DUST Search - Dust Reports [concerns about a ELO Ranking report]

Parson Atreides wrote:

I think it's a cool idea, but your main paragraph about problems might be too important/pervasive to ignore, especially if your goal is to have an accurate measurement and to avoid trolling/epeen stroking. We all know once those numbers hit, people are going to look at the top and bottom and post accordingly.

I know from personal experience if you fight a few hard-fought battles and lose more clones than you can replenish in a day, then it's possible to lose a district despite winning essentially every battle. Or you can win a few, lose one, and then have so few clones left on a district it becomes basically impossible to defend. Both situations will lower your ELO and raise theirs despite the fact that you beat the opponent more often than you lost to them.

Another problem I see is how would you know who was attacking a district if they never take it? For example if you see Corp A under attack constantly, but they never lose the district, wouldn't it be impossible to calculate an ELO? Maybe Corp B, who is really good, is hitting them and they're successfully defending, which would increase Corp A's ELO a lot. On the other hand it could be their alt corp attacking and they're just practicing. How could you calculate the adjustments to their ELO?

I like the idea a lot and it would be nice to see implemented, but unfortunately I don't think you have sufficient access to information to make this accurate enough to be of any real worth.

Pages: 1 :: [one page]

Author	Thread Statistics \| Show CCP posts - 0 post(s)
The Info Broker Expert Intervention Caldari State 365	Posted - 2013.07.06 22:24:00 - [1] - Quote As I've been working on a couple of new reports I've been getting increasingly worried that they might in the wrong context detract from what I am trying to do with them. I'm hoping to explain the goal of the new report I am working on, and see if you guys thinks it's worth me continuing work on it, if I should modify my process, or if I should scrap altogether to work on other things. My Goal Creating a PC ELO ranking for every Dust Corporation that has been involved in PC. (If you don't know what an ELO is, it is basically a number that represents your relative level of success in PC by looking at who you have won and lost against, and factoring in the success [score] of the opponents you've played in each win and loss. Each week I would update the Win/Loss of a corporation based on that weeks combat, and recalculate the ELO to show who is moving up, and who is moving down in the Dust rankings. Primary Issues with the new Report The biggest issue with this report is that I don't have a full data set. The only signs I see as a third party observer to most PC matches is district flips. For this reason I am forced to make successfully flipping a district a win, and failing to defend a district a loss. This creates an inherent unfairness as groups who attack and then sell or give districts away will can't suffer losses. It also makes politics, and ringing just as important as combat to determining your PC ELO. While I agree these are all serious problems there is no way I can see to get around them, and I think they can be overlooked as long as you are not trying to draw to much additional meaning out of the ranking. The rankings are here to show who has had the most success in PC, it's not meant to suggest a higher ranked group will win in a 16v16 match. Things like Ringers or Politics are extremely important in PC, and I don't really mind them being extra factors in the process of calculating a PC ELO. At the end of the day all my information and calculations will be 100% public, and if you disagree with a ranking you can take a district from them and change it. The Process of Ranking Corporations I've assigned a value of 1500 to every single Corporation who entered PC. Nobody got any adjustments due to expected performance, or reputation. Using a K-Value of 30 I've ran a calculation for every single district flip that has happened in PC so far. The group that took the district gained ELO, the group that lost it lost an equal amount. The amount they gained or lost in every single calculation was based on the score of both parties involved. Some fights are only worth 3 points because of the ranking differences (Teamplayers defeating ZionTCD), while others have been worth 50+ points if they are serious upsets, say Zion beats Teamplayers. I've basically made this post to see where you guys land on this topic. Is this a meaningful statistic or report, or does it come off as just e-peen measuring or sensationalist reporting? I know other Dust sites have used 'shadow rankings' or just straight up editorialized rankings, only to come out looking foolish a few months later. I'm personally more interested in bringing content I can back up with numbers, even still I'd prefer to throw away a backed up report like this is even after it's near completion if it would hurt the overall thing we are trying to build, or be primarily be misinterpreted. If I do end up publishing it I guess I ask that it not be used for trolling or talking **** if possible. Yes some groups had extensive losses and no wins, it's a part of the report because it includes all of PC. I personally respect any group that got involved in PC, even if things didn't go well. I'm asking that if I put a lot of time into making this report that you have some respect for what I did and don't use it to troll to much. Thanks!
