EPT[Patch 4.20] Blaze It ◯０o。(ー。ー)y~~ - Pag…
| Forum Index > LoL General |
|
Kupon3ss
時の回廊10066 Posts
| ||
|
IamPryda
United States1186 Posts
| ||
|
chalice
United States1945 Posts
On December 08 2014 09:21 Kupon3ss wrote: There are some stats that can be derived from controlled situations like LCS but nobody cares about the lol equivalent of your stats from a pickup game on the street that's an attitude problem, not an argument against statistics and analytics. LCS teams should be looking at solo queue as the NCAA to their NBA, not a worthless pick up game. | ||
|
ShaLLoW[baY]
Canada12499 Posts
On December 08 2014 03:20 chalice wrote: it's not that the other smites are bad (except for lolpoacher's), it's that even if you don't need smite to clear effectively, not smiting the birds for the oracle's buff on cooldown is almost always a mistake. while probably not optimal, poachers is fucking hilarious on shaco | ||
|
Prog
United Kingdom1470 Posts
On December 08 2014 09:34 chalice wrote: that's an attitude problem, not an argument against statistics and analytics. LCS teams should be looking at solo queue as the NCAA to their NBA, not a worthless pick up game. Random teams/teammates and the largest amount of players having no professional relation to lol at all definitely sounds more like a pick up game than any attitude change could make up for. I'd say if the overwhelming amount of players treat something as a pick up game, then it won't change because a selected few treat it like the NCAA. | ||
|
Goumindong
United States3529 Posts
On December 08 2014 08:16 chalice wrote: i'm not asking for completed solutions, just ideas. like do people not think that it's at least possible there is a correlation between first blood/first dragon kill participation and winrate for someone who is jungling? or how about first tower and first dragon for bot laners? an adc that participates in an early game mid tower kill? Why would there be? As you go up in rank and start playing better players your "first dragon %" should fall. So having a high first dragon percent only tells us that your other areas are either disproportionately weak or that you're still climbing. It doesn't give us information about how good you are. The only thing that does that is your elo. I mean look, the reason we have elo is because chess masters were unable to objectively evaluate the play of people playing chess in front of their eyes. The same problem exists in league, not only is it hard to define objectively good play, but the actual information we have is thrown off by who you're playing in such a way that elo is the best. In a team game like league you would just look at the highest ranked players and then see if you thought their play style was right for your team. I don't think you could apply stats to pull hidden gems out since there are only so many people good enough to play at the top of the ladder and anyone else isn't really worth considering. | ||
|
chalice
United States1945 Posts
like when you're putting together maknoon's NA team, you don't think there is any way that with enough research you might be able to find a challenger or master tier jungler with the potential to be better than nintendude instead of just sorting by elo and name recognition? | ||
|
Slusher
United States19143 Posts
Or like when I play blotz, in games I win I kill myself at least once a game to set my team up or save somebody Kda is just really disconnected from context | ||
|
GolemMadness
Canada11044 Posts
| ||
|
Dark_Chill
Canada3353 Posts
I think stats in LoL are kind of useless in solo queue because pretty much nothing is controlled. Didn't get your best role? That'll make you worse. Your team has good synergy? That'll make you better. Judging individual players sounds like it you'd have to take one person, record his win percent against X(n) teams while playing on X(n) before you can even start to look at him individually. From there you might be able to see some patterns on that specific player which contribute to his wins across games. | ||
|
JimmiC
Canada22817 Posts
| ||
TheYango
United States47024 Posts
On December 08 2014 09:34 chalice wrote: that's an attitude problem, not an argument against statistics and analytics. LCS teams should be looking at solo queue as the NCAA to their NBA, not a worthless pick up game. The analogy fails because the game the NCAA is playing is a lot more similar to the game the NBA is playing than solo queue is to competitive LCS play. Which is why Kupon's "pickup game" comment is apt, because that's a lot closer to what solo queue actually is given the random teammates/no communication aspect. On December 08 2014 10:14 chalice wrote: like when you're putting together maknoon's NA team, you don't think there is any way that with enough research you might be able to find a challenger or master tier jungler with the potential to be better than nintendude instead of just sorting by elo and name recognition? Sure, and you don't need inaccurate and disconnected statistics to do it. The problems with Western organizations not getting new talent arguably has more to do with the organizations not wanting to take the risk than with them being unable to find talent. | ||
|
Goumindong
United States3529 Posts
On December 08 2014 10:14 chalice wrote: the point isn't to find some hidden Diamond 5 gem who will take the LCS by storm, but rather to evaluate the play of those near the top of the ladder in comparison with one another. like when you're putting together maknoon's NA team, you don't think there is any way that with enough research you might be able to find a challenger or master tier jungler with the potential to be better than nintendude instead of just sorting by elo and name recognition? Yes, you filter by champions played and take out the ones which don't play champions conductive to your strategy. But otherwise there isn't really a way. Maybe the basketball metrics people could come up with an idea, but I suspect not since there is very little individual matchups(evaluating shooting/passing essentially break down to a matchup) or independent events(shooting, more or less). And these area are where you find the value in those type of advanced metrics Additionally your biggest factors are communication and team cohesion, which can't be measured in soloqueue (except well, by ignoring ragers/ego's/drama queens) without playing with those people and straight up seeing how they play. | ||
|
chalice
United States1945 Posts
