Ability to react to overheads

[Hollysmoke]

Reaction Times:

Average Reaction Time: 265ms

Median Reaction Time: 254ms

Source: http://www.humanbenchmark.com/tests/reactiontime/statistics

Overheads (Dusts)

Axl - 433ms

Bedman - 433ms

Chipp - 450ms

Elphelt - 400ms

Faust - 433ms

I-No - 433ms

Ky - 466ms

Leo - 416ms

May - 450ms

Millia - 466ms

Potemkin - 450ms

Ramlethal - 466ms

Sin - 483ms

Slayer - 400ms

Sol - 400ms

Venom - 416ms

Zato-1 - 416ms

Average dust animation time: 434.21ms. Converting the time to FPS based on the overhead average and human reaction time average, the average player would have about 10.14 frames to reaction to a dust.

Some interesting things I noticed were the more mix-up/setplay oriented characters or speedier characters in fact had a slower overhead. More straight-forward characters were faster. Sin's dust should be slightly

faster to keep in tune with the rest of the cast. The way animations are played out can cut down on the number of frames you'll physically be able to react too (for example, Millia rolls and then the dust glow appears). Feel free to discuss or point out anything incorrect. I just thought this was interesting.

[TheRealBobMan]

This is a very complex topic.  I've written up some stuff on this in the past, but never came up with a revised draft that I was happy to publish.  Either my "newbie" sample reader couldn't understand, or my target audience sample readers found it boring.  Maybe I'll just info dump in this thread.  I'll start with a few major points:

 

1.  There isn't a single "reaction time" value that we can use as a benchmark for everything.  The human benchmark website you've linked only tests for a Simple Reaction (and at this point we'll assume the reactions I'm talking about are visual reactions).  A Simple Reaction is when you execute response Y to stimulus X.  See light turn green, press button.  The values I find tend to vary, but hold around 200ms.  Keep in mind that the nature of the test can skew the results, such such as a computer based test like this one being run on monitors that have substantial display latency.

 

2.  There are also Complex Reactions, also known as Recognition Reactions.  If A, do B.  If X, do Y.  You have to differentiate between two different visual stimuli and react with the appropriate response.  Light turn green, left click; light turn red, right click.  I'm generally finding 380ms here.

 

3.  Adding more potential stimuli to account for will increase the amount of time it takes to appropriately respond.

 

4.  You actually react faster than you can commit to acting.  As far as visual stimuli goes with fighting games the difference is really minor (21ms).

 

5.  Distraction makes a substantial difference in reaction time.  If you receive a second stimulus within 100-500ms after the first, you will react slower to the new one.  The number I'm getting is 174ms.  You also generally don't notice that you're reacting more slowly when this happens

 

6.  Auditory reaction times are faster than visual reaction times.  The average for auditory is 140-160ms.

 

 

Here are some sources:

http://biology.clemson.edu/bpc/bp/Lab/110/reaction.htm

http://www.livescience.com/6041-reactions-faster-actions-study-finds.html

http://www.cbc.ca/news/technology/reaction-faster-than-action-study-finds-1.965676

http://discovermagazine.com/2010/nov/15-the-brain-router-in-our-heads-processing-bottleneck

 

 

Now there are a few other things to consider:

 

A)  When you react to something, where is the telegraph that you're reacting to?  The sooner you can recognize something, the sooner you can react to it.

If you react to Millia rolling away from you, you're reacting much sooner than if you wait for the red flash.  This is easier said than done, but if she's doing her forward roll there's a couple of things she could be doing.  If she rolls backwards the only thing that's going to happen is the overhead, or a YRC which stops time long enough for you to react to it.

 

B)  How can you reduce the number of things you need to react to?  If I crouch block by default I don't have to worry about the 5-frame lows, and only have to switch blocking on reaction to overheads or jump/poke to deal with throws.

GG is designed well, so even though the overheads are slow enough to react to if they're the only thing you have to watch out for, throws are instant.  Jumping to avoid a throw makes you vulnerable to lows (you're not holding downback if you're holding up), and poking out makes you vulnerable to frame traps.

 

C)  Characters with strong mixup tend to have slow dusts, but they have other mixup tools that are more difficult to react to.  To compensate, the slower Dusts tend to have extra utility.  Millia rolls backward and low profiles under things.  I-No has decent range and a lot of active frames.  Zato floats over lows.  They also tend to have strong mixup specifically because they have many high/low/throw tools and many different avenues to utilize those tools from.  One of Leo's biggest weaknesses in back-turn stance is that you can just block low and reaction block the fast overhead.  He's not going to throw you, and he doesn't have any other high/low/throw tools besides his really slow stomp.  His crossup options really help, but overall his high/low is actually not scary.

 

 

Going to take a break here to let other people post and start a discussion.

