Massman

Current system: Top-5 hardware masters = top-6 team Top-10 hardware masters = 44k team (= 12k more than current #1 team) Estimations on how much people you need to be the #1 team in a rev4 design are very likely to be incorrect. For a good estimation, we need to see a beta version in action.

I also said before that things NEED to change. I've said it from the beginning, a couple of times in the other rev4 thread and said it several times in this thread as well. Things NEED to change. Period.

These bonus points are not introduced with the eye on 'balancing'. It's introduced because rev3 awards skill based on a difficulty-scale 1) Rev3 weighs results based on amount of competition. Being 1st amongst 5 is worth less than being 1st amongst 50. Whether that's in the 3x CPU global category or a random hardware category 2) Running a 5-pot configuration is significantly more difficult than running a 1-pot configuration. The little amount of points awarded for this (due to Rev3 design) was compensated by the introduction of these bonus points. NOT because it's the world record, BUT because multi-GPU LN2 configurations are inherently more difficult to handle. The 'get some slaves'-argument makes no sense. I can use the exact same argument to make a single CPU pot configuration less difficult ... "get someone to control the temps for your CPU so you can focus on the OS".

Wrong. All his golden cups will be worth more to the team than before. All the scores where he's got the best score of the team in that specific ranking will be worth more to the team. Wrong. See above. Double wrong. Being good individually will get you somewhere (pretty much like Turrican would be a top-35 team on his own in the current system), but it will not give you a free pass to the top of the rankings. Also, having support (and therefore golden samples) will be equally beneficial in terms of adding points to your team, but less beneficial in terms of taking away points from the other teams. In addition, two people on the same team with great connections (2x golden sample) have a lot less effect in the Rev4 as having the top-2 spots in the user ranking doesn't mean anything for the spots in the PowerTeam ranking. Financially less-capable overclockers still have an effect on the team ranking in two ways: increasing their own points (eg: bench old hardware for hardware points) and putting the team in those old-skool rankings in 1st/2nd/3rd in the PowerTeam ranking. Partially wrong. One man team will not 'dominate the team rankings indefinitly'. A team with a series of very-capable, well-motivated overclockers will do just fine. And what is wrong with that? "very capable", "well motivated" ... isn't that a good thing? Ironic. When I first published the plans, people almost killed me under the assumption that all my intentions were marketing-driven. Now that ONE(!) single aspect of the plans is changed, I'm being accused of not being marketing-driven enough. What a strange world we live in.

Just a random fact I'd like to throw at you. HWBOT currently counts: - 46 Global rankings - more than 10864 HW CPU rankings (didn't count the multi-CPU rankings) - more than 7260 HW GPU rankings (didn't count the 3/4-VGA rankings) People also always want more rankings where they can get point for. Introducing a simple new benchmark for the CPU (not split up per core) would currently increase the amount of HW CPU rankings with 1552 (amount of CPU models in our database) Are you now going to tell me that with a 6-man team you can 'win' the team rankings? Or that there's no way for newcomers to help the team by having the team's #1 score in some ranking? Numbers argue differently. I have no idea how they will use the bone I threw them in the LOC-incident. Basically, they didn't use the information before (at all) and only after given incentive to have a look at some scores the info they had came in handy. I'm unsure if there are plans to continue the research on score/system sharing. In any case, whatever Futuremark will do, it doesn't help us. FM has no information on the specific VGA/CPU identity and can, in other words, not differentiate between two CPU or VGA samples. Or, to make it even more clearer: they can not see the difference between your E8600 and my E8600.

Yes. I think the correct term for this is 'compromise'. Also, consider the following: Any system that is very beneficial for newcomers is also very beneficial for hardware sharing. Any system that is very beneficial for newcomers is also very beneficial for large-sized teams, or the merging of several large teams. The latter has happened before and many people (not always in public) critisized this. (url = http://hwbot.org/forum/showthread.php?t=1118) The whole point of changing the current system is to make it less prone to pure quantity and more prone to pure quality. Both in terms of skill, effort as hardware.

