@Riot: 10% Crit chance is actually 8%

Comment below rating threshold, click here to show it.

Riot Morgageddon

Player Support Tech Specialist

01-31-2013

Quote:
Originally Posted by SplendidSorrow View Post
Have you considered the fact that what critical chance is displayed to you is balanced around your attack speed?

By that I mean if you have a 10% crit rate and a 1.2 attack speed, if you artificially slow your attack speed (as you did) you won't actually see that 10% crit rate because you're loosing data points artificially. If the application (LoL) manipulates individual attack critical percentage based upon expected attacks per second...and you artificially slow it down you may actually be causing your data to be off.
Interesting idea.
I suppose then to test that, I should run an *ideal* rate where my attack speed matches up or is slightly slower than what my program optimizes to, therefore i don't attack and press 'stop' after each registered hit. Rather, i just attack over and over, while my client records...

Poor mundo, he might not be able to take all those hits...


Comment below rating threshold, click here to show it.

Lux Eternus

Senior Member

01-31-2013

bumpidy


Comment below rating threshold, click here to show it.

yuki1314

Senior Member

01-31-2013

Quote:
Originally Posted by SplendidSorrow View Post
Have you considered the fact that what critical chance is displayed to you is balanced around your attack speed?

By that I mean if you have a 10% crit rate and a 1.2 attack speed, if you artificially slow your attack speed (as you did) you won't actually see that 10% crit rate because you're loosing data points artificially. If the application (LoL) manipulates individual attack critical percentage based upon expected attacks per second...and you artificially slow it down you may actually be causing your data to be off.
That does not explain the early 12% crit rate then falls back into 8% lately.


Comment below rating threshold, click here to show it.

Bland

Senior Member

01-31-2013

Quote:
Originally Posted by yuki1314 View Post
That does not explain the early 12% crit rate then falls back into 8% lately.
Interesting point... I know Riot uses an algorithm that "enforces" the statistical crit chance, by increasing your chance to crit on a non-crit, and decreasing it on a crit. Could it be that the error lies in that algorithm, and it tried to compensate for the 12% crit chance at the beginning too strongly?
Would be interesting to actually do this testrun several times, to see whether different spikes in the beginning lead to different results in regards to overall chance.


Comment below rating threshold, click here to show it.

TiBe

Senior Member

01-31-2013

Riot should be clarifying how it works, still don't get it why are they hiding this kind of information...


Comment below rating threshold, click here to show it.

SplendidSorrow

Senior Member

01-31-2013

Quote:
Originally Posted by yuki1314 View Post
That does not explain the early 12% crit rate then falls back into 8% lately.
Peaks and valleys are expected. However as more data points are missed from the slowing of attacks artificially the data is going to get more and more skewed one way or the other.

Consider that during the first the attacks that are missed are normals...and towards the end they shift more on the critical side?


Comment below rating threshold, click here to show it.

Belegdhor

This user has referred a friend to League of Legends, click for more information

Senior Member

01-31-2013

Is it possible riot uses a double Random Number Generator? Taking the average of two random numbers as opposed to just the single random number when determining criticals? This works in a way that anything less than 50% has slightly less chance to hit and anything more than 50% has slightly more chance to hit.

edit: this is a good explanation of what I mean - http://fireemblem.wikia.com/wiki/Ran...mber_Generator

"Note that in all games released outside Japan, the Random Number Generator has generated two numbers instead of one for hit chance, and the average of the number is the true hit value; in other words, a displayed hit (hit shown on screen) of 75 will actually hit on any two numbers whose average is less than or equal to 75. The effect on overall accuracy of attacks makes a "bell curve" effect: it becomes exceedingly rare for an accurate attack (90+ hit) to miss, while inaccurate attacks (Hit lesser than 50) will land far lest often than its hit rate would suggest. This is beneficial to the player, as their units usually have higher hit rates while the enemy units tend to have lower hit rates."


Comment below rating threshold, click here to show it.

Riot Morgageddon

Player Support Tech Specialist

01-31-2013

I don't know if slowing the attacks is skewing the data. I can time it different, as well as optimize the time, however, there still in lies another problem. On normal games, you right click things to have attacks disappear (didn't finish the animation before it died) or swapping targets so you cancel one attack to start another. That's inherent in both real games, and my testing. I dunno how the whole 'almost finished an attack' relates to the chance, whether the probabilities are calculated and then ignored, or stored for the next completed attack.


Comment below rating threshold, click here to show it.

Hello Darkness

Senior Member

01-31-2013

You could always do this on summoners rift, having mundo or w/e stand on the summoner platform. this way you can just flash into the base and auto attack him there the moment you have enough for the zeal. Thus, you won't have to waste 40 minutes building up tank mundo


Comment below rating threshold, click here to show it.

alcibiades5410

Senior Member

01-31-2013

I look forward to your research. Your grant of a case of mountain dew and a package of doritos has been approved.