Reconsider How Ties Receive Points

[xxbassplayerxx]

Right now, a tie on the bot ends up with the user who submitted first taking the higher position. Because this hobby is one that is largely based on competing with old hardware and time is not a very relative metric, I suggest the following changes.

 

Take this example (me!!) for consideration:

 

 

 

Mothod 1:

For a tie, the points normally given to the users would be added together and divided by the number of users (an average). In the rankings, all users would show as having the same rank.

 

For example:

 

(39.8+33.5+29.7)/3 = 34.3 points each

 

1. 4sec 437ms 49.9pts

2. 4sec 453ms 34.3pts

2. 4sec 453ms 34.3pts

2. 4sec 453ms 34.3pts

5. 4sec 464ms 27.2pts

 

 

Method 2:

The other way to divvy points would be to give them all whatever 2nd place normally receives, in this case 39.8 points. The 5th place user would still receive the points that 5th place normally awards.

 

For example:

 

1. 4sec 437ms 49.9pts

2. 4sec 453ms 39.8pts

2. 4sec 453ms 39.8pts

2. 4sec 453ms 39.8pts

5. 4sec 464ms 27.2pts

 

 

Note that all users receive 2nd place as well as the same amount of points (as is fair with a tie). Each would also receive a silver cup, in this instance.

 

I prefer Method 2. This way, no one loses points (as would happen with some people in Method 1) and other people tying would not cause the points value to drop.

 

What are your thoughts and opinions?

Edited March 1, 2013 by xxbassplayerxx

[GENiEBEN]

Im for Method 2 or a Method 3 where ties are sorted by cpu frequency efficiency (lower mhz > better place).

[Massman]

We can't do efficiency, people will clock down for screenshot purposes and lie about the used frequency. We can't enforce, so we shouldn't make a rule that encourages lying.

 

I like both ideas, but I'm also not really against the current "whoever submits first is ranked higher"-principle (although I guess that's more relevant to competitions rather than a continuous ranking). Whichever is the least amount of effort to code, I'd opt for. Dennis, what do you think?

[Massman]

By the way, I suppose this thread could also be in the public forums? I think it would be interesting to hear more than just staff opinions about this.

[GENiEBEN]

Let's see Dennis' response.

[knopflerbruce]

I don't like method 3 either, efficiency is good, but points are based on scores, not tweaking level. Maybe the same user didn't have the skills to push the last few mhz out of the chip, too - by that reasoning we could reward the max points to the score with the HIGHEST frequency.

 

Method 2 is the usual method in sports, i'd go with that if possible.

[der8auer]

If there was a poll I'd vote for method 2

 

Current system is unbalanced at this point as all deserve the same amount of points for equal results.

[Christian Ney]

method 2 ftw

[Gamer]

I would keep it like it is, happend to me a few times.

New drivers, tweaks,....

 

If it would change, method 2.

[knopflerbruce]

Though, IF there are drawbacks, like increased server load by a fair margin, or it takes too much coding time, the current situation is not a disaster, by all means.

[xxbassplayerxx]

Thanks for the responses everyone. For some reason I couldn't reply to this after I posted it but Dennis got it all sorted!

 

I agree that it wouldn't be too bad if the change could not be implemented, but it would definitely be appreciated. In my case, for instance, there's a difference of 10 points and we all have the same score/time!

 

By the way, I suppose this thread could also be in the public forums? I think it would be interesting to hear more than just staff opinions about this.

 

If any of you guys with mod permissions would move this somewhere public, I'd appreciate it!

Edited March 1, 2013 by xxbassplayerxx

[Devroush]

We've been working on a new scoring algorithm, main reason is that if you've got some obscure hardware (like 60-core cpu or something), you can easily get 60 UGP because there's no competition. This is likely to change in the near future so that there's no longer a minimum of 60UGP given out for first place.

 

As for your question: it's a valid point and I don't think it would be hard to implement. I'll probably test it out and see if there are... unforseen consequences .

[xxbassplayerxx]

Great! Thanks for the input

[Massman]

Moving to public area, Rev5 section. Also adding a POLL.

