Massman Posted October 3, 2015 Posted October 3, 2015 Running the numbers ... Excel is going mad Quote
Bones Posted October 3, 2015 Posted October 3, 2015 One thing I haven't seen mentioned yet is the fact there are a few here with an insane amount of points vs others to the point if and when they should decide to swap teams it skews the team rankings for a bit. The top 3 teams haven't had much activity lately related to swapping places but all three are within 2000 points of each other. We at CP have been bumped by OCF a few times since aquiring the 4th place spot so there's competition in that as well with OCN nipping at OCF's heels. This means at least the top 6 spots are indeed being contested, it's just not so noticeable for now with the points spread between 3rd and 4th place and with all three top teams just going at it and keeping pace with each other..... Or maybe just not doing alot ATM but I doubt it. There is alot of competition going on, I don't see a need to once again change things around for now - Let it ride because I'm sure once the Country Cup starts we'll see some competition. Quote
Massman Posted October 3, 2015 Posted October 3, 2015 First things first, let's make sure we're all on the same page. The current Teams League is based on the following algorithm: SUM[GTPP + HTPP + (UP)/10], with GTPP = Global Team Power Points = "best score per team per global ranking"; eg. 3DMark03 1xGPU HTPP = Hardware Team Power Points = "best score per team per hardware ranking"; eg. 3DMark03 1xGPU GeForce GTX 480 UP = Sum of all team member user points, without limit on GL, HW, or WR The suggestion of Der8auer affects the GTPP only. His suggestion is to rank the teams in the global rankings differently. Instead of taking into account just the best score of the team, he suggests to take into account the best 10 scores of the team. The caveat: one hardware can only be used once (so no spamming of 6700K) and one user can account for multiple results (so no hardware sharing). Der8auer asked me to make a couple of examples to show the effect. This calculation is very different from the existing one, so I'm running the numbers manually at the moment. 3DMark11 Performance 1xGPU: Current [table=head;autonum=1]score|team|gpu 35852|Australia OC|GeForce GTX 980 Ti 35011|TecLab OC Team|GeForce GTX 980 Ti 34718|Team MLG|GeForce GTX 980 Ti 34616|Hardwareluxx.de|GeForce GTX 980 Ti 34330|PURE|GeForce GTX 980 Ti 32367|Extreme Box OC TEAM|GeForce GTX 980 Ti 31939|Elite Overclockers of Korea (republic)|GeForce GTX 980 Ti 31578|Team KATANA|GeForce GTX 980 Ti 30464|Asus ROG|GeForce GTX 980 Ti 30280|Madshrimps Belgium OC Team|GeForce GTX 980 Ti 30278|Klan-OC|GeForce GTX 980 Ti 29517|OcTeamDenmark|GeForce GTX Titan X 29257|lab501.ro|GeForce GTX Titan X 29020|OverCleX|GeForce GTX 980 Ti 28809|Team South Africa|GeForce GTX 980 [/table] 3DMark11 Performance 1xGPU: Der8auer method (10 best) [table=head;autonum=1]score|team 24757,9|Team MLG 23964,3|TecLab OC Team 23267|Australia OC 22825|PURE 22548,1|Hardwareluxx.de 22413,6|Extreme Box OC TEAM 21519,9|Klan-OC 21164,3|Overclockers.com 21033,2|Overclock.net 20314,5|HW Legend OC 19889,3|XtremeSystems 19849,2|Team Russia 19571,4|Asus ROG 19500|HwBox Hellas O/C Team 19316,8|ForumOverclocking.ro [/table] 3DMark Performance 1xGPU: Der8auer method (20 best) [table=head;autonum=1]score|team 18952,2|Team MLG 17726,05|Australia OC 17662,7|TecLab OC Team 16867,4|Hardwareluxx.de 16412,25|HW Legend OC 16412,1|Overclockers.com 16378,85|Overclock.net 16373,25|PURE 16300,7|Klan-OC 15885,4|Team Russia 15771,8|HwBox Hellas O/C Team 15529,5|PC Games Hardware 15167,4|XtremeOverdrive OC team Italy 15097,35|XtremeSystems 15056,4|lab501.ro [/table] Comparison hardware used: 5th place Current - Pure: GeForce GTX 980 Ti (Xtreme Addict) Der8auer (10) - Hardwareluxx.de: GeForce GTX 980 Ti (Dancop), GeForce GTX Titan X (Dancop), GeForce GTX 980 (Dancop), Radeon R9 290X (Dancop), GeForce GTX 780 Ti (Dancop), GeForce GTX Titan (Dancop), GeForce GTX 780 (Dancop), Radeon HD 7970 (BenchBros), GeForce GTX Titan Black Edition (paralyzer2005), GeForce GTX 680 (Freakezoit). Der8auer (20) - HW Legend OC: GeForce GTX 980 (rsannino), GeForce GTX 980 Ti (delly), GeForce GTX 780 Ti (rsannino), Radeon R9 290X (NATA 58), GeForce GTX 780 (rsannino), GeForce GTX 970 (delly), Radeon HD 7970 (rsannino), GeForce GTX Titan Black Edition (NATA 58), GeForce GTX Titan (delly), Radeon R9 280X (NATA 58), Radeon R9 280 (delly), Radeon HD 7950 (delly), GeForce GTX 960 (delly), Radeon R9 285 (delly), GeForce