EPTAnyways, riot changed the crits a while back when people complained about streaks of luck/unluck(sp?).
[Patch 3.07: Nerf Everything Not Nami] General Discussion…
| Forum Index > LoL General |
Looks like we're back to status quo. Hope more of you lurkers unburrow and talk with us. :3 If you have any issues or comments about the new design, feel free to PM Neo. | ||
|
LaNague
Germany9118 Posts
Anyways, riot changed the crits a while back when people complained about streaks of luck/unluck(sp?). | ||
|
Diamond
United States10796 Posts
Btw anyone know what runes/masteries Alex Ich ran on jungle eve @ allstars? Want to re-add that to my ranked pool and curious what he went with. Checked a VOD and did not see it mentioned, but the lolchampseries youtube is missing some of the games from that weekend. Edit: I mean troll as non-serious, not literally troll the game. | ||
|
nafta
Bulgaria18893 Posts
edit:Diamond runs ap quints/ad red/ar yellow/5 flat mr and 4 scaling and 21/9 masteries with mpen. | ||
MoonBear
Straight outta Johto18973 Posts
On June 06 2013 07:00 LaNague wrote: well, im pretty sure the mathematicians can differentiate between a really random distribution and a "not so quite random, but still a bit random" distribution. Anyways, riot changed the crits a while back when people complained about streaks of luck/unluck(sp?). Most forms of randomisation are really just pseudorandomisation. Even computers cannot truly generate random numbers and end up being pseudorandom. There are very few things which can truly generate random numbers, and often you start getting into impractical for a game territory, like hardware randomisation. Fwiw iirc LoL uses a Markov chain to generate its RNG. | ||
|
thenexusp
United States3721 Posts
On June 06 2013 06:45 bmn wrote: I'm confused what you mean by "truly random". A coin flip that has a 99% chance of yielding head and 1% chance of yielding tail is still "truly random". Uniform distributions maximize entropy, but that doesn't make non-uniform distributions "not random". If you have a 1% chance of critical strike on every hit that's just as much a "truly random" event as when you have 50% or 99% critical strike chance. 0% and 100% are degenerate cases, sure. If we use your algorithm and assume that a 6 means 'critical strike' I can now easily exploit the behavior to achieve far more critical strikes (as a fraction of attacks) than I previously could: always 'waste' the two hits after a critical strike on a minion, and all other attacks have a 1/4 chance of being a critical strike. It'll just encourage people to time attacks the way they do with caitlyn to maximize harass; some champs are a lot more amenable to that as others. (It's easier if you're ranged and can disengage at will, it's harder if you have to go for a hard engage and exchange attacks in bursts.) But as Seuss pointed out that's a property of your algorithm, and the sequence itself is far too short to yield useful analysis like that. yeah, "truly random" isn't the right word, what I mean was "truly uniformly random" Most methods of smoothing out the randomness do have their abuse cases, like you mentioned. I can't speak to how Riot handles crit, and it's definitely not as extreme as "you can't crit twice in a row", so figuring out how to abuse it is probably more effort than it's worth. (I could also see an easy implementation where you run separate markov chains for autoattacking champions and minions, so you wouldn't even be able to buffer your crits that way) EDIT: Also my original post was trying to touch on that humans tend to perceive uniformly random sequences to be less random than sequences generated from less than uniformly random processes (excluding long runs of the same number is a big part of this) | ||
|
Dark_Chill
Canada3353 Posts
Plz educate. | ||
|
thenexusp
United States3721 Posts
On June 06 2013 07:09 Dark_Chill wrote: I'm not sure I understand the difference between truly random and truly uniformly random. I understand Riot uses a different system to make sure that when a crit doesn't happen on one attack, it's more likely to happen on the next attack, but I don't see why that's related to uniform randomness. Plz educate. Uniformly random means, if you have a 30% crit chance, every single attack has a 30% chance to crit regardless of the attacks that came before. That's not the system that LoL has - if you hit two crits in a row, the next attack has a lower than 30% chance to crit. The numbers are still set up so that on average, 30% of your attacks will crit. EDIT: "uniform" isn't really the right word to use since your crit chance isn't 50%, but just go with it. Pretend what's really being generated is a number from 1-100, and you crit if the number generated is less than your crit chance. On June 06 2013 07:05 MoonBear wrote: Most forms of randomisation are really just pseudorandomisation. Even computers cannot truly generate random numbers and end up being pseudorandom. There are very few things which can truly generate random numbers, and often you start getting into impractical for a game territory, like hardware randomisation. Fwiw iirc LoL uses a Markov chain to generate its RNG. Any software RNG is going to be pseudorandom and deterministic based on its original seed, but the sequences generated by the good ones are essentially undifferentiatable from a sequence generated by a real random process (with the randomness tests we currently have). | ||
