|
On July 13 2010 07:25 teamamerica wrote:
I tried to read through most of this and I understand most of it (I hope!). I'm just not sure about how promootions work. I'm stuck in a gold league with NEXGenius at the top. He's has an amazing record and is consistently matched against (and beating) diamond leaguers (from the crazy league with Idra and Whitera and all the others). He was about 700 pts, 200 more then 2nd place, which is about 200 more then the next place. How come he hasn't been promoted. In 2v2 I've been promoted twice already, despite never reaching the top of my bronze or silver league (now gold).
Some ppl just aren't getting promoted for some reason, we aren't sure why... I believe.
|
In theory this would probably make sense... but in practice the problem is that the players are not NOT randomly distributed. Most of the best players will probably get placed into one of the earlier divisions, while the skill level will gradually decrease as the date of creation of a new division is later.
|
My personal opinion on the subject (as if anyone cares = 0) is that the division system is actually pretty cool. It does what blizzard wanted it to for me a casual player. I do however think there should be some kind of overall comparison for those people that actually are near the top.
Something as simple as a top 100 players list separate from the current divisions.
|
That may sound a bit harsh, but you even say:
I think blizzard may implement some way to weed out inactive accounts from the calculations for rank in the future.
So: How can you claim to have math on your side if one of your premises is wrong?
If inactive accounts are not removed from leagues your whole analysis is not valid, because of the many "dead" accounts in the different divisions.
|
On July 13 2010 07:45 Zocat wrote:That may sound a bit harsh, but you even say: Show nested quote +I think blizzard may implement some way to weed out inactive accounts from the calculations for rank in the future. So: How can you claim to have math on your side if one of your premises is wrong? If inactive accounts are not removed from leagues your whole analysis is not valid, because of the many "dead" accounts in the different divisions.
It doesn't have all that big an effect to be honest. You'd probably be pretty hard-pressed to find a division with enough inactive players that are inactive long enough to actually skew the data. Frankly I don't even know for sure that Blizzard doesn't have a system to remove these players from divisions already.
|
I can be rank #1 in my division for months, or even years, because the other 99 people stopped playing. That doesn't make me god's gift to Starcraft.
|
On July 13 2010 07:49 Issorlol wrote:Show nested quote +On July 13 2010 07:45 Zocat wrote:That may sound a bit harsh, but you even say: I think blizzard may implement some way to weed out inactive accounts from the calculations for rank in the future. So: How can you claim to have math on your side if one of your premises is wrong? If inactive accounts are not removed from leagues your whole analysis is not valid, because of the many "dead" accounts in the different divisions. It doesn't have all that big an effect to be honest. You'd probably be pretty hard-pressed to find a division with enough inactive players that are inactive long enough to actually skew the data. Frankly I don't even know for sure that Blizzard doesn't have a system to remove these players from divisions already.
Unless they are stupid they must have thought of this. I'd guess that have something in mind for inactive players, though what is anyone's guess.
|
On July 13 2010 07:19 FabledIntegral wrote:Show nested quote +On July 13 2010 06:22 Issorlol wrote:On July 13 2010 06:21 FabledIntegral wrote:
100 is not gigantic whatsoever statistically speaking... almost no survey will use a sample as small as 100 ppl if you're trying to compare to the rest of the population. You are incorrect. .... almost any survey ever will use thousands of responses to try to give any relatively accurate results. If you only have 100 people you can not hope to say you are coming close... I'm definitely rusty on statistics but after doing a bit of googling the first ~2 pages I came up with all suggested sample sizes much larger than 100. Just found another... (keep in mind I'm just doing some googling) and I found this, look at the second page http://www.stocktongov.com/auditor/documents/ExplanationofSurveySampleSize.pdf"A sample size of 400 is one commonly chosen by local governments for resident policy surveys because a margin of error of ±5% is felt to be acceptable to government officials and the public at large." Although I'm confused because I remember the "±5%" number not being on a static number like 400... but depending on a multitude of other factors... including standard deviations... which is going into something completely different (man and I just took econ stats like a year ago :S).
The margin of error for 100 is going to be like 10%ish and with fluidity of the ranking system, that's really quite ok. 100 is a large sample, if you're really worried about it and want to have a really small margin of error, you go to ~1100. I don't think for this that's necessary for this though. 500 or 1000 people league would be really convincing samples (IF they take care of a lot of the other flaws, random shuffling of the people within each league would take care of a lot of it).
I think at diamond having one big bracket (or at the very least, larger like 1000 person brackets) would be really nice and for the lower divisions keep it as is, so the more casual player can have the feel of achievement/be able to see themselves move up the rank at a more percentile level (imagine you're an average silver or gold player and each win/loss moves you up or down literally thousands because you're right in the middle) while at diamond the much more competitive players can see how they rank more exactly.
|
On July 13 2010 07:49 Issorlol wrote:Show nested quote +On July 13 2010 07:45 Zocat wrote:That may sound a bit harsh, but you even say: I think blizzard may implement some way to weed out inactive accounts from the calculations for rank in the future. So: How can you claim to have math on your side if one of your premises is wrong? If inactive accounts are not removed from leagues your whole analysis is not valid, because of the many "dead" accounts in the different divisions. It doesn't have all that big an effect to be honest. You'd probably be pretty hard-pressed to find a division with enough inactive players that are inactive long enough to actually skew the data. Frankly I don't even know for sure that Blizzard doesn't have a system to remove these players from divisions already.
