|
On September 23 2015 12:30 Ketara wrote:
Most players I would imagine, if they are first pick, play the same role every time.
It seems to me like it would be easy to use data from champ select to figure out what roles a person is likely to prefer to play, and then average that out amongst the while population.
I have the data. Do you want to work on it and report back?
|
On September 23 2015 12:28 Sufficiency wrote:Show nested quote +On September 23 2015 12:26 Gahlo wrote:On September 23 2015 12:24 Sufficiency wrote:On September 23 2015 12:20 Gahlo wrote:On September 23 2015 12:15 Goumindong wrote:
Yea, i have no idea how to construct a statistic which is going to tell you your average number of roles per 5 games or what not Where are you getting /5 games from? I'm sure somewhere Riot has a log of all the ranked games played this season. Do they have Teleport? Probably a top laner. Smite? Probably a jungler. Playing an ADC? Probably AS. Have a support item? Probably support. You can make very general assumptions to narrow down the most possible position played in that game without even having to look at full rosters. Yes, I realized that I didn't list anything for Mid, but that falls under process of elimination. That part is easy. I have all of those in my own database. The problem is how Lyte defines a player to "play position X". If I played top Vladimir once because I was forced to, am I a top lane player now? Or is it based on self-reports like Ketara suggested? By the way I highly doubt that's what they did because that's really complicated. On September 23 2015 12:13 Gahlo wrote:On September 23 2015 12:06 Sufficiency wrote:On September 23 2015 11:58 Gahlo wrote:On September 23 2015 11:55 Sufficiency wrote:Lyte said: On average, Ranked players only play 2.2 positions There is no way this is true. Then again, it's not clear at all what he actually means, so who knows. The average doesn't fit into a while #. If somebody like me plays Jungle/Support and another player plays Top/Jungle/Mid, we play an average of 2.5 positions. The problem is what he means by "play 2.2 positions". What does he mean by "play X position"? If I am forced to play top in one game out of one hundred (even though I clearly prefer jungle), am I considered to have played the top position? They probably look at it and go "This guy plays 40% jungle, 30% support, 15% top, 10% AD, and 5% mid. Probably safe to say he plays 2 positions." So you are saying I can name my own cutoff (you set it at 30%, for example) to suit my own needs? What is so magical about 30%?
I made up the position splits off the top of my head for a nonexistent player. I'd think if they were a top/jungle/support player they'd have more top games. Either way, like I said, Riot probably has all the games played this season to look at and find patterns to crunch #s on that provides a pretty clear cutoff depending on how many positions a player seems to play.
|
On September 23 2015 12:38 Gahlo wrote:Show nested quote +On September 23 2015 12:28 Sufficiency wrote:On September 23 2015 12:26 Gahlo wrote:On September 23 2015 12:24 Sufficiency wrote:On September 23 2015 12:20 Gahlo wrote:On September 23 2015 12:15 Goumindong wrote:
Yea, i have no idea how to construct a statistic which is going to tell you your average number of roles per 5 games or what not Where are you getting /5 games from? I'm sure somewhere Riot has a log of all the ranked games played this season. Do they have Teleport? Probably a top laner. Smite? Probably a jungler. Playing an ADC? Probably AS. Have a support item? Probably support. You can make very general assumptions to narrow down the most possible position played in that game without even having to look at full rosters. Yes, I realized that I didn't list anything for Mid, but that falls under process of elimination. That part is easy. I have all of those in my own database. The problem is how Lyte defines a player to "play position X". If I played top Vladimir once because I was forced to, am I a top lane player now? Or is it based on self-reports like Ketara suggested? By the way I highly doubt that's what they did because that's really complicated. On September 23 2015 12:13 Gahlo wrote:On September 23 2015 12:06 Sufficiency wrote:On September 23 2015 11:58 Gahlo wrote:On September 23 2015 11:55 Sufficiency wrote:Lyte said: On average, Ranked players only play 2.2 positions There is no way this is true. Then again, it's not clear at all what he actually means, so who knows. The average doesn't fit into a while #. If somebody like me plays Jungle/Support and another player plays Top/Jungle/Mid, we play an average of 2.5 positions. The problem is what he means by "play 2.2 positions". What does he mean by "play X position"? If I am forced to play top in one game out of one hundred (even though I clearly prefer jungle), am I considered to have played the top position? They probably look at it and go "This guy plays 40% jungle, 30% support, 15% top, 10% AD, and 5% mid. Probably safe to say he plays 2 positions." So you are saying I can name my own cutoff (you set it at 30%, for example) to suit my own needs? What is so magical about 30%? I made up the position splits off the top of my head for a nonexistent player. I'd think if they were a top/jungle/support player they'd have more top games. Either way, like I said, Riot probably has all the games played this season to look at and find patterns to crunch #s on that provides a pretty clear cutoff depending on how many positions a player seems to play.