 

Fyi, the poll outcome will be used in case we find no major difficulties regarding the implementation of any of the suggested methods. This is an update with very low priority as it's not a bug and not a really useful feature. We're not going to spend hours and hours of development time on. If the implementation of the prefered method is a two-second job, we will follow the outcome of the poll.

[Mr.Scott]

A tie doesn't beat anything. Why should the first guy in lose points? Example: Guy #1 has a #1 score for 10 years. It takes 10 years for somebody else to TIE said score, not beat it, just tie it, and the 10 year vet LOSES points. That's BS.

Voted 'leave it alone'.

Edited March 3, 2013 by Mr.Scott

[GENiEBEN]

A tie doesn't beat anything. Why should the first guy in lose points? Example: Guy #1 has a #1 score for 10 years. It takes 10 years for somebody else to TIE said score, not beat it, just tie it, and the 10 year vet LOSES points. That's BS.

Voted 'leave it alone'.

 

How is he loosing points with method 2?

[Mr.Scott]

How is he loosing points with method 2?

With method 2 the #1 score would receive the same points as the later tied score. No points lost but no benefit for being #1 for an extended length of time either if the recent tie gets the same points.

 

Bottom line is, you want to reward for ties. There was nothing beaten to warrant more points. Just my opinion. You asked, I gave.

[GENiEBEN]

You are entitled to your opinion and it's appreciated, but I hope you know time does not equal 'skill'. We shouldn't be benching old gen hardware and let the old folks who ran them at their time get all the credit

Edited March 3, 2013 by GENiEBEN

[Mr.Scott]

You are entitled to your opinion and it's appreciated, but I hope you know time does not equal 'skill'. We shouldn't be benching old gen hardware and let the old folks who run them at their time get all the credit

Oh yes, lets reward the entitlement generation for not getting the job done yet again.

I agree, time does not equal skill, but with all the advancements in OC'ing and sub ambient technology, these scores should be beaten more readily if the bencher so chooses. One just needs to apply themselves, the hardware is still out there, you see it in almost every team comp. All you have to do is beat the score above you and you'll be rewarded. It's that simple.

[GENiEBEN]

but with all the advancements in OC'ing and sub ambient technology.

 

Other than improving pots to ease the work of the bencher there's hardly any advancement in that department, it's still down to CB/CBB luck. There's been no new magical cooling solution, cpus run on the same old boards and memory (with few exceptions ofc) and used OS` are the same as 7 yrs ago.

 

For the example provided here by Thomas, which current advancement in technology would you think it applies? :celebration:

[dinos22]

No2

[Mr.Scott]

 

For the example provided here by Thomas, which current advancement in technology would you think it applies? :celebration:

Bin a few more chips and try a little harder. Efficiency can always be improved. No such thing as perfect.

[Antinomy]

Keep it the say it is: rank by date (oldest = highest points)

Is it really this way? I remember being able to get the gold cup by having a tie with the #1 while my score was newer than the existing.

 

If all that 10 years of development in methods, hints, cooling and software has given you is a tie - you really suck compared to the guy who was able to clock it back in those days. Waza, maxmem weren't familiar to most people back there.

 

I totally agree with Mr.Scott. Many people should be glad Holicho didn't apply HWBot

[Devroush]

That was probably a bug then, oldest scores definitely rank first at the moment. xxbassplayerxx's example: http://www.hwbot.org/benchmark/wprime_-_32m/rankings?hardwareTypeId=processor_1741#start=0#interval=20

[Rasparthe]

A tie doesn't beat anything. Why should the first guy in lose points? Example: Guy #1 has a #1 score for 10 years. It takes 10 years for somebody else to TIE said score, not beat it, just tie it, and the 10 year vet LOSES points. That's BS.

Voted 'leave it alone'.

 

I can kind of see where your coming from, a 10 year score that stands should be recognized, but what a month, what about a day, what about an hour? Should a guy that submitted a score an hour earlier than someone deserve the extra 10 points simply because he didn't go grab that dinner the way the second guy did.

 

The reason I do not agree with your argument is scores get beaten all the time and you lose points because your score wasn't good enough. Same should apply for ties, if the first guy had applied himself better, binned more chips, pushed harder there wouldn't have been a tie. Now there is, and everyone should be awarded the same points. The same argument you apply to the second guy "stealing" points should also be applied to the first guy.

 

I voted for 2.