EPT[Patch 5.19] League of Legends General Discussion - Page 6
| Forum Index > LoL General |
|
Goumindong
United States3529 Posts
| ||
|
wei2coolman
United States60033 Posts
On October 01 2015 10:01 Amarok wrote: For the people defending Crit, do you actually think having Crit (pseudo random or otherwise) is preferable to just having "Increases auto-attack damage by X%" where X is the Crit chance of the item? Would the game be better or worse if Crit was changed in that way? Assuming you changed procs that work off Crit to scale correctly (IE, Frenzy mastery, crit damage runes) I don't see how it could be anything but an improvement. So why defend Crit? I mentioned it earlier, it creates a "threat" in the midgame. For example if you plan on trying to dive botlane 3v2, and the enemy adc has the threat of "crit" from an IE, it makes the dive significantly more risky, than if he was just to have "auto attack damage by x%" where X is equal to the damage difference in DPS via IE crit, because of the RNG of crit chance having a variable to do more damage than expected. Thus you need to play around the idea of assumed "worst case" scenario. | ||
|
Caiada
United States3052 Posts
On October 01 2015 10:16 wei2coolman wrote: I mentioned it earlier, it creates a "threat" in the midgame. For example if you plan on trying to dive botlane 3v2, and the enemy adc has the threat of "crit" from an IE, it makes the dive significantly more risky, than if he was just to have "auto attack damage by x%" where X is equal to the damage difference in DPS via IE crit, because of the RNG of crit chance having a variable to do more damage than expected. Thus you need to play around the idea of assumed "worst case" scenario. Also all very true. My completely unformed/iterated proposition was something along the lines of guaranteed crit as an item stat. For example you'd pick up IE and you wouldn't get '20% chance to do blah', you'd get 'your 5th AA within X seconds does blah'. The issues I immediately saw were that it would stack strangely (does everything have 1 in 5 and you get an additional 1 in 5? And so on. What about Tforce's crit? Is 1 in 10 the base, when 1 in 10 wouldn't actually even really be useful? How useful is 10% crit right now?) and that it creates a strange counterintuitive trading minigame for some ADCs (it would only be on full ADC items and things like Shiv, but how would it interact with Caitlyn itemization, Vayne itemization? and so forth.) I also don't think AD mains would like it, lol. I've already basically concluded that they're changing the entirety of core AD itemization from what they've said (or god I hope they are because it's a fucking wasteland of design choices right now), it's just a question of how far they're going and what idea they have. | ||
|
Gahlo
United States35174 Posts
On October 01 2015 10:16 wei2coolman wrote: I mentioned it earlier, it creates a "threat" in the midgame. For example if you plan on trying to dive botlane 3v2, and the enemy adc has the threat of "crit" from an IE, it makes the dive significantly more risky, than if he was just to have "auto attack damage by x%" where X is equal to the damage difference in DPS via IE crit, because of the RNG of crit chance having a variable to do more damage than expected. Thus you need to play around the idea of assumed "worst case" scenario. Honestly, that's just a result of the shit state of marksmen with "I need 2 items to be good against low durability targets/tower and 3 items to be good against tanks." | ||
|
Sufficiency
Canada23833 Posts
On October 01 2015 10:13 Goumindong wrote: Sufficiency you do realize that if crit is memory less it has all the problems indescribed necessarily right? I find it funny your suggesting that without realizing what it means I don't think you know what it means =\ We get it. You are an economist. But googling the topic for 2 seconds then splitting out jargons doesn't mean anything nor does it demonstrate that you know anything. I think we have gone through this too many times already. | ||
|
killerdog
Denmark6522 Posts
The only skill there is working out when a gamble is worth it and when it isn't, which is what you explicitly told me to not do like 2 posts ago... | ||
|
wei2coolman
United States60033 Posts
On October 01 2015 10:19 Gahlo wrote: Honestly, that's just a result of the shit state of marksmen with "I need 2 items to be good against low durability targets/tower and 3 items to be good against tanks." That is partialy it, it's a result of a powercurve created by multiplicative scaling. I think champions like the new one (Kindred right?), is trying to do away with that. Adding a lot more innate power that doesn't require crit scaling. On October 01 2015 10:26 killerdog wrote: The only skill there is working out when a gamble is worth it and when it isn't, which is what you explicitly told me to not do like 2 posts ago... What I said is, there are solutions to dealing with a gamble. If you have higher crit chance than the enemy, then you should play out the 1v1 gamble. If you don't have higher crit chance ,than you should avoid the gamble. Knowing when to play the hand you're dealt is a poker skill as well. | ||
|
Goumindong
United States3529 Posts
On October 01 2015 10:20 Sufficiency wrote: I don't think you know what it means =\ We get it. You are an economist. But googling the topic for 2 seconds then splitting out jargons doesn't mean anything nor does it demonstrate that you know anything. I think we have gone through this too many times already. I am quite familiar with markov chains thank you very much. Do you know what memoryless means? | ||
|
zer0das
United States8519 Posts
| ||
|
Nos-
Canada12016 Posts
| ||
|
TheHumanSensation
Canada1210 Posts
On October 01 2015 12:45 Nos- wrote: sufficiency and goumindong having a go at each other is a fucking riot jesus christ please continue discussing this They should settle it with a deathmatch - first to crit wins. | ||
|
Goumindong
United States3529 Posts