Or I can invent a number to fit my narrative, like what you did with the 30%.
The fact of the matter is, "2.2 positions" is problematic. If this is the kind of evidence-based decision making that Riot engages in, it's terrible.
|
On September 23 2015 12:32 Sufficiency wrote:Show nested quote +On September 23 2015 12:30 Ketara wrote:
Most players I would imagine, if they are first pick, play the same role every time.
It seems to me like it would be easy to use data from champ select to figure out what roles a person is likely to prefer to play, and then average that out amongst the while population. I have the data. Do you want to work on it and report back?
I am on a train in Myanmar going through a war zone at the moment, sorry.
|
Maybe they did a survey (probably not)?
|
On September 23 2015 12:42 Sufficiency wrote:Show nested quote +On September 23 2015 12:38 Gahlo wrote:On September 23 2015 12:28 Sufficiency wrote:On September 23 2015 12:26 Gahlo wrote:On September 23 2015 12:24 Sufficiency wrote:On September 23 2015 12:20 Gahlo wrote:On September 23 2015 12:15 Goumindong wrote:
Yea, i have no idea how to construct a statistic which is going to tell you your average number of roles per 5 games or what not Where are you getting /5 games from? I'm sure somewhere Riot has a log of all the ranked games played this season. Do they have Teleport? Probably a top laner. Smite? Probably a jungler. Playing an ADC? Probably AS. Have a support item? Probably support. You can make very general assumptions to narrow down the most possible position played in that game without even having to look at full rosters. Yes, I realized that I didn't list anything for Mid, but that falls under process of elimination. That part is easy. I have all of those in my own database. The problem is how Lyte defines a player to "play position X". If I played top Vladimir once because I was forced to, am I a top lane player now? Or is it based on self-reports like Ketara suggested? By the way I highly doubt that's what they did because that's really complicated. On September 23 2015 12:13 Gahlo wrote:On September 23 2015 12:06 Sufficiency wrote:On September 23 2015 11:58 Gahlo wrote:On September 23 2015 11:55 Sufficiency wrote:Lyte said: On average, Ranked players only play 2.2 positions There is no way this is true. Then again, it's not clear at all what he actually means, so who knows. The average doesn't fit into a while #. If somebody like me plays Jungle/Support and another player plays Top/Jungle/Mid, we play an average of 2.5 positions. The problem is what he means by "play 2.2 positions". What does he mean by "play X position"? If I am forced to play top in one game out of one hundred (even though I clearly prefer jungle), am I considered to have played the top position? They probably look at it and go "This guy plays 40% jungle, 30% support, 15% top, 10% AD, and 5% mid. Probably safe to say he plays 2 positions." So you are saying I can name my own cutoff (you set it at 30%, for example) to suit my own needs? What is so magical about 30%? I made up the position splits off the top of my head for a nonexistent player. I'd think if they were a top/jungle/support player they'd have more top games. Either way, like I said, Riot probably has all the games played this season to look at and find patterns to crunch #s on that provides a pretty clear cutoff depending on how many positions a player seems to play. Or I can invent a number to fit my narrative, like what you did with the 30%. The fact of the matter is, "2.2 positions" is problematic. If this is the kind of evidence-based decision making that Riot engages in, it's terrible.
Fuck it, I'm done. I can't believe you're this dense. Have fun figuring it out with your head that far up your ass.
|
|
|
~Analysis is more important than statistics as always~
But good luck convincing people of that.
|
|
|
Yeah I'm not really understanding the senseless excuse to bash Riot out of sufficiency, here.
|
On September 23 2015 12:59 Ketara wrote:
Yeah I'm not really understanding the senseless excuse to bash Riot out of sufficiency, here.
It's not a senseless bash.
I was merely pointing out the illogical part of Lyte's answer. Then for some reason Gahlo believes he can solve the problem by setting a random 30% cut off, which is completely arbitrary and chosen to simply fit his narrative. Now he calls me an ass for no reason.
It's really confusing.
|
Complaining about completely arbitrarily picked numbers and implying Riot is doing the same sure is a thing.
|
On September 23 2015 13:03 Caiada wrote:
Complaining about completely arbitrarily picked numbers and implying Riot is doing the same sure is a thing.
That's a fair point, which is why I am taking out the data and try to figure out how Riot is arriving at this magical 2.2.
No, I do not believe Lyte just randomly made it up. Yes, I do think the number is problematic in a variety of different ways - hopefully this is obvious.
|
On September 23 2015 12:30 Ketara wrote:
Most players I would imagine, if they are first pick, play the same role every time.
It seems to me like it would be easy to use data from champ select to figure out what roles a person is likely to prefer to play, and then average that out amongst the while population.
Average doesn't work because its all the same value. And roles aren't cardinal so there isn't a way to make any sense of an average.
