EPTZerg rule finally toppled - Page 6
| Forum Index > SC2 General |
|
a_flayer
Netherlands2826 Posts
| ||
|
FeyFey
Germany10114 Posts
| ||
|
Kenny_oro
Germany368 Posts
 | ||
|
Canucklehead
Canada5074 Posts
| ||
|
Otolia
France5805 Posts
On April 19 2013 02:20 Canucklehead wrote: Errr I dunno what polt has done lately to be ranked number 2. And forgg 4th? He's not very good. I dunno if he's still 1-1-1 against every race in Hots, like he did in Wol though. Any list that has forgg in the top 5 shouldn't be taken seriously. These lists are an interesting talking point, but not that accurate I find. Except for life. He's the number 1 bonjwa in the scene right now. This is not a list, it's a ranking. It's not an appraisal or an opinion but the result of a scientifically correct representation of the players skill. Unless you have mathematically correct definition of accuracy in this present case, what you said is just an opinion and a not very forgiving one at that. | ||
|
Caladbolg
2855 Posts
| ||
|
m0nt
Australia80 Posts
| ||
|
Canucklehead
Canada5074 Posts
On April 19 2013 02:26 Otolia wrote: This is not a list, it's a ranking. It's not an appraisal or an opinion but the result of a scientifically correct representation of the players skill. Unless you have mathematically correct definition of accuracy in this present case, what you said is just an opinion and a not very forgiving one at that. You must not be a very critical thinker and take things at face value. If you were you would know the ranking is based on a formula made by a human and is not infallible. I'm saying the formula used to come up with this ranking has flaws in it. I'm not sure why you think it produces an accurate ranking just because it uses fancy stats. Do you take any list a statistician comes up with as fact without questioning the methods behind them? That's a serious question, since you just assumed everything was a "scientifically correct representation of the players skill" because he told you it was. | ||
|
Otolia
France5805 Posts
On April 19 2013 02:37 Canucklehead wrote: You must not be a very critical thinker and take things at face value. If you were you would know the ranking is based on a formula made by a human and is not infallible. I'm saying the formula used to come up with this ranking has flaws in it. I'm not sure why you think it produces an accurate ranking just because it uses fancy stats. Do you take any list a statistician comes up with as fact without questioning the methods behind them? That's a serious question, since you just assumed everything was a "scientifically correct representation of the players skill" because he told you it was. I have access to code. I've read it and understood it. I doubt you can say the same. The most stupid part in your post is you considering that your opinion on who should stand at the top is relevant to the accuracy of the ratings. The ratings are accurate, the formula may not be the best or the most precise but the ratings are accurate according to the system. There is nothing more to it. So unless you actually have a worthwhile suggestion to make, you should stop being an ass to people who are dedicating a significant part of their life to a project like this one. Edit : And so we are clear, I am not pretending that everything is right and that criticism isn't acceptable. I don't like ForGG being 4th either as it isn't reflecting his actual skill level but saying that the system overall is bad because of a single case is beyond absurd. | ||
|
Grovbolle
Denmark3811 Posts
On April 19 2013 02:37 Canucklehead wrote: You must not be a very critical thinker and take things at face value. If you were you would know the ranking is based on a formula made by a human and is not infallible. I'm saying the formula used to come up with this ranking has flaws in it. I'm not sure why you think it produces an accurate ranking just because it uses fancy stats. Do you take any list a statistician comes up with as fact without questioning the methods behind them? That's a serious question, since you just assumed everything was a "scientifically correct representation of the players skill" because he told you it was. Please remember that all our site does is track and distribute points. I refer to my post on page 3 with a big-ass picture explaining how points are distributed. Please don't just come in bashing the entire project because ForGG is pretty highly rated. | ||
|
Canucklehead
Canada5074 Posts
On April 19 2013 02:52 Grovbolle wrote: Please don't just come in bashing the entire project because ForGG is pretty highly rated. Why not? That is a pretty big flaw in the system if it produces a ranking that has forgg in the top 5. You guys are pretty sensitive and not open to any criticism I see. A person who took their system more seriously and strove for accuracy would try to figure out why forgg got ranked so high and try to fix the flaws instead of pretending no flaws exist. | ||
