843.
Steady, Auto, Auto, Steady: MISS
1/30 21:28:32.365  Your Auto Shot hits Shadowy Necromancer for 827.
Steady, Auto: MISS
1/30 21:28:34.427  Your Multi-Shot hits Shadowy Necromancer for 989.
Auto: MISS
1/30 21:28:38.068  Your Auto Shot crits Shadowy Necromancer for 1716.
Steady, Auto: MISS
1/30 21:28:42.287  Your Auto Shot hits Shadowy Necromancer for 718.
1/30 21:28:44.458  Your Auto Shot hits Shadowy Necromancer for 751.
1/30 21:28:46.490  Your Banshee Curse is removed by Anyak 's Remove Lesser Curse.

Mattaos · 03/05/08, 12:22 PM

Could I get a clarification on the amount I should plug-in to the spreadsheet for latency based on 300ms?

EDIT: Found the solution I needed.

Indora · 03/05/08, 12:30 PM

Originally Posted by Kathucka

In the end, though, the real question we want answered is, "How does %hit affect the actual number of crits and blocks you get?" The point I'm making is that knowing whether we have a one-roll or two-roll system is not enough information to answer this question, since there could be other factors. The definitive answer can come only from a big test to see if changing the %hit changes the actual crit rate, or from an explanation from a Blizzard developer.

That's exactly what I want to say, too.

Don't try to test 1- or 2-roll-systems as long as you don't even know if exactly this system is used!
Just test whether your critchance is reduced by misschance.
Nothing else.
The simplest and only way to check this is a huge test with 10.000 and more shots.
Of course there is a chance that the result ist wrong. But as long as this chance is very low (<1% or something like this) we can be quite sure. Feel free to even calculate the chance of the result being wrong.

How did the caster find out that their spells are using a 2-roll-system?
If they have reliable calculations that the 2-roll-system works this way and no other, (there can be a LOT of alternatives) we can do tests like these.
To my knowledge they only know that misses reduces their critchance and don't know whether 2 rolls are used!

RogueLeaderX · 03/05/08, 1:55 PM

Originally Posted by Kathucka

Perhaps, take a level 64 hunter out to meet Dr. Boom. That's a 12% miss rate, right? Gear the hunter with lots of crit gear, but no hit gear. If the hunter has a 20% paper doll crit rate, that's a 16% crit rate against a target 4 levels above him. (Apologies if these numbers are wrong. They're from memory, and I'm only using them as an example).

Of course, this requires finding a level 64 hunter, getting him to cooperate, and firing away for around three hours. I'm not sure who is going to bell that particular cat, but a fast, low-power ranged weapon would sure be a good idea. Maybe dropping a whole lot of gold on the test subject would be an appropriate thing to do....

Wouldn't it also be possible for a level 70 to go out, keep their crit% static (or as close to it as possible) but change their hit% and see if their crit% goes up? Violet Signet of the Master Assasin comes to mind. Any other easily obtainable gear that has HR but no agi or CR? Maybe putting a cenarian helm enchant on some crappy 'of the owl' green? It'd cost less than paying a level 64 1k for his time.

That would give us a test without any buffs or debuffs on the target or player that potentially change how things work.

Kathucka · 03/05/08, 8:36 PM

Lactose wrote:

1/30 21:28:38.068 Your Auto Shot crits Shadowy Necromancer for 1716.

Well, that says a lot. If you're getting both crits and hits with a 66% miss debuff, it can't be using a standard single-roll table in that situation. With that %miss, either %crit or %hit would push the other off the table.

I still would like to see a big statistical test, so we know we have the full story.

RogueLeaderX wrote:

Wouldn't it also be possible for a level 70 to go out, keep their crit% static (or as close to it as possible) but change their hit% and see if their crit% goes up?

The problem with that is that Dr. Boom is level 68, leaving only a base 3% miss rate for a level 70 hunter. With an adjusted 20% crit rate, and firing 1,000 shots, that's the difference between 200 and 194 crits. That is well within normal random variation.

In other words, with a 70, you'd have to fire a hell of a lot of shots to make the results statistically meaningful.

With a level 64, though, you'd get a base 12% miss rate. The difference with 1000 shots and a 20% adjusted crit rate would be expected values of 200 or 176. You'd need far fewer shots to overcome random variation.

Where the heck is Olgas? This sort of thing is right up her alley....