Parson Atreides Ahrendee Mercenaries Omega Commission 419	Posted - 2013.07.07 11:31:00 - [2] - Quote I think it's a cool idea, but your main paragraph about problems might be too important/pervasive to ignore, especially if your goal is to have an accurate measurement and to avoid trolling/epeen stroking. We all know once those numbers hit, people are going to look at the top and bottom and post accordingly. I know from personal experience if you fight a few hard-fought battles and lose more clones than you can replenish in a day, then it's possible to lose a district despite winning essentially every battle. Or you can win a few, lose one, and then have so few clones left on a district it becomes basically impossible to defend. Both situations will lower your ELO and raise theirs despite the fact that you beat the opponent more often than you lost to them. Another problem I see is how would you know who was attacking a district if they never take it? For example if you see Corp A under attack constantly, but they never lose the district, wouldn't it be impossible to calculate an ELO? Maybe Corp B, who is really good, is hitting them and they're successfully defending, which would increase Corp A's ELO a lot. On the other hand it could be their alt corp attacking and they're just practicing. How could you calculate the adjustments to their ELO? I like the idea a lot and it would be nice to see implemented, but unfortunately I don't think you have sufficient access to information to make this accurate enough to be of any real worth.
Severance Pay Krullefor Organization Minmatar Republic 550	Posted - 2013.07.07 13:48:00 - [3] - Quote What is ELO?
Thor Odinson42 Molon Labe. League of Infamy 555	Posted - 2013.07.07 17:45:00 - [4] - Quote Severance Pay wrote: What is ELO? It's explained in the post, Why'd you leave HS?
Rogatien Merc Ill Omens EoN. 296	Posted - 2013.07.07 20:07:00 - [5] - Quote Parson Atreides wrote: I think it's a cool idea, but your main paragraph about problems might be too important/pervasive to ignore, especially if your goal is to have an accurate measurement and to avoid trolling/epeen stroking. We all know once those numbers hit, people are going to look at the top and bottom and post accordingly. I know from personal experience if you fight a few hard-fought battles and lose more clones than you can replenish in a day, then it's possible to lose a district despite winning essentially every battle. Or you can win a few, lose one, and then have so few clones left on a district it becomes basically impossible to defend. Both situations will lower your ELO and raise theirs despite the fact that you beat the opponent more often than you lost to them. Another problem I see is how would you know who was attacking a district if they never take it? For example if you see Corp A under attack constantly, but they never lose the district, wouldn't it be impossible to calculate an ELO? Maybe Corp B, who is really good, is hitting them and they're successfully defending, which would increase Corp A's ELO a lot. On the other hand it could be their alt corp attacking and they're just practicing. How could you calculate the adjustments to their ELO? I like the idea a lot and it would be nice to see implemented, but unfortunately I don't think you have sufficient access to information to make this accurate enough to be of any real worth. As stated, this is about who has been successful in PC; it's about the war, not the individual battles - something I find harder and harder to explain to people these days. The one issue I see in this ranking is that it is an offensive ranking and will eventually favor new, aggressive corporations. Incomplete data tracking mechanics (FU CCP) mean it is impossible for a report like this to track successful defenses or failed offenses. Which is hilarious because the entire PC community is currently against being aggressive in PC. Ya'll are so weird haha. THAT SAID, I for one like it and will be interested in seeing how it plays out.
CHICAGOCUBS4EVER TeamPlayers EoN. 663	Posted - 2013.07.07 21:46:00 - [6] - Quote Honestly, ranking corporations in PC will prove extremely difficult. You have zero data on matches, who is in each or as you said districts that get sold/traded/change hands by intention rather than attrition. Dust does not give us the tools to make this kind of report mildly accurate at this time, which can create a bias from different points of view and detract from the service of dustreports. You really are limited to what you can prove, not assume and I don't want to see any reason for player x to frown upon the great service dust reports has done for the community Just my 2 isk :P
Zatara Rought TeamPlayers EoN. 456	Posted - 2013.07.08 03:41:00 - [7] - Quote Welp I can tell you I've played damn near 150 pc matches in the last month (If you don't see me on the video you didn't scroll down far enough), and I'm salivating at the prospect of a new war with all this blue donut mess. Someone attack us yo. We need to keep the knives sharpened before the tourney.


Pages: 1 :: [one page]