|
Swift118
United Kingdom335 Posts
On April 19 2013 02:20 Canucklehead wrote: Errr I dunno what polt has done lately to be ranked number 2. And forgg 4th? He's not very good. I dunno if he's still 1-1-1 against every race in Hots, like he did in Wol though. Any list that has forgg in the top 5 shouldn't be taken seriously. These lists are an interesting talking point, but not that accurate I find. Except for life. He's the number 1 bonjwa in the scene right now. Any post saying Forgg is not very good should not be taken seriously. | ||
|
MasterOfPuppets
Romania6942 Posts
On April 19 2013 02:13 a_flayer wrote: I refuse to understand how this works, but the fact that ForGG is in that list makes it pretty moot to me. If you refuse to put any effort into understanding this, then why post at all? | ||
|
Conti
Germany2516 Posts
On April 19 2013 02:57 Canucklehead wrote: Why not? That is a pretty big flaw in the system if it produces a ranking that has forgg in the top 5. You guys are pretty sensitive and not open to any criticism I see. A person who took their system more seriously and strove for accuracy would try to figure out why forgg got ranked so high and try to fix the flaws instead of pretending no flaws exist. Perhaps there should be a FAQ with the most common and/or obvious "fixes" to the ranking system somewhere. Believe me when I say that a) the ranking system is not final and is being worked on, and b) many things have been considered and are being considered. And most of the obvious fixes turn out not to be so useful and/or easy once you start to think about them. The ranking definitely does have flaws, and I'd say nobody has ever called it perfect, or even close to it. It's not. It will most likely never be, not with a game with so many random factors such as Starcraft 2. It is, however, still pretty damn accurate, compared to all the other ranking systems out there. Just try it: Go to the prediction page and use aligulac's predictions for your liquidbets. You will do significantly better than the average. | ||
|
Grovbolle
Denmark3811 Posts
On April 19 2013 02:57 Canucklehead wrote: Why not? That is a pretty big flaw in the system if it produces a ranking that has forgg in the top 5. You guys are pretty sensitive and not open to any criticism I see. A person who took their system more seriously would try to figure out why forgg got ranked so high and try to fix the flaws instead of pretending no flaws exist. As already posted, we are all aware of exactly why forgg is ranked as highly as he is. We also explained why. We are also debating how to improve this internally, but currently haven't come up with a solution. The best current solution is to wait for him to lose points (which is inevitable). This system uses likelihood optimization, which basically means that if a player gets a better score than he was expected to, we raise the rating of the player so that the overall likelihood prediction of the entire model fits the data set available. With ForGG plowing players, 2-0'ing lots of players he was only predicted to beat, let's say 2-1, his rating improved. And because he played quite a few players, it increased a lot. You quite simply bash something you do not understand without adding anything to the discussion, not even having the courtesy to read the thread since I clearly adressed this on page 3. | ||
|
SniXSniPe
United States1938 Posts
Who is that #1 Terran? Is that Lucifron, or something? That person I think finished #1 last WoL season too. That's a ridiculous amount of games with even a higher win-rate than ForGG. | ||
|
Mortal
2943 Posts
Also, Protoss having a shitter of a time? ;_; | ||
|
a3den
704 Posts
Just like no rating system has currently the KeSPA players at their respected perceived strength (imo), there needs to be more point exchange (matches) between the different scenes (esf, kespa, international) for everything to balance in a more meaningful way. | ||
|
Depom
Sweden3 Posts
It would be interesting to know if those posters think it's possible to improve on the model (and if so, how) or if it's just generic internet-based bashing of things you don't agree with. Since both the code and the data is available for download, anyone with the required skill (or funds to pay someone with the required skill) could alter the model in any way they choose. You could factor in age, shoe size and twitter followers if you wanted. Side note; as someone who loves starcraft and is currently studying statistics and python in my spare time, this project is a god-send. Thank you and keep up the good work. | ||
|
Blargh
United States2103 Posts
| ||
| ||