GTX 680 (delly), GeForce GTX 580 (rsannino), GeForce GTX 770 (NATA 58), Radeon R9 270X (NATA 58), GeForce GTX 670 (Slime), GeForce GTX 950 (isonzo) First impression By limiting to different hardware, the GTPP will be skewed towards the high-end of the latest generation. Especially in product series where there are multiple revisions of the top performance hardware, this will be an issue. For example: GTX 980, 980 Ti, Titan X. These are three different models, but all from the same product category. This effect should apply to all benchmarks. What could be another interesting approach is to go by unique core architectures (Skylake, Haswell, Ivy Bridge) or product families (GeForce 900, GeForce 700, GeForce 600). Quote
JunkDogg Posted October 3, 2015 Posted October 3, 2015 Well that makes a bit more sense. Ye be warned, this will piss a lot of people off. Especially teams who drop. Quote
Mr.Scott Posted October 3, 2015 Posted October 3, 2015 By limiting to different hardware, the GTPP will be skewed towards the high-end of the latest generation. This effect should apply to all benchmarks. I knew it. That's bunnyextraction. You effectively kill my team with that change. See post #21. Quote
ObscureParadox Posted October 3, 2015 Posted October 3, 2015 Actually dude i don't see it making one hell of a difference since the global points have always been about the top scores of the latest hardware. Permitting the hardware points side doesn't change you guys should be fine Quote
Mr.Scott Posted October 3, 2015 Posted October 3, 2015 Actually dude i don't see it making one hell of a difference since the global points have always been about the top scores of the latest hardware. Permitting the hardware points side doesn't change you guys should be fine Do you really think that the globals will be the only change? You're kidding yourself. HW will be hot on it's heels if not at the same time because somebody will b1tch. That is and has always been the MO. Give an inch, take a mile. Quote
Rasparthe Posted October 3, 2015 Posted October 3, 2015 Actually dude i don't see it making one hell of a difference since the global points have always been about the top scores of the latest hardware. Permitting the hardware points side doesn't change you guys should be fine Actually this puts smaller teams at a very large disadvantage. You can see the example in the charts that Massman has put up. When it comes to 3DMark11, OCN is completely uncompetitive, not even showing up on the board. Under the (10) system, OCN moves to 9th, and under the (20) system up to 7th. Under the current system, smaller teams can still be competitive, since no matter how large the competition is, they can only post one score. More members means they have more people trying to get that competitive score but they are limited to one contributing score. For a team like CP, at least in the global portion, it would a large blow. They have only 30 members, PC Games Hardware, has 2000 (granted not all active I'm sure). I can see that for the global portion of the team score it would be heavily biased towards, PC Games Hardware and OCN and other large teams. As you said though, the global portion is already incredibly skewed to the newest hardware. So I guess my question is, why does this new system attract more attention to Team Points? Why would guys running the newest hardware be more interested in team play when they are already receiving a "bonus" to their team scores because they are running the newest generation hardware they currently have? Quote
der8auer Posted October 3, 2015 Author Posted October 3, 2015 Actually that's not how I thought about the ranking haha Will talk to Massman again I guess. Quote
Mr.Scott Posted October 3, 2015 Posted October 3, 2015 For a team like CP, at least in the global portion, it would a large blow. They have only 30 members, PC Games Hardware, has 2000 (granted not all active I'm sure). I can see that for the global portion of the team score it would be heavily biased towards, PC Games Hardware and OCN and other large teams. Actually, we only have 16 members, 6 of which I consider active. That 30 figure has been wrong since the beginning of time.........because we don't fix bugs here. We just make changes to add more bugs that go unfixed. Quote
MetalRacer Posted October 3, 2015 Posted October 3, 2015 What could be another interesting approach is to go by unique core architectures (Skylake, Haswell, Ivy Bridge) or product families (GeForce 900, GeForce 700, GeForce 600). This approach would be better. Quote