I think it would be better to assume that there is some probability that your current rank represents your percentile overall at non-top-level-Diamond play. Because unless each division has the same distribution of skill as another division (which I doubt 5 games will do accurately, unless there is some math we don't see in the background that moves people around TO make the divisions do this) then your rank is not representative of your percentile. The top 1 in a division can be just as shitty as the 60th in another, how can you tell? I am not considering top-Diamond players here.
|
Doesn't this whole argument assume that all divisions in a league are roughly the same skill?
~LoA
|
I still think a single ladder system would be better than divisions. The system is designed to make children feel better about themselves much like how schools award half a mark for a completely wrong answer ("they tried"). It is just ridiculous.
I want an accurate representation of how well I am doing so I can clearly see improvement and clearly see how I compare to other people on the ladder. The division system does not deliver.
Show true ranks or don't bother.
|
I understand exactly what the OP means and theoretically he is correct. However, as some people have mentioned, this is not going to be accurate for top diamond players. For example, Diamond Division "Blah Blah" may have 3 players all better than the #1 player in another diamond division. So you cannot say that player A, B, and C in Blah Blah are in the top 1%, 2% and 3% respectively and then that the other number 1 player in the other league is in the top 1%. That would be unfair to players B and C in division Blah Blah. To say after time, the diamond divisions will all even out is just plain ignorant.
So my point is that overall this divisional system will do well for the most part, but players should not be robbed of a global ranking, or at least regional ranking (US, EU, Asian). I think that there are plenty of players out there that what to know where they stand as a rank and not a theoretical percentage.
|
I read through the op, and i understand what is being suggested... But it is simply wrong... There is nothing to say that entire divisions can't be within the top 1 or 2 percentiles, or the 98 or 99 conversely. There is absolutely nothing to suggest that a player who is rank 1 in a poorly skilled division will be better than a rank 80 in a highly skilled division...
The op is assuming that the each division will always contain a wide gamut of skill levels. If this was the case I would definitely agree... But a sample of 100 players out of what could potentially be well over 100,000 playera in each league, is far from wide enough to give u a true percentile based ranking...
I would actually really like to see a statisticians point of view on the op's theory.
|
I don't see why they can't just have the divisions and a global rank at the same time. You could know that you are 5th in your division and feel a sort of satisfaction for that I guess but then you could just go and get a simple number of how you stand against everyone.
With a global rank you could still go and see what % you are of your overall league, just find out how many total people are ranked in diamond globally, then do some simple math to figure out what percentage of the diamond level you are.
|
On July 13 2010 08:22 Forlorn wrote:
I don't see why they can't just have the divisions and a global rank at the same time. You could know that you are 5th in your division and feel a sort of satisfaction for that I guess but then you could just go and get a simple number of how you stand against everyone.
With a global rank you could still go and see what % you are of your overall league, just find out how many total people are ranked in diamond globally, then do some simple math to figure out what percentage of the diamond level you are.
This is exactly what they need to do :/
|
On July 13 2010 08:00 Andtwo wrote:Show nested quote +On July 13 2010 07:19 FabledIntegral wrote:On July 13 2010 06:22 Issorlol wrote:On July 13 2010 06:21 FabledIntegral wrote:
100 is not gigantic whatsoever statistically speaking... almost no survey will use a sample as small as 100 ppl if you're trying to compare to the rest of the population. You are incorrect. .... almost any survey ever will use thousands of responses to try to give any relatively accurate results. If you only have 100 people you can not hope to say you are coming close... I'm definitely rusty on statistics but after doing a bit of googling the first ~2 pages I came up with all suggested sample sizes much larger than 100. Just found another... (keep in mind I'm just doing some googling) and I found this, look at the second page http://www.stocktongov.com/auditor/documents/ExplanationofSurveySampleSize.pdf"A sample size of 400 is one commonly chosen by local governments for resident policy surveys because a margin of error of ±5% is felt to be acceptable to government officials and the public at large." Although I'm confused because I remember the "±5%" number not being on a static number like 400... but depending on a multitude of other factors... including standard deviations... which is going into something completely different (man and I just took econ stats like a year ago :S). The margin of error for 100 is going to be like 10%ish and with fluidity of the ranking system, that's really quite ok. 100 is a large sample, if you're really worried about it and want to have a really small margin of error, you go to ~1100. I don't think for this that's necessary for this though. 500 or 1000 people league would be really convincing samples (IF they take care of a lot of the other flaws, random shuffling of the people within each league would take care of a lot of it). I think at diamond having one big bracket (or at the very least, larger like 1000 person brackets) would be really nice and for the lower divisions keep it as is, so the more casual player can have the feel of achievement/be able to see themselves move up the rank at a more percentile level (imagine you're an average silver or gold player and each win/loss moves you up or down literally thousands because you're right in the middle) while at diamond the much more competitive players can see how they rank more exactly.