|
Diamond
United States10796 Posts
| ||
|
thenexusp
United States3721 Posts
On June 06 2013 07:16 Diamond wrote: Speaking of crits, I always here the LCS caster babble how your crit chance increases every time you don't crit until you do. Which seems stupid as hell. Am I misunderstanding this? If it's a real thing what's the factor it increases by? I have looked pretty hard and have yet to find the info. Conversely, if you hit a crit, your crit chance is lowered until you miss a couple. One way to visualize this is to imagine a deck of 100 cards labeled from 1-100. When you autoattack, you draw a card, if the number is less than your crit chance, you get a crit. However, you don't replace the card and you only reshuffle when you go through the deck. Missing a crit increases your crit chance because there's now comparatively more "crit cards" in the deck, and vice versa. On the whole you're still going to get a proportion of crits equal to your crit chance. | ||
|
RoieTRS
United States2569 Posts
| ||
|
Sufficiency
Canada23833 Posts
On June 06 2013 07:16 Diamond wrote: Speaking of crits, I always here the LCS caster babble how your crit chance increases every time you don't crit until you do. Which seems stupid as hell. Am I misunderstanding this? If it's a real thing what's the factor it increases by? I have looked pretty hard and have yet to find the info. This is most likely true. It is known that the chance of consecutive crits is lower than what champion stats describe, so the chance of consecutive non-crits must also work that way. | ||
|
bmn
886 Posts
On June 06 2013 07:07 thenexusp wrote: yeah, "truly random" isn't the right word, what I mean was "truly uniformly random" Most methods of smoothing out the randomness do have their abuse cases, like you mentioned. I can't speak to how Riot handles crit, and it's definitely not as extreme as "you can't crit twice in a row", so figuring out how to abuse it is probably more effort than it's worth. (I could also see an easy implementation where you run separate markov chains for autoattacking champions and minions, so you wouldn't even be able to buffer your crits that way) EDIT: Also my original post was trying to touch on that humans tend to perceive uniformly random sequences to be less random than sequences generated from less than uniformly random processes (excluding long runs of the same number is a big part of this) I see, what you're talking about is better described as crits being independent (i.e. each attack is considered independently of whether the previous one was a critical strike). A loaded coin is not a good analogy there, since just loading will still leave the individual outcomes independent. I was thrown off by that example. Your algorithm still produces a sequence with uniform distribution of die outcomes, by the way, it's just that those individual outcomes aren't _independent_. I always figured that the whole point of critical strikes was to add luck to the game, so I don't really care much about the exact mechanics as long as it's not trivially predictable. (It'll inevitably be unfair at some points -- the trivial example being two people who are each 1 critical strike from death duking it out) | ||
|
Jojo131
Brazil1631 Posts
| ||
|
JeosAdn
Costa Rica432 Posts
On June 06 2013 07:29 Jojo131 wrote: This discussion did nauti end well ._. FTFY | ||
|
Tula
Austria1544 Posts
| ||
onlywonderboy
United States23745 Posts
| ||
|
SimulatedAnneal
765 Posts
On June 06 2013 06:45 bmn wrote: I'm confused what you mean by "truly random". A coin flip that has a 99% chance of yielding head and 1% chance of yielding tail is still "truly random". Uniform distributions maximize entropy, but that doesn't make non-uniform distributions "not random". If you have a 1% chance of critical strike on every hit that's just as much a "truly random" event as when you have 50% or 99% critical strike chance. 0% and 100% are degenerate cases, sure. If we use your algorithm and assume that a 6 means 'critical strike' I can now easily exploit the behavior to achieve far more critical strikes (as a fraction of attacks) than I previously could: always 'waste' the two hits after a critical strike on a minion, and all other attacks have a 1/4 chance of being a critical strike. It'll just encourage people to time attacks the way they do with caitlyn to maximize harass; some champs are a lot more amenable to that as others. (It's easier if you're ranged and can disengage at will, it's harder if you have to go for a hard engage and exchange attacks in bursts.) But as Seuss pointed out that's a property of your algorithm, and the sequence itself is far too short to yield useful analysis like that. They track (non) crits separately against players and minions. | ||
|
Craton
United States17294 Posts
| ||
|
Badboyrune
Sweden2247 Posts
| ||
|
wei2coolman
United States60033 Posts
On June 06 2013 08:00 Craton wrote: I understand it to not be a true Markov chain, just something related to it / an offshoot that incorporated some of the streaking reductions. It's probably just a large markov chain that's constantly modifying. | ||
| ||
