05/17/08, 2:35 PM #3751 coredumperror Piston Honda     Hibernicus Tauren Druid   Vek'nilash If you're in a fight where taunt actually works (not very many) you'd be pretty unlikely to survive for more than a few seconds if you're not in full tank gear. Combining your desperately low armor with desperately low HP, your healers probably won't know to be healing you till you're already dead.
05/17/08, 3:46 PM   #3752
Tinweasele
Von Kaiser

Night Elf Druid

Firetree
 Originally Posted by raffy Well, obviously if the proc rate is 4% then, the chance of getting the proc after n attacks is: 1 - (1 - p)^n The other logic I use is like, "if within 1.5 seconds you will get another 20 energy from regen, then include that extra 20 energy in the energy calculation." So it currently looks like: energy = (current Energy) + (20 energy if energy tick occurs within 1.5 sec) + (20 energy if 1-(1-p)^n > 95%) I model p like a lazy moving average, incase the proc rate actually does increase with haste effects. p = ((20 - 1) * p_old + p_new) / 20 where p_new = 1 / number of hits since last proc while wearing 2.4 in Cat form Maybe I should just change it to: p = ((# - 19) * p_old + p_new) / (20 + #) where # is number of procs for the wow session (this way it should eventually re converge to 4% after considerable number of samples.)
extrapolation does that actually block the powershifting? or is it just used for describing lost energy? cause if you are comparing to the threshold then anythreshold under 20 has a .5 window of opportunity to powershift, which seems low to me:-/

05/17/08, 4:03 PM   #3753
Mijae
Don Flamenco

Tauren Druid

Tichondrius
 Originally Posted by raffy Well, obviously if the proc rate is 4% then, the chance of getting the proc after n attacks is: 1 - (1 - p)^n The other logic I use is like, "if within 1.5 seconds you will get another 20 energy from regen, then include that extra 20 energy in the energy calculation." So it currently looks like: energy = (current Energy) + (20 energy if energy tick occurs within 1.5 sec) + (20 energy if 1-(1-p)^n > 95%) I model p like a lazy moving average, incase the proc rate actually does increase with haste effects. p = ((20 - 1) * p_old + p_new) / 20 where p_new = 1 / number of hits since last proc while wearing 2.4 in Cat form Maybe I should just change it to: p = ((# - 19) * p_old + p_new) / (20 + #) where # is number of procs for the wow session (this way it should eventually re converge to 4% after considerable number of samples.)
That's what I'm saying. You cannot model the probability that way. There is no moving average. It doesn't matter how many hits you have had since the last proc. The next attack will always have the same proc chance, you cannot change it to a variable chance. They are independent events.

You are just as well off using a basic average procs per minute and then turning that into a time threshold.
timeThresh = 60 / PPM
OoC = 2 PPM
2t4 = (hitsPerMinute * 4%) PPM

Even this will only give an average time till proc. You will end up being wrong very often.

Either way, I don't see this being very useful. Whether the energy proc happens before or after the powershift, as long as current energy is being checked it shouldn't really matter imo.

05/17/08, 4:46 PM   #3754
raffy
Piston Honda

Night Elf Druid

Suramar
 Originally Posted by Mijae That's what I'm saying. You cannot model the probability that way. There is no moving average. It doesn't matter how many hits you have had since the last proc. The next attack will always have the same proc chance, you cannot change it to a variable chance. They are independent events. You are just as well off using a basic average procs per minute and then turning that into a time threshold. timeThresh = 60 / PPM OoC = 2 PPM 2t4 = (hitsPerMinute * 4%) PPM Even this will only give an average time till proc. You will end up being wrong very often. Either way, I don't see this being very useful. Whether the energy proc happens before or after the powershift, as long as current energy is being checked it shouldn't really matter imo.
Assuming 4% proc rate, if you haven't got a proc in the last n hits, then your chance of getting the proc within the next few attacks increases, since over extremely large n, the proc rate must be 4%.

I do see your point as the PPM example would basically do the same thing, so maybe what I really should be modeling, is like the inverse of proc. For example, after you got a few lucky procs, your chance of getting more lucky procs is decreases as your number of hits increases.

On a longer fight, I'm sure that if you modeled the bloodlust proc you could have some success predicting its occurrence. How much energy would you save? Who knows? But why not try to figure it out? ;p Since we're ultimately limited by mana, avoiding shifting near high probability of Bloodlust procs doesn't seem like that bad of an idea.

05/17/08, 6:30 PM   #3755
Anedris
King Hippo

Troll Priest

Steamwheedle Cartel
 Originally Posted by raffy Assuming 4% proc rate, if you haven't got a proc in the last n hits, then your chance of getting the proc within the next few attacks increases, since over extremely large n, the proc rate must be 4%.
What?

Probability has no memory. If I flip a coin a hundred times and get heads every time, my next flip still has only a 50% chance of being tails. If I flip that coin ten thousand times I can assume I will get approximately 50% heads and 50% tails, but there is no law of causality that dictates I must get that - it is simply probable.