On December 08 2014 10:43 Dark_Chill wrote: Well, when someone before said the only useful stat was nexus died, let's look at that. Since LoL is a team game, isn't it more useful to first analyze a team before we go on individual players? I think stats in LoL are kind of useless in solo queue because pretty much nothing is controlled. Didn't get your best role? That'll make you worse. Your team has good synergy? That'll make you better. Judging individual players sounds like it you'd have to take one person, record his win percent against X(n) teams while playing on X(n) before you can even start to look at him individually. From there you might be able to see some patterns on that specific player which contribute to his wins across games. yeah, you definitely have to filter by role and for most categories probably champion as well. there's obviously no point in averaging in wild turtle's stats playing the support role in with his ADC games and his early game stats between games played with vayne and those with say lucian or corki are likely significantly different. the good thing is that there's enough data available that breaking things down by champion should still give you relatively decent sample sizes. | ||
|
killerdog
Denmark6522 Posts
On December 08 2014 09:21 Kupon3ss wrote: There are some stats that can be derived from controlled situations like LCS but nobody cares about the lol equivalent of your stats from a pickup game on the street Well if i could analyse the data of distance from the hoop every shot was fired from, which hand was used, player height and whatever else i wanted for thousands (or millions if just scraping general soloq stats) of pickup games of basketball, I'd imagine I'd spot some trends. Which while maybe not technically "useful" (whatever that word means in this context anyway) but the correlation between the various variables (what's the average relationship between height and accuracy, is distance -> accuracy a linear relationship, does left vs right handed accuracy correspond to the proportion of people who are right/left handed? etc etc) You're doing a massive disservice to the potential for data analysis in soloq by implying that we can't draw larger samples, and more precise data from scraping lol stats from soloq then from watching one pickup game of street basket... | ||
|
Prog
United Kingdom1470 Posts
On December 08 2014 11:11 killerdog wrote: Well if i could analyse the data of distance from the hoop every shot was fired from, which hand was used, player height and whatever else i wanted for thousands (or millions if just scraping general soloq stats) of pickup games of basketball, I'd imagine I'd spot some trends. Which while maybe not technically "useful" (whatever that word means in this context anyway) but the correlation between the various variables (what's the average relationship between height and accuracy, is distance -> accuracy a linear relationship, does left vs right handed accuracy correspond to the proportion of people who are right/left handed? etc etc) You're doing a massive disservice to the potential for data analysis in soloq by implying that we can't draw larger samples, and more precise data from scraping lol stats from soloq then from watching one pickup game of street basket... As soon as you bring in larger samples you can ignore anything but win/loss though. The dilemma is this: either you focus on particular stats of single games, which are meaningless, or you take a larger sample in which anything but win/loss is unnecessary. Either way more data (than win/loss) seems pretty useless to capture skill. | ||
|
Goumindong
United States3529 Posts
On December 08 2014 11:11 killerdog wrote: Well if i could analyse the data of distance from the hoop every shot was fired from, which hand was used, player height and whatever else i wanted for thousands (or millions if just scraping general soloq stats) of pickup games of basketball, I'd imagine I'd spot some trends. Which while maybe not technically "useful" (whatever that word means in this context anyway) but the correlation between the various variables (what's the average relationship between height and accuracy, is distance -> accuracy a linear relationship, does left vs right handed accuracy correspond to the proportion of people who are right/left handed? etc etc) You're doing a massive disservice to the potential for data analysis in soloq by implying that we can't draw larger samples, and more precise data from scraping lol stats from soloq then from watching one pickup game of street basket... Actually we can probably get more info out of a pickup baseketball game. Since basketball has a lot of individual contests(which we can measure spatially) and independent events for us to measure while league does not. edit: Basketball also is won/lost by how many points each team has, so any action which directly leads to points necessarily directly leads to winning/losing. League has no such stat in addition to the problems expressed earlier in deriving meaning from the stats it has. The only stat I can really think of that you could use would be "Attempted CS %" Or "of cs you're not competing for with allies and which you attempt to get how often do you get that CS". But that stat only works for 3 roles and isn't tracked. | ||
|
Kupon3ss
時の回廊10066 Posts
The West thinks that actual 5v5 matches are the same as soloQ Wheras the Asian teams know that they are not and treat the two as completely different things | ||
|
chalice
United States1945 Posts
On December 08 2014 12:20 Kupon3ss wrote: That is actually the biggest difference between Eastern and Western LoL The West thinks that actual 5v5 matches are the same as soloQ Wheras the Asian teams know that they are not and treat the two as completely different things yeah, korean solo queue talent has basically never been able translate into competitive success. | ||
|
red_
United States8474 Posts
On December 08 2014 11:56 Goumindong wrote: edit: Basketball also is won/lost by how many points each team has, so any action which directly leads to points necessarily directly leads to winning/losing. League has no such stat in addition to the problems expressed earlier in deriving meaning from the stats it has. Done. I'm not saying there are literally no possible metrics for League(or Dota or w/e video game), but basketball, and baseball(the crown jewel of sports metrics) are decided by one single value: points(runs), which can be accrued in a limited amount of time(or innings). I really have absolutely no clue how you're going to find the League equivalent to league and defense adjusted ERA or true shooting %, or whatever, when there is no clear stat that says 'win' other than a destroyed nexus. You're basically resigned to making up stats and looking for correlations until you find one that maybe sticks(until a patch changes the game and the stats). | ||
| ||
