Pondering: How much penalty for being human versus being a simulator?

[Apra]

So, I was pondering this today. The enhancement simulator is a machine. It isn't going to have reaction times as long as a human does, it doesn't have visual distractions or information overload. Or that moment of indecision when you try to figure out which item should be next in your rotation. I wonder how much of a DPS reduction being a human being is versus being a simulator.

Lets assume the fight is Patchwerk and there's no moving. You just stand a deliver. I'm guessing maybe an 8% reduction. It is really hard to say because there's a lot of variation in what the log parses seem to say.

[Fosorrow]

You also have to consider the RNG. Sometimes you can get lucky and others you get really unlucky. Enhance can be very proc'y with lightning shield, maelstrom and such, but you raise a good point. Obviously the amount would differ for each person against the machine. The only real way to find out would get a large group together and have them all run the sim for their char, then do the fight, get the results together and then compare. I would say the average is probably around or less than 5%, taking into consideration that a lot of damage is non controllable (white hits, windfury, shield), and the fractions of seconds wouldn't add up to too terribly much IMO.

[Endolicious]

Yeah, Enhancement is pretty much one of the worst specs to ask this question for. Maybe a Hunter or Rogue? Meh.

[Phosphorous]

AFAIK, a VERY GOOD human reaction time is 0.3 seconds.

It comes down to you being prepared as a player ; You will do more DPS if you make the decision before the moment, than at it. If you're making the decision at the second the event happens, you have 0.3 seconds delay. If you make it beforehand, you have 0.0 seconds, as you're beating the button before the event happens.

This is where things like Watcher, etc, come in handy.

"Human" models vs. Simcraft is based on how much time the human spends reacting to the GCD. Theoretically you could model this on simcraft if you measured your average decision-making delay and applied it to the spell cast rotation in the simcraft (thereby reducing the simcraft DPS).


Enhancement is a hard model because of maelstrom weapon, no matter what you will have a delay on that spell by being a human, unless you get the proc in the middle of a GCD. That's very hard to math.


TLDR : Decide what buttons to push before the GCD is ready and you will do more DPS.

[harky]

I'm not sure I understand the worry here. You will at most be losing two applications Maelstrom Weapon due to 'human timing'. Even then it's very unlikely. You'd need a reaction time of over a second, which even with lag is pretty substantial. Sims should be properly evaluating GCD usage, so there shouldn't be anywhere near an 8% DPS loss due to timing. If there is then you either have extremely bad timing, or don't understand your rotation. This isn't something like Hot Streak where you can't see it coming due to flight time and huge random chances.

If you ever have a "moment of indecision when you try to figure out which item should be next in your rotation" then you need to practice... a lot. Priority is LB > SS > ES > LL. It's one of the simplest rotations in game and even 'bad' rotations aren't much worse than 'perfect' rotations. I remember someone on EJ did a study with intentionally screwed up rotations compared to the best rotation and came up with a 50 DPS difference. Even Ret Paladins have a more complex rotation to monitor. If you're really having that much trouble install something that will alert you to MW procs so you know to use your next free GCD on LB. :

[Apra]

You don't always know what is next. For instance, you can get 2-3 MW5 stacks while you're thinking that ES or SS will come up first, and unfortunately your brain locked into your best guess before you registered you could have pressed the LB button. So you have to wait through the GCD before using MW5. I've also had the cast of ES and SS coming off CD at nearly the same time. My countdown counter shows they've both got 1 second left but one will be ready before the other, so another guess there which can cause a small delay. Although maybe I could reconfigure my cooldown display to show 10ths of seconds.

And of course, the more environmental things you have to pay attention to, the worse it gets as you can't lock your eyeballs onto your cooldown counters/SAA or whatever you use.

And then again, I don't have the reflexes of an 18 year old, hopped up on Mtn. Dew either. Never did, not having the right equipment for that.

I recently sat down at the sim and experimented with the rotation and even small changes made fairly sizable differences. For instance someone was proposing adding in MW3-MW4 and yup, it made about a 50dps difference.