So it doesn't matter how many attacks I've made that failed to proc. The next attack has exactly a 4% chance to proc. The same as the attack after that, and so on. The hits you have already made have zero influence on the hits you have yet to make.

 05/17/08, 7:01 PM #3756 Astrylian Rawr     Celestrylian Night Elf Monk   Stormrage Gambler's fallacy - Wikipedia, the free encyclopedia Rawr!
05/17/08, 7:09 PM   #3757
raffy
Piston Honda

Night Elf Druid

Suramar
 Originally Posted by Anedris What? Probability has no memory. If I flip a coin a hundred times and get heads every time, my next flip still has only a 50% chance of being tails. If I flip that coin ten thousand times I can assume I will get approximately 50% heads and 50% tails, but there is no law of causality that dictates I must get that - it is simply probable. So it doesn't matter how many attacks I've made that failed to proc. The next attack has exactly a 4% chance to proc. The same as the attack after that, and so on. The hits you have already made have zero influence on the hits you have yet to make.
Yeah, but we're not flipping coins in a textbook example, we're trying to model Blizzards RNG for a 4% proc rate based off melee attacks.

Have you ever seen more than two chain Bloodlust procs? Three? Ten in a row? I'm sure Blizzards RNG uses its own output as feedback to keep the random distribution relatively sample-independent to avoid RNG bias in short-interval combat.

The criteria is probably similar to: "a 4% proc over 1000 attacks such that the proc rate of any substring of the attacks falls within a window of 0 to 8%."

Proof of this can be shown by looking at any WWS report and realizing that the overall proc rates of various random actions (crits or procs or whatever) match their expected values within a reasonably small window of error.

ie. no one is sustaining 80% crit with 50% paper doll.

@Astrylian - Gambler's fallacy again implies we know nothing about the distribution (ie. its random) which is not the case for WoW's combat RNG.

 05/17/08, 7:21 PM #3758 Krazen Don Flamenco   Xiz Blood Elf Warlock   Turalyon Hey guys, I'm looking to gear up my friend's feral druid by buying a couple t6 tokens, and hopefully some other BT pieces. He's currently geared out in a set of full t4/badge gear. I'm primarily looking to optimize cat dps, bear tanking is secondary. The World of Warcraft Armory I'm planning on keeping the t4 hat and shoulders, gemmed for dps (cursed vision isn't an option), and he has enough gold to buy 3 pieces of BT/Hyjal loot, maybe 4. Rarer, desirable dps pieces like [Shadowmoon Destroyer's Drape] are also not for sale. What order should he be looking for pieces in? T6 gloves look like the clear best option, since they can be used in both dps and tanking roles. From there, I'm looking the following pieces. [Insidious Bands] T6 legs (enchanted for tanking) T6 chest (gemmed for tanking) For reference, he currently has no dps legs (using t4 legs), and uses [Chestguard of the Conniver] in the chest slot. The Teron bracers will let me regem [Band of the Swift Paw] for tanking purposes. Last edited by Krazen : 05/17/08 at 7:34 PM.
 05/17/08, 7:25 PM #3759 Anedris King Hippo   July Troll Priest   Steamwheedle Cartel Huh? Why would Blizzard implement a bunch of complicated code to make every random element in the game conform to its average over time? People don't get 80% crit rates with 50% paper doll crit because the chances of rolling that many crits over an entire encounter, given a 50% crit chance, are simply very small. You do however see occasional WWSs of warlocks who got lucky with the RNG and did 2900 DPS or whatever due to higher-than-average crit rates. (I don't know what you mean by a "bloodlust proc" since you're obviously not talking about the shaman ability.) When flipping coins, you know the chance of a given event occurring (50%). I don't see why the gambler's fallacy would apply to that but not to a 4% proc chance. (Unless Blizzard's RNG was not, in fact, random at all but designed instead to produce average results over some given interval, which, again, makes little sense to me, though if that is indeed the case then sure, what you say makes sense...) If Blizz did put in this check to make sure occurrences never stray too far from their averages, they obviously didn't implement it for loot drops at any rate...
05/17/08, 8:25 PM   #3760
coredumperror
Piston Honda

Tauren Druid

<SiN>
Vek'nilash
 Originally Posted by raffy Have you ever seen more than two chain Bloodlust procs? Three? Ten in a row? I'm sure Blizzards RNG uses its own output as feedback to keep the random distribution relatively sample-independent to avoid RNG bias in short-interval combat.
Yes, I have in fact seen several chain bloodlust (2t4) procs in a row. I once pulled off 4 procs in under 5 seconds.

I think you're confusing blizzards "ppm" system of putting an internal cooldown on most procs, preventing them from happening too often, with the observed/documented 4%, no cooldown proc rate of 2t4. 2t4 has no internal cooldown, and it is ONLY the luck of the draw that might net you several procs in a row, or none at all for a long period of time.