[Hollysmoke]

-Snipples-

I forgot to mention that latency on certain input devices are to be considered as well. These are very rough calculations but the they are on they basis of "You see glow, you press button". How the animation plays out was something I also pointed out, and causes people to react to it differently. Although 10.14 frames to react would be "optimal" it is by and far from realistic. I'd see it more half that to react. If any bio majors want an excuse for both playing guilty gear and writing a thesis, now's your chance. If you also subtract your latency on netplay from these times, you'll get a rough idea as to why overheads are a pet peeve for most people.

[Anne]

Just not at the moment.

[mAc Chaos]

This was locked during the height of the "can you react to 18f/14f overhead" stuff that was blowing up Twitter and DL at the time. That's old news now so go ahead and discuss.

[skd]

Average dust animation time: 434.21ms. Converting the time to FPS based on the overhead average and human reaction time average, the average player would have about 10.14 frames to reaction to a dust.

not quite, 434.21ms ends up at a bit more than 26f (1 frame is 16.666...ms).

[Kikirin]

not quite, 434.21ms ends up at a bit more than 26f (1 frame is 16.666...ms).

Probably meant a ~10 frame leeway between the average/median reaction time (265ms/254ms) and the average Dust startup? ( 434.21 - 265 ) / 16.666 = ~10 frames.

In either case, I'm super free to dust, both online and off, in spite of their "slowness". =(

[Hollysmoke]

I had a calculation written out but I've long lost it since then. I don't care enough to bother writing it out again, but basically based on human reaction times, we take away that from time of the start-up because our brain is not going to be able to process that. It's about a 10F window that you begin comprehension and react accordingly. There are also a lot more factors involved that make the calculation more complex but my drive for discussion was snuffed out.

[Coffeeling]

10F is only for raw twitch reaction with no processing involved - a muscle spasm in reaction to anything at all happening. Reaction with processing is much, much slower than that, though you can speed it up by narrowing down what you're looking for (say, the opposing character's feet leaving the ground).

[TheRealBobMan]

Oh shit, we can talk about this again! 

@Coffeeling:  It's closer to 12f for what you're thinking of.  Everyone seems to be talking about different sets of numbers.

 

Simple Reactions - the "twitch" reaction.  This is 200ms / 12f

Complex Reactions - If A, then B, if C, then D.  This is 384ms / 23f

 

Holly's 10f comes from taking an average dust and subtracting the time it takes for a Simple Reaction.  This basically how wide the margin for error is.  Blocking a 5D on reaction is not at all hard if it's the only thing you have to worry about.

Blocking TK Badmoon is hard even when it's the only thing you have to worry about, just because it's so fast.  But it's doable with training.  It's also doable to watch for this and several other things at the same time, though you do open yourself up to other things.  Haircar has a little hop, so it's really easy to mess up if you block low and switch on reaction to the jump.  If you watched for both, you couldn't block either, ever.  If you watch for the hop, you turn it into a 50/50, or worse if the other guy picks up on your weakness and starts spamming Haircar to trip you up, switching just to put the fear of the TK Badmoon in you and keep you trying to switch.  If you watch Millia's feet you wont fault on Haircar, but can catch TK Badmoon.  You're also going to catch 2K, 2S, and 2D by virtue of blocking low until you see the cue.  The only problem is that you're vulnerable to 6K, throw setups, and crossups, and probably wont be able to punish a 5D as well as you should (though you may be able to block it in time because it's that slow).

At some point you have to take a guess or make a read based on your opponent's patterns.  It's just that if you learn your opponent's gatling tree and where their mixups may come from, you significantly reduce the amount of things you have to watch for.  If Millia isn't covered by a disc or Roses and has no meter, the TK Badmoon / Haircar mixup is much riskier and gives much less reward.  In such a situation she's more likely to rely on 6K, throw, and traps that bait you into trying to punish her (using 5D to dodge a reversal or feint and punish a reversal BS).

[Kikuichimonji]

I would like to mention that most people who think they're reacting to Leo's overhead in my experience are just fuzzy blocking and get hit if I do bt.PK. I agree that it's entirely possible to block on reaction, though.

It's definitely more possible to react to things you're expecting. But looking for specific offense to counter often makes you extremely predictable on defense. The aggressor can either introduce false positives (eg Haircar instead of Bad Moon), or he can simply avoid your anticipated reaction (dash up throw instead of Bad Moon).

Also, I think it's incredibly important to talk about what I think of as "time horizons." Essentially, when you know something will happen during a very specific time period, you can ready yourself to react in a much stronger fashion than over the long term. To compare this, think of Millia's 6K. There are players who can get pretty good at blocking Millia 6K on wake-up. However, if Millia can extend her pressure, those same players will have much worse odds of blocking 6K.

This is why I think the Millia Blocker flash is not a strong test case for blocking Millia mixups; the variability in timing is too great, and in a realistic game, nobody is forced to respect the threat of 6K for five seconds straight. People use fuzzy jab, fuzzy jump, and FD to encourage aggressors to run "clean" offense. This of course doesn't even consider fuzzy block options which can make 50/50's much harder to execute.

Sent from my LG-H811 using Tapatalk