Rasparthe Posted October 3, 2015 Posted October 3, 2015 Actually that's not how I thought about the ranking haha Will talk to Massman again I guess. Okay, I totally misunderstood your OP compared to how Massman set up his rankings, so perhaps I'm still completely confused on the idea. Just as my general opinion though, I honestly had no idea there was a problem with the Team Ranking. It is one thing that I thought was fairly well balanced. You have teams that concentrate on newest generation in the top (Team Russia), teams that concentrate on old school (CP) and teams that balance the two (PURE & OCN).(And yes, I know I'm generalizing a lot) To me, that says its a balanced scoring system. Lots of ways to help out your team's rankings, no matter the member you may join or where their interests may lay. If you are trying to garner interest in team rankings perhaps messing with the scoring system isn't the best route. There is a Team Cup, but the the top 3 teams barely even participate. The why of that, would be the first question I would want answered. Maybe there is just little interest in team play amongst those teams. I can tell you that at OCN there was a great deal of excitement around Team Cup, our Skype channel literally pukes out an ocean of posts every morning when I checked it, and now that its over its down to a dull roar. Just my opinion, of course, but I actually like the current system. Quote
der8auer Posted October 3, 2015 Author Posted October 3, 2015 Honestly I don't want to change anything if the majority is happy with the current team ranking. It's just brain storming guys. The ranking I'm talking about is pretty much what we see in country-cup or team-cup - just without time and specific hardware limit. It has nothing to do with the current way of a team ranking. Maybe call it Competitive-Benchmark-Team-Ranking. I think this also went wrong in the communication between me and massman - sorry for that Just an example with 2 Benchmarks so you can see how it would build up. So before you guys get the torches out to burn my house down - how would you think about this if it would be a seperate solution/ranking and the current ranking would still be present? Quote
Mr.Scott Posted October 3, 2015 Posted October 3, 2015 It was never on your shoulders bro, just want to clarify that. Nobody's gonna burn your house down. Quote
Rauf Posted October 3, 2015 Posted October 3, 2015 (edited) Will write my full opinions later. But with this ranking, how would you compensate for an average when the team does not have 10 or 20 different hw submissions? Also it does seem wrong to not make it about hw ranking instead of actual benchmark score. A top score with an old but golden cpu would be rather pointless in the global rankings. Look at msi beat the fastest comp for example. Team points could be the sum of individual hw points for top 10/20 team members submissions in each benchmark. Edited October 3, 2015 by Rauf Quote
Mr.Scott Posted October 3, 2015 Posted October 3, 2015 Will write my full opinions later. But with this ranking, how would you compensate for an average when the team does not have 10 or 20 different hw submissions? Also it does seem wrong to not make it about hw ranking instead of actual benchmark score. A top score with an old but golden cpu would be rather pointless in the global rankings. Look at msi beat the fastest comp for example. Team points could be the sum of individual hw points for top 10/20 team members submissions in each benchmark. And if your team does not have 10/20 members, what then? Quote
Massman Posted October 4, 2015 Posted October 4, 2015 Der8auer explained the difference in ideas over messenger. So originally, he was thinking about a competitive Teams League that solely looks at the GTPP and takes the average score of the best 10/20. We can only do something like this alongside the current format. The goal is to make it more appealing to be part of the team again, something which is working well for some teams, but for a lot of teams it's not. Der8auer made this suggestion in the private XOC Focus Group (link for those who have access)) and the response was a bit more positive. Thank you for coming up with ideas on how to improve the League, Der8auer! Moving forward, I think we need to have more exact information on the following items: What is the balance of GTPP and HTPP in the Teams League What is the weight of the single-man heavy-hitters on teams What would a GTPP ranking look like based on the average top 