 05/17/08, 8:37 PM #3761 Astrylian Rawr     Celestrylian Night Elf Monk   Stormrage Indeed, there's no special magic with the proc chance of 2T4. 4% chance on every hit, completely independent, that's it. Regarding buying MH/BT items... If you want to stick to the T6 tokens, due to the high drop rate, yeah, sounds right. If you want to gamble on it dropping, Nether Shadow Tunic, Grips of Damnation, and Shady Dealer's Pantaloons may be slightly better. Ask Rawr to find out. Rawr!
 05/17/08, 8:38 PM #3762 raffy Piston Honda   Edgy Night Elf Druid   Suramar There's no point in arguing about this. All I'm saying is, I believe the results from WoW do not conform to a perfectly random distribution, thus information can be extracted and ultimately predictions can be made about future data with relative accuracy. If I'm wrong, the predictions will not be accurate, however any improvement would result in increased DPS because you'd have 20 more energy per missed proc at the end of the fight. Even if thats only 20 energy, thats still ~2k damage or about 0.5% more damage over the course of an average fight. I am not trying to say I can predict a random variable, so please stop mentioning the freaking coin flip :p
05/17/08, 9:52 PM   #3763
Touf
Don Flamenco

Tauren Warrior

Ner'zhul
 Originally Posted by raffy There's no point in arguing about this. All I'm saying is, I believe the results from WoW do not conform to a perfectly random distribution, thus information can be extracted and ultimately predictions can be made about future data with relative accuracy. If I'm wrong, the predictions will not be accurate, however any improvement would result in increased DPS because you'd have 20 more energy per missed proc at the end of the fight. Even if thats only 20 energy, thats still ~2k damage or about 0.5% more damage over the course of an average fight. I am not trying to say I can predict a random variable, so please stop mentioning the freaking coin flip :p
Wow, wow, wow. Let me guess, you "feel" that this is true? Do you have ANY sort of proof that the programmers would go through about 100 times the work necessary to implement such a system?

Imagine it this way, a 5 minute fight has ~300 attacks. If you look at Exact Confidence Interval for a Proportion
you can get a feel for how hard it is to even get a 5% difference in crit rate with that many samples.

When you go down to 50 samples (most casters) you see outliers all over the place. But still, the odds of getting an 80% crit rate instead of 50% with 50 samples is only 1 in 100,000.

And there is a point, because you're trying to model something on unbased assumptions.

05/17/08, 10:40 PM   #3764
raffy
Piston Honda

Night Elf Druid

Suramar
 Originally Posted by Touf Wow, wow, wow. Let me guess, you "feel" that this is true? Do you have ANY sort of proof that the programmers would go through about 100 times the work necessary to implement such a system? Imagine it this way, a 5 minute fight has ~300 attacks. If you look at Exact Confidence Interval for a Proportion you can get a feel for how hard it is to even get a 5% difference in crit rate with that many samples. When you go down to 50 samples (most casters) you see outliers all over the place. But still, the odds of getting an 80% crit rate instead of 50% with 50 samples is only 1 in 100,000. And there is a point, because you're trying to model something on unbased assumptions.
1. You can implement such a system with a few lines of code.

2. Yes, I'm trying to model something that I cannot prove one way or another about how it works, which is why making an assumption and then attempting to model it and make predictions. If it's truly random, then I cannot model it accurately to extract any useful information. However, if it is modelable (which I believe it is, especially at really low or really high proc rates) and thus not perfectly random and independent), what harm is there in attempting to model it?

3. I find it ignorant and silly to believe that the attack and proc RNGs do not operate with target constraints. So much of this game relies on short bursts of combat, and if the RNG provided truely random results, you'd see far DPS outliers (ie. like 80% sustained crit).

4. Lets continue this discussion in the other thread so the Feral thread can be left alone.

05/17/08, 10:55 PM   #3765
Valerian
King Hippo

Night Elf Druid

Blackhand
 Originally Posted by Krazen Hey guys, I'm looking to gear up my friend's feral druid by buying a couple t6 tokens, and hopefully some other BT pieces. He's currently geared out in a set of full t4/badge gear. I'm primarily looking to optimize cat dps, bear tanking is secondary. The World of Warcraft Armory I'm planning on keeping the t4 hat and shoulders, gemmed for dps (cursed vision isn't an option), and he has enough gold to buy 3 pieces of BT/Hyjal loot, maybe 4. Rarer, desirable dps pieces like [Shadowmoon Destroyer's Drape] are also not for sale. What order should he be looking for pieces in? T6 gloves look like the clear best option, since they can be used in both dps and tanking roles. From there, I'm looking the following pieces. [Insidious Bands] T6 legs (enchanted for tanking) T6 chest (gemmed for tanking) For reference, he currently has no dps legs (using t4 legs), and uses [Chestguard of the Conniver] in the chest slot. The Teron bracers will let me regem [Band of the Swift Paw] for tanking purposes.
From the look of the gear he has he needs:

Chest, Legs, Gloves as first priority. T6 gloves and legs work well. Nether Shadow Tunic would be the best BT/Hyjal chest available. You could go for Teron bracers, but really the S3 ones are nearly the same for dps. Further once S4 comes out the new bracers are better than anything besides T6 for dps (and have only a minor arena rating requirement).