The key concept, the PowerTeam ranking, is missing in your example. That's the biggest difference between Alternative 5 and alternative 3. Alternative 5: "Team ranking = SUM(points best submissions per ranking) + [sUM(all non-best submissions) / 10]" Look at the example you worked out. Then look again at above formula. Alternative 3 would be (following your example scores + assuming team/user ratio of 1/2 (which basically means team_weight/user_weight equals 1/2)): PowerTeam contributes: 49,8 / 2 = 24,9p Users contribute: (49,8 + 38,7 + 33,3 + 29,5 + 26,9) / 10 = 17,82 Thanks to math-magic, you can also look at things like this: PowerTeam contributes: 10x (49,8 / 2) = 249p Users contribute: (49,8 + 38,7 + 33,3 + 29,5 + 26,9) = 178,2 PS: there's an aspect of irony here. Although you're saying you are looking at things from a team-sprit point of view, you're still thinking too much in terms of user points. The PowerTeam points are points awarded to the team for being better in comparison to other teams.

K404, What you are describing is Alternative 5, not 3 .

Also, we're not giving users only 10% of their points. Personal totals will have 100% of the points; the team contribution algorithm just only uses 10% of that personal total. In practical terms: if you now get 25p for a submission, the new teams league will NOT change that. If someone beats you, you'll have still the same amount in decrease as you have now (like 24,8p or so). Please make a note that that 10% is just relative weight. As mentioned here, the formula can easily be adjusted so there's no decrease in contribution from a personal point of view. The question is whether that makes things better or worse: do you want a team ranking with the top team at 5k points, 50k points or 500k points. The idea is to have the new lay-out in this style: Core i7 980X INFO -- USER RANKING -- (POWER)TEAM RANKING -- ... -- In other words, the PowerTeam ranking will be on a separate tab. The scores listed on that page will ONLY contain the team name and NOT the user name. Clicking on the result will direct to the score detail page (which shows the user name).

That's the PowerTeam variable: the best score of the team is used to determine the PowerTeam ranking of your team within a hw/bm ranking. If I understand you correctly, you are angry with the idea that someone from your team, who pushed really hard to get the team higher up the PowerTeam ranking, is contributing more to the team in this specific hw/bm ranking? How do you line that up with "all we fight for is the team spirit"? As far as I can see, option 3 would bring forum replies such as:

Lol. In option 3 ... they are not useless. That's the big difference between what was suggested in the very first document: first design rendered newcomers useless for the team ranking, option 3 design embraces newcomers and make them useful. Option 3: Team points = (10x PowerTeam points) + SUM(all member points)

4 is indeed too difficult both in terms of understanding and coding-wise. Shouldn't be there, but is there anyways. Option 5 and option 3 give, in practical terms, pretty much the same results, but option 5 is more prone to hardware sharing as displayed here: http://hwbot.org/forum/showpost.php?p=73945&postcount=11 Many people seem to have issues with the 'best score per ranking'-concept. Maybe I need to describe it differently: in essence, we would have a separate tab for user and team results. Whereas the current system allows users, through hardware sharing, to take away points from your team, a system with the PowerTeam points still gives the team what it deserves based on the team ranking. In practical terms: even if there are 25 overclockers of the same team in front of you, if you are able to make your team second in the PowerTeam ranking, your team will receive a fair amount of points. Instead of getting points for being 26th, the team will get points for being 2nd. Option 3 is a combination of Powerteam and user points, so in essence you are still contributing to the team total with every single submission, but thanks to the PowerTeam variable there's a buffer for hardware sharing (and hardware spamming). This also means that the team will benefit more from you pushing to beat another team. I wasn't even planning on adding that last option for the simple reason that change IS necessary. I think no-one can deny that over the last couple of months, hardware sharing (and score sharing) has been the most discussed issue (along with corporate overclocking). It's not only annoying for the community, as points are taken away from the team, but also for the HWBOT staff, as it's very difficult to prove hardware sharing. Therefore, a team ranking system change is necessary to reduce the effect of potential hardware sharing. In the end, I added the option so people could vote for both the alternative that looks best and still ventilate the fact that they don't like change.

Have you ever tried handling 5 pots all by yourself? . It definitly doesn't look as difficult as it really is. The key is here amount of time per pot: with only a CPU pot, you can spend 100% of the time on this single pot, but if you have 1 CPU and 4 GPU pots, you have to divide the time in 5x 20% of the time for each pot.

Never noticed how clean the setup looks. Well done mate!

3/4-way GPU on LN2 requires a lot more skill than running WPrime on a 4x 6-core system ...

Not if we do the 10x at the very end of the calculation. What I wanted to show with that 10x calculation is that a newly proposed algorithm can be made 'better' by artificially increasing the points. It's not because everyone gains points that the system is in fact better, but it's perceived as better. In fact, the relative standing of the teams remains exactly the same. In the rankings where there are very few results, the PowerTeam points will be lower than in competitive rankings, just like it is now in the user rankings. I'm pretty sure we'll have to work with a min_weight variable to make sure the PowerTeam points aren't totally meaningless (eg: always 2p if your team is first), but that's something we need to work out once we get an entire calculation done.