On October 01 2015 12:52 TheHumanSensation wrote: They should settle it with a deathmatch - first to crit wins. Oh god. Anywho a markov chain is memoryless which means that the outcome depends only on the immediate state. Which is to say that the state is "crit" or "not crit" of only the prior attack and each of crit/not crit has a different probability of critting. Which is to say that suppose the two ADs start with crit/not crit and not crit/not crit. At this point the two ADs are in the same state and so their future expectations are precisely the same. So unless no/no and crit/no. (Or crit/crit no/crit) cannot happen* then a markov chain produces the same problems expressed earlier. Because crit/no + E(future attacks). > no/no+ E(future attacks) Because markov chains are memoryless and that is what memoryless means. *and even if it can't so long as the initial state is effectively random you still have the problems as shown earlier in the thread. Since the fully deterministic 50% crit markov chain is simply 0,1 for row 0 and 1,0 for row 1 and we already know the outcome for that. Edit2: Since this is so much fun lets define the problem very specifically. There is a function for the RNG. It follows some form, geometric, whatever, doesn't matter. What matters is that it generates a count of crits over a number of attacks/time Such P(Count[crit%, t]=Count[crit%,t]) =/= 1 for any t. That is, given any time period that a fight might exist in, it is not the case that two players with the same critical strike rate will always have the same amount of crits. If this is the case then we can talk about the advantage of a fight as a binary system. It produces an advantage for one side when Count[crit%, t]=/=Count[crit%,t]* and records a 1, and doesn't when Count[crit%, t]=Count[crit%,t] and produces a zero. Now we have independent binary distribution for each fight, and we can start talking about how likely it is that crit will effect a fight. And since we know that the probability of not hitting a 1 at least once is zero as you continue to run random numbers on a binary distribution we know that no matter the RNG function they use it must be the case that, so long as everyone doesn't always crit the same over any time period that a fight might possibly exist in that the randomness in crits is effecting the game *we might instead talk about the relative advantage in crit count relative to expected for two different players. But this is more or less the same thing but easier to write | ||
|
Alaric
France45622 Posts
On October 01 2015 10:19 Gahlo wrote: Honestly, that's just a result of the shit state of marksmen with "I need 2 items to be good against low durability targets/tower and 3 items to be good against tanks." Save for those who start with Triforce or BotRK, figures. | ||
|
Osmoses
Sweden5302 Posts
On October 01 2015 14:32 Goumindong wrote: Oh god. Anywho a markov chain is memoryless which means that the outcome depends only on the immediate state. Which is to say that the state is "crit" or "not crit" of only the prior attack and each of crit/not crit has a different probability of critting. Which is to say that suppose the two ADs start with crit/not crit and not crit/not crit. At this point the two ADs are in the same state and so their future expectations are precisely the same. So unless no/no and crit/no. (Or crit/crit no/crit) cannot happen* then a markov chain produces the same problems expressed earlier. Because crit/no + E(future attacks). > no/no+ E(future attacks) Because markov chains are memoryless and that is what memoryless means. *and even if it can't so long as the initial state is effectively random you still have the problems as shown earlier in the thread. Since the fully deterministic 50% crit markov chain is simply 0,1 for row 0 and 1,0 for row 1 and we already know the outcome for that. And how is this relevant when it's already been established that riots crit code is not, in fact, memoryless? I like crit, some randomness is good for a game, any and all anectodes mentioned about how crit change the outcome of a game mean nothing because if all it took was a crit then the game was clearly close enough to be fair either way. Who the hell wants to play a game that is 100% deterministic anyway? | ||
|
M2
Bulgaria4186 Posts
| ||
|
killerdog
Denmark6522 Posts
On October 01 2015 17:05 Osmoses wrote: And how is this relevant when it's already been established that riots crit code is not, in fact, memoryless? I like crit, some randomness is good for a game, any and all anectodes mentioned about how crit change the outcome of a game mean nothing because if all it took was a crit then the game was clearly close enough to be fair either way. Who the hell wants to play a game that is 100% deterministic anyway? What kind of logic is it that it's ok for close games to be decided by chance? In a game where a vast majority of games are stomps pretty much decided from the start, that has to be the most frustrating possible. Why should the few games which actually have the chance to be carried by one or two amazing displays of skill be reduced to being rng'd? That has to be the least player satisfying approach to game design i've heard so far :o If you get a game where it's neck and neck for 30-40 minutes, do you really thing the best way to end that is with a coinflip, and not with someone pulling off a crazy outplay? Having a super close tense game get decided based off a lucky crit/series of crits is absolutely terrible for the loser, and really not fun/gratifying for the winner either... | ||
|
Osmoses
Sweden5302 Posts
Every decision you make in a game that has RNG is a calculated risk, from the behaviour of the enemy team to diving a weak adc that has 5% crit. Sometimes you lose even when the odds were in your favor, but if that wasn't the case the game would be boring as shit, wouldn't it? | ||
|
killerdog
Denmark6522 Posts
On October 01 2015 17:31 Osmoses wrote: What are these amazing displays of skill and crazy outplays you're talking about? The WHOLE reason why close fights are exciting is because you can't tell how they're gonna end and the ONLY reason you can't tell is randomness. Crazy outplays happen because someone juked left when the opponent thought they would juke right, there is literally no difference between that and crit. When the juke works everybody yells "DAT META" but nobody says a word when it fails. Every decision you make in a game that has RNG is a calculated risk, from the behaviour of the enemy team to diving a weak adc that has 5% crit. Sometimes you lose even when the odds were in your favor, but if that wasn't the case the game would be boring as shit, wouldn't it? Did you actually just say that there is "literally no difference between an outplay and a crit" | ||
|
Fildun
Netherlands4124 Posts
| ||
|
nafta
Bulgaria18893 Posts
| ||
| ||