But I was just curious about how much the human factor accounts for less DPS. It'd be really tough to quantify because the tools are so imperfect. I noticed that the log parsers tend to vary hugely.

[TheMediator]

AFAIK, a VERY GOOD human reaction time is 0.3 seconds.

It comes down to you being prepared as a player ; You will do more DPS if you make the decision before the moment, than at it. If you're making the decision at the second the event happens, you have 0.3 seconds delay. If you make it beforehand, you have 0.0 seconds, as you're beating the button before the event happens.

This is where things like Watcher, etc, come in handy.

"Human" models vs. Simcraft is based on how much time the human spends reacting to the GCD. Theoretically you could model this on simcraft if you measured your average decision-making delay and applied it to the spell cast rotation in the simcraft (thereby reducing the simcraft DPS).


Enhancement is a hard model because of maelstrom weapon, no matter what you will have a delay on that spell by being a human, unless you get the proc in the middle of a GCD. That's very hard to math.


TLDR : Decide what buttons to push before the GCD is ready and you will do more DPS.

This post wraps everything up nicely.

[Deleted]

Brain meltdown.

[U1000]

the reaction time from your brain to your finger is 0.26 seconds exactly

[Apra]

the reaction time from your brain to your finger is 0.26 seconds exactly

That's pretty slow compared to a computer. I've read where we actually make the decision before we're conscious of it.

I notice a lot of times my finger is traveling to a button but I can't seem to interrupt myself and recommit to a different button. I suppose that could be happening during that quarter of a second that the impulse is going from brain to finger, the new impulse can't arrive quick enough to cancel it.

[sanderke]

the reaction time from your brain to your finger is 0.26 seconds exactly

It depends on the person.

[Apra]

I'm not sure I understand the worry here.

No worries at all. Just pondering things.

[Deleted]

So, I was pondering this today. The enhancement simulator is a machine. It isn't going to have reaction times as long as a human does, it doesn't have visual distractions or information overload. Or that moment of indecision when you try to figure out which item should be next in your rotation. I wonder how much of a DPS reduction being a human being is versus being a simulator.

Lets assume the fight is Patchwerk and there's no moving. You just stand a deliver. I'm guessing maybe an 8% reduction. It is really hard to say because there's a lot of variation in what the log parses seem to say.

You do realise that you can configure the sim to answer precisely this question?? You simply modify the amounts min & max lag. Lag isn't just client to server lag its meant to be human reaction time lag as well. That's why anyone who posts an EnhSim config file with 0 lag should get told to come back when they have a sensible config file.

Normally your lag figures in the config file should be 200-400.

From memory by default there is a 0.25 sec human reaction lag built into the sim. However this assumes really REALLY good reaction times.

[Apra]

I was thinking that maybe lag could simulate human reaction times. So add 250 lag (if you're an 18 year old boy with lots of Mtn. Dew in your system) to whatever your network latency is and you might get a better approximation. I gotta figure out what the lag is for being a 50 year old woman, hopped up on Trader Joe's coffee. No smart-ass comments now!

Actually, it wouldn't be that hard to figure out my own lag... I could write a little program to measure it. I could even try to approximate the information overload factor.

Lastly, if you can calculate this properly that's step one. The next step is to see how far off you are from the projections from the simulator and to do that you'd need a very accurate DPS parser. I find they vary extremely. Skada was about 20% lower than WMO. Does anyone have a DPS parser they think is actually accurate? Real time or offline? Probably prefer offline.

Whoa! Increasing my lag to 350-400 reduced my DPS by about 300.

[Apra]

Ah ha! I feel slightly gloaty right now.
http://en.wikipedia.org/wiki/Reaction_time

Reaction time (RT) is the elapsed time between the presentation of a sensory stimulus and the subsequent behavioral response. RT is often used in experimental psychology to measure the duration of mental operations, an area of research known as mental chronometry. The behavioral response is typically a button press but can also be an eye movement, a vocal response, or some other observable behavior.