10, allowing one score per architecture or generation That's stuff for next week! Actually this puts smaller teams at a very large disadvantage. You can see the example in the charts that Massman has put up. When it comes to 3DMark11, OCN is completely uncompetitive, not even showing up on the board. Under the (10) system, OCN moves to 9th, and under the (20) system up to 7th. Under the current system, smaller teams can still be competitive, since no matter how large the competition is, they can only post one score. More members means they have more people trying to get that competitive score but they are limited to one contributing score. This analysis is correct, but not complete. Like you said, it would put smaller teams of which no member does high-end as a disadvantage. It would require larger teams or teams which have a couple of high-end benchers to be more active/competitive in the GTPP space. Thank you for the civil discussion and relevant contributions, Rasparthe. Quote
Mr.Scott Posted October 4, 2015 Posted October 4, 2015 (edited) Oh yes, lets handicap the small teams. That will promote more and better benchers. What are you thinking man. Maybe the big teams should think about culling their dead weight inactive benchers. Maybe there should be a time limit on inactivity before team points are stripped. Eg. after say a year of inactivity, team points from the inactive members are stripped from said teams. That would shake up the standings a little bit wouldn't it. Edited October 4, 2015 by Mr.Scott Quote
max1024 Posted October 4, 2015 Posted October 4, 2015 New system is not interesting at all. Our small Belarus OC Team got 5-th place on Team CUP 2015. Our team is very small, active overclockers is about 7 people, one man like old retro HW, one modern, and one mobile devices and so on. If I'm right undertand, new team rating will be calculated as middle value for one HW devided by all persons in team. But for example: people A have cpu A, people B do not have cpu A because it is not interesting to it, he prefer old HW or mobile HW. In result team ranking for A cpu will be result people A devided by all persons in team. Think it is illogically and do not right at all different people likes different HW, and one can it offer another is not. New ranking like middle temperature in clinic, one people with pneumonia is about +40 another is -20 in mortuary. But in generally they are both healthy and have +36.6C Quote
der8auer Posted October 4, 2015 Author Posted October 4, 2015 No you understand it wrong it's not related to the amount of team members Quote
Mr.Scott Posted October 4, 2015 Posted October 4, 2015 (edited) No you understand it wrong it's not related to the amount of team members But the amount of members on a team DOES have an impact. Like you said, it would put smaller teams of which no member does high-end as a disadvantage. It would require larger teams or teams which have a couple of high-end benchers to be more active/competitive in the GTPP space. Edited October 4, 2015 by Mr.Scott Quote
max1024 Posted October 4, 2015 Posted October 4, 2015 The main trouble in this is in averaging results. It is loses individuality of person and stand on the first place Team that is not a right. Underlying all must first be a specific person - overclocker, even if that one person is the Team. Quote
der8auer Posted October 4, 2015 Author Posted October 4, 2015 The main trouble in this is in averaging results. It is loses individuality of person and stand on the first place Team that is not a right. Underlying all must first be a specific person - overclocker, even if that one person is the Team. Well you have several steps in the ranking. Like in the Country- or Team-Cup it should be possible to click on the average score to see who is contributing with what hardware. But the amount of members on a team DOES have an impact. No it doesn't. It doesn't matter if you have 5, 50 or 500 users in a team. You need people to serve high-end, mid-end and legacy hardware. That can be done with 200 users or 2. It's about the right communication (team-work) to split up the work and find people for all required scores. Quote
TerraRaptor Posted October 4, 2015 Posted October 4, 2015 You need people to serve high-end, mid-end and legacy hardware. Is this a hobby or work? Why should one ask teammates buy and bench high-end if the team is for example built of students of one university? Are you sure those students are bad team?) Quote
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.