How is 10% "really quite ok," especially when it's ±10%... when you're dealing with a population in the hundreds of thousands, being 10,000 off is kinda a big thing to me personally. Maybe it's ok with you but I see zero benefit in using this system.
|
On July 13 2010 08:20 skipgamer wrote:
I read through the op, and i understand what is being suggested... But it is simply wrong... There is nothing to say that entire divisions can't be within the top 1 or 2 percentiles, or the 98 or 99 conversely. There is absolutely nothing to suggest that a player who is rank 1 in a poorly skilled division will be better than a rank 80 in a highly skilled division...
The op is assuming that the each division will always contain a wide gamut of skill levels. If this was the case I would definitely agree... But a sample of 100 players out of what could potentially be well over 100,000 playera in each league, is far from wide enough to give u a true percentile based ranking...
I would actually really like to see a statisticians point of view on the op's theory.
Those of us with degrees in mathematics have already explained that he's absolutely wrong, and why he's wrong. And other people apparently didn't need the background to see it as well. If you read through the thread, you'll see it 
|
On July 13 2010 04:42 Issorlol wrote:
Your division is a sample of the entire population of the league in which you place. This means that your rank in your division is, over a long period of time with many games played by the community as a whole, indicative of your rank as a whole compared to everyone in your league by percentile
That means if you play 300 games and you're rank 89 in your division... You're as good as 89% of the other people in your league.
So if I'm rank 89 I'm then "as good" as someone at rank 1? Huh?
Or, do you mean I'm "as good" as everyone in the bottom 89%? (I'm "as good" as someone at rank 12 if I'm at rank 89.)
I'm sorry, but if you can't communicate what you're thinking with clarity and accuracy then that makes it hard for me to believe you know what you're talking about. Perhaps you do, but I'm not going to bother deciphering your confusing statements.
As it is right now, I don't think any of us truly know what ranking criteria Blizzard is using. I also don't think any of us yet know what meaning the different ranks within leagues might have for true competition (if there is any meaning in them at all). For instance, saying that all rank 89 players in platinum league are of equivalent skill seems like a gigantic logical leap to me. (Heck, even saying all Platinum players are better than Silver players might not be exactly accurate.) That could be the way it works, but has Blizzard revealed any of the mechanics behind the system yet? If you could inform of us of the precise data that firmly tells you everything you're claiming to know, that would also be helpful to alleviate my skepticism. Thanks.
|
I see the point you're trying to make, but there are a few things I have problems with.
If you're placed in a diamond division, with no room to go up, and everyone in your devision is really good, then your rank would not be indicative of your percentile rank. For example, if everyone in your division is winning with a 80% win rate, and you're rank 80 with a 75% win rate, it's clear that you're better than 80% of diamond players (assuming the matchmaking system is doing its job and not matching you vs silvers etc). If there was some way implemented to shuffle players through different divisions so that each division has an equal sum of skill level, then I would agree more readily with your post. (What works in theory doesn't always work in practice)
My second point is if the leagues still existed, but divisions were removed, then how is that less accurate of a skill ranking? If you can see that you are rank 4000 out of the 25000 people in your division, then doesn't it just take some simple math to realize that you are in the top 16%? That seems a lot more accurate in theory than the division system, because theres no chance of being put into a "division of death".
|
On July 13 2010 08:51 Tiptup wrote:Show nested quote +On July 13 2010 04:42 Issorlol wrote:
Your division is a sample of the entire population of the league in which you place. This means that your rank in your division is, over a long period of time with many games played by the community as a whole, indicative of your rank as a whole compared to everyone in your league by percentile
That means if you play 300 games and you're rank 89 in your division... You're as good as 89% of the other people in your league. So if I'm rank 89 I'm then "as good" as someone at rank 1? Huh? Or, do you mean I'm "as good" as everyone in the bottom 89%? (I'm "as good" as someone at rank 12 if I'm at rank 89.) I'm sorry, but if you can't communicate what you're thinking with clarity and accuracy then that makes it hard for me to believe you know what you're talking about. Perhaps you do, but I'm not going to bother deciphering your confusing statements. As it is right now, I don't think any of us truly know what ranking criteria Blizzard is using. I also don't think any of us yet know what meaning the different ranks within leagues might have for true competition (if there is any meaning in them at all). For instance, saying that all rank 89 players in platinum league are of equivalent skill seems like a gigantic logical leap to me. (Heck, even saying all Platinum players are better than Silver players might not be exactly accurate.) That could be the way it works, but has Blizzard revealed any of the mechanics behind the system yet? If you could inform of us of the precise data that firmly tells you everything you're claiming to know, that would also be helpful to alleviate my skepticism. Thanks.
Think of "as good or better than." Aka if you're 89%, 11% are better than you.
|
|
|
|
|
|
EPT