Role 1% + Role 2% + so on so forth = 1 trivially. The "average" of roles you play sums to one for the same reason. You have to either play multiple roles per game, or mess around with how you count, in order to get values that aren't 1.
|
It's not no reason.
Your posting has the tendency to really bother people. You act like you're so much smarter than everyone while at the same time rarely doing much to actually improve whatever conversation youre a part of.
You know what Lyte is trying to say. Based off common sense it's completely reasonable. But you're taking issue with the lack of detail in the statement, then using that lack of detail to insert your own narrative that it's problematic, impossible, terrible or whatever, as if there's no possible way Riot could have more data than you, better statisticians and analysts than you, or more manpower and time than you to come to these conclusions.
Then when we are like hey, his statement seems reasonable and plausible, you attack us because we don't have the data or whatever. You make a personal attack against me that's ridiculous and petty, and rather than report you for it I just make a joke and shrug it off, but I can definitely see why Gahlo is angry.
That doesn't mean his post is warranted, but the position he's coming from is perfectly reasonable.
|
On September 23 2015 13:10 Ketara wrote:
It's not no reason.
Your posting has the tendency to really bother people. You act like you're so much smarter than everyone while at the same time rarely doing much to actually improve whatever conversation youre a part of.
You know what Lyte is trying to say. Based off common sense it's completely reasonable. But you're taking issue with the lack of detail in the statement, then using that lack of detail to insert your own narrative that it's problematic, impossible, terrible or whatever, as if there's no possible way Riot could have more data than you, better statisticians and analysts than you, or more manpower and time than you to come to these conclusions.
Then when we are like hey, his statement seems reasonable and plausible, you attack us because we don't have the data or whatever. You make a personal attack against me that's ridiculous and petty, and rather than report you for it I just make a joke and shrug it off, but I can definitely see why Gahlo is angry.
That doesn't mean his post is warranted, but the position he's coming from is perfectly reasonable.
Yes, I do have a lot of data stored. No, I have not tried to use it to justify my argument or even remotely analyzed it. I am actually not that interested. I also don't have the time.
If you think I "personally attacked you" in a way that is "ridiculous and petty" because I asked if you want to analyze the data, you are sadly mistaken. I am just curious how someone else will approach the problem. Sorry if you think I was being passive aggressive, that was not my intention. I apologize if I offended you.
Like I have said, Riot has misused their statistics in the past. I have already mentioned in a previous post that one of the games' tips is something like "XX% of the players who follows the Summoners' Code wins YY% more games" or along that line, which you can see the similar kind of problems.
|
I would imagine that they put much more work into stuff like that 2.2 figure than they do in champ select tips.
Can you really tell me that if it was your paid job to figure out how many roles people played on average, you had access to ALL of Riots game data (including things like chat logs and reports that you might not currently have) and had weeks to solve this problem, that it would be impossible?
Would it even be difficult?
|
On September 23 2015 13:25 Ketara wrote:
I would imagine that they put much more work into stuff like that 2.2 figure than they do in champ select tips.
Can you really tell me that if it was your paid job to figure out how many roles people played on average, you had access to ALL of Riots game data (including things like chat logs and reports that you might not currently have) and had weeks to solve this problem, that it would be impossible?
Would it even be difficult?
A help string on the client that is frequently displayed requires far more vigour than a number that is on a AskFM post.
Now, for your question as to whether or not the problem can be solved - yes, if you ask it that way. In fact most real life problems can be solved by throwing resources and time onto it.
Would the problem be difficult though? It could be difficult - depending on what underlying assumptions are taken and what Lyte means by "Player X plays Role Y"
|
On September 23 2015 13:25 Ketara wrote:
I would imagine that they put much more work into stuff like that 2.2 figure than they do in champ select tips.
Can you really tell me that if it was your paid job to figure out how many roles people played on average, you had access to ALL of Riots game data (including things like chat logs and reports that you might not currently have) and had weeks to solve this problem, that it would be impossible?
Would it even be difficult?
Only if you are riot, and stubbornly cling to your assumptions regardless of evidence .
|
Aren't there already formulas to calculate what would be a statistically significant difference between certain percentages of gameplay in a certain role?
+ Show Spoiler +I mean, if you have someone who plays ever role equally, that's 20/20/20/20/20% and he would play 5 roles.
It becomes alot harder for someone who plays 30/15/20/15/20 or someone with even less elegant numbers. But I guess there must be some cutoff. You can even go case by case, add all the ones that vary within a certain range, let's take previous example, which could become 30[+-3]/15[+-3]... and come up with a cutoff that could be relevant.
Also, different groups with different percentages would get a different deviation.
Then after calculating howmany roles each group plays, you can simply go for a mean calculation.
The 2.2 roles isn't really that problematic in that way imo.
Another maybe statistical intriguing question is: what kind of duo has the highest win%; is it the duo's bot? duo's jungle/mid? random roledivision?
|
|
|
|
|
|
EPT