(CLICK for LARGER VERSION) Full version

At least I try, right? . FYI, people voting for Alternative 2 are basically saying that currently there's no problem at all. Given the latest complain-threads and -articles, I find that very strange.

Getting the information is not the problem, it's the hwboint algorithm that is. I need to calculate the weight and individual points for each record as the algorithm does not support team rankings. Anyways, I have an approx ranking now ... the question is whether I should make it look better or not. A simple trick with numbers (multiplying the end results x10) makes ALL the teams GAIN points instead of lose points in comparison to the current system. Some teams just gain more points than the others. Algorithm as in Alternative 3: [Team points = "PowerTeam points" + SUM(all points) / 10] is equal to [Team points = 10x "powerteam points" + sum(all points)] Ah, I'll just add them both so you can see it with your own eyes

'Feed' = manual calculations in excel, in this case I'll give it a shot.

In comparison to the current situation? And in comparison to some other alternatives?

Possibly because some services are enabled, explorer scans for all active applications, 'detects' the wprime and puts it in the taskbar?

Not entirely. The effect of hardware sharing is more prominent in Alternative 5 than in Alternative 3, as it affects the top score of the team. Maybe an example to show what I mean: User Ranking: 1. OCA 1 - 50p 2. OCA 2 - 40p 3. OCA 3 - 30p 4. OCX 1 - 20p 5. OCX 2 - 10p 6. OCX 3 - 5p 7. HOT 1 - 3p 8. HOT 2 - 2p 9. HOT 3 - 1p Assuming team/user ratio for this ranking of 1/2, PowerTeam ranking: 1. OCA - 25p 2. OCX - 20p 3. HOT - 15p Alternative 5: - OCA = 50 + (40 + 30)/10 = 57p - OCX = 20 + (10 + 5)/10 = 21,5p - HOT = 3 + (2 + 1)/10 = 3,3p Alternative 3: - OCA = 25 + (50 + 40 + 30)/10 = 37p - OCX = 20 + (20 + 10 + 5)/10 = 23,5p - HOT = 15 + (3 + 2 + 1)/10 = 15,6p Fighting for a better ranking for the team will be MORE BENEFICIAL with alternative 3 than alternative 5. In addition, hardware sharing will be LESS of an issue in alternative 5. Beating the best score of another team is a lot more beneficial in alternative 3 ... isn't that good for the team spirit? //EDIT: First case study - hardware sharing within top team// Just as example, this is what happens when a OCA 4 comes in 4th in the ranking: Alternative 5: - OCA = 50 + (40 + 30 + 20)/10 = 59p (before: 57) - OCX = 10 + (5 + 3)/10 = 10,8p (before: 21,5) - HOT = 2 + (1 + 0)/10 = 2,1p (before: 3,3) Alternative 3: - OCA = 25 + (50 + 40 + 30 + 20)/10 = 39p (before: 37) - OCX = 20 + (10 + 5 + 3)/10 = 21,8p (before: 23,5) - HOT = 15 + (2 + 1)/10 = 15,3p (before: 15,6) //EDIT2: Second case study - someone beating other overclockers and team// Maybe a more relevant situation: what happens when HOT3 jumps from 9th to 3rd (and therefore HOT jumps over OCX to 2nd place in PowerTeam ranking): Alternative 5: - OCA = 50 + (40 + 20)/10 = 56p (before: 57) - HOT = 30 + (2 + 1)/10 = 30,3p (before: 3,3) - OCX = 10 + (5 + 3)/10 = 10,8p (before: 21,5) Alternative 3: - OCA = 25 + (50 + 40 + 20)/10 = 36p (before: 37) - HOT = 20 + (30 + 2 + 1)/10 = 23,3p (before: 15,6) - OCX = 15 + (10 + 5 + 3)/10 = 16,8p (before: 23,5) In alternative 3, beating any given overclocker doesn't change that much to the ranking, but improving your team's PowerTeam ranking is highly rewarding. In alternative 5, the entire team total goes down quite rapidly if only ONE other person beats some scores.

To those voting for Alternative 5: could you please explain why you don't like a specific teams league (as in Alternative 3)?

Damned if I do, Damned if I don't