RT is fastest when there is only one possible response (simple reaction time) and becomes slower as additional response options are added (choice reaction time). According to Hick's law, choice reaction time increases in proportion to the logarithm of the number of response alternatives. The law is usually expressed by the formula RT = a + blog2(n + 1), where a and b are constants representing the intercept and slope of the function, and n is the number of alternatives.[1]

Reaction time is quickest for young adults and gradually slows down with age. It can be improved with practice, up to a point, and it declines under conditions of fatigue and distractions.[2]

http://en.wikipedia.org/wiki/Reaction_time

I wonder if Blizzard considers this when designing classes? The simpler your rotation, with the fewer random things happening, the better your reaction time should be.

Oh wow, it gets more interesting...

Simple reaction time is the time required for an observer to respond to the presence of a stimulus. For example, a subject might be asked to press a button as soon as a light or sound appears. Mean RT for young adults is approximately 215 milliseconds to a detect visual stimulus, and approximately 160 milliseconds to detect an auditory stimulus.[5]

I knew there was a good reason why I use all those audible warnings in SAA and MSBT!

Due to momentary attentional lapses, there is a considerable amount of Ramell variability in an individual's reaction time. To control for this, researchers typically require a subject to perform multiple trials, which are then averaged to provide a more reliable measure.

Yup, yup. All those encounters where you have to pay attention to things in the environment definitely take a toll on the DPS.

Oh, here's something that measures reaction times with choices involved: http://cognitivelabs.com/test5.htm

[Shaamwow]

The difference in reaction time between a person and a computer is quite considerable.

Lag is allowed for in the simulator but is not precisely the same thing (nor is it reasonable to expect a simulator to correct for human error like this).


A cool demonstration of this--for anyone who thinks they have cat like reflexes this will put you in your place:

1. Find a dollar bill and a friend.

2. Steady your arm on a table or something while holding it out in front of you.

3. Hold your thumb and forefinger about 1 inch apart (with the same orientation they'd be in if you were holding a cup or something).

4. Have your friend dangle the dollar in between your fingers longways, with the center of the bill being even with your fingers.

5. Have your friend drop the dollar at a random time.

6. Try (and fail) to catch it when your friend drops it.


Now, its important not to jump the gun--this is not a test of how well you can anticipate something, but rather a show of just how slow our response time is. Only try to catch the dollar when it is dropped.

Your brain just can't send the messages fast enough--unless you have jumped the gun or anticipated the bill being dropped you will never catch it.

This goes back to what was said earlier, about DPS being faster if you decide your actions ahead of time. This is inevitably true given the short span of time between GCD's. You should always be thinking of your next move as you are executing your current one.


As far as simming goes for enhancement--we're probably the class which can take the most meaningful advantage of a sim, being as a huge portion of our dps comes from proc based effects.

[TwistedPower]

For allot of classes it doesnt realy matter about the initail timing of a move. This is becuase most specs have a rotation (and each spell has a universal cooldown) so as long as a player clicks his rotation in an order and clicks his moves one after another then the speed at which the spells are cast are constant.

Thus the universal cooldowns ensures that even if the first cast is .5 seconds off, this will not happen again as you have an entire second to press the next button in the line of buttons to press.

In a fight where you stand still you would only ever lose that inital .5 seconds.

[mapp]

shaamwow i'm gonna try that sounds like a fun waste of time =D (no sarcasm)

[Shaamwow]

For allot of classes it doesnt realy matter about the initail timing of a move. This is becuase most specs have a rotation (and each spell has a universal cooldown) so as long as a player clicks his rotation in an order and clicks his moves one after another then the speed at which the spells are cast are constant.

Thus the universal cooldowns ensures that even if the first cast is .5 seconds off, this will not happen again as you have an entire second to press the next button in the line of buttons to press.

In a fight where you stand still you would only ever lose that inital .5 seconds.

Just say no to clicking.

Please.

[underdogba]

The point of a simulator is not necessarily to tell you how much dps you should be doing, the point is to accurately evaluate differences between gear choices based on actual real-world stats and mechanics rather than thinly-justified generalizations.