Total Posts:68|Showing Posts:1-30|Last Page
Jump to topic:

The leaderboard...

Xer
Posts: 7,776
Add as Friend
Challenge to a Debate
Send a Message
6/4/2009 3:40:14 PM
Posted: 7 years ago
The leaderboard/percentile should be reformed in my opinion. There should be more emphasis on win ratio instead of wins. For example, on the first page of the leaderboard, a person who is 65-47 is ahead of a person who is 48-3. This should not happen. That is it for now, thoughts?
Nik
Posts: 552
Add as Friend
Challenge to a Debate
Send a Message
6/4/2009 3:41:59 PM
Posted: 7 years ago
Yeah agreed, but theres not going to be any changes for a while, so get used to the current system cus its gonna be like this for a while.
"If you could tell the world but one truth, I could convince it of a thousand lies"
Maikuru
Posts: 9,112
Add as Friend
Challenge to a Debate
Send a Message
6/4/2009 4:53:55 PM
Posted: 7 years ago
At 6/4/2009 3:57:18 PM, mongeese wrote:
I think that the entire voting system needs reformation. The general masses fail all too often in choosing the victor.

Don't the general masses always pick the victor? Unless you're referring to multi-account voting, in which case I concur.
"You assume I wouldn't want to burn this whole place to the ground."
- lamerde

https://i.imgflip.com...
wjmelements
Posts: 8,206
Add as Friend
Challenge to a Debate
Send a Message
6/4/2009 4:56:45 PM
Posted: 7 years ago
But the general 7-bomb makes blind votes more valuable than educated ones...

I do think that the percentile system needs to be changed. InquireTruth, below me, is a much better debater than I.
in the blink of an eye you finally see the light
TheSkeptic
Posts: 1,362
Add as Friend
Challenge to a Debate
Send a Message
6/5/2009 12:36:31 AM
Posted: 7 years ago
Win ratio is too flimsy to base it on. If my record is 1-0, then I have a perfect 100% record. So basically, any newcomer can come onto this site, win in a ridiculously easy debate, and trump me.
I-am-a-panda
Posts: 15,380
Add as Friend
Challenge to a Debate
Send a Message
6/5/2009 3:03:13 AM
Posted: 7 years ago
I still think it' show much Phil loves you. Strangely, my percentile has gone up.
Pizza. I have enormous respect for Pizza.
theitalianstallion
Posts: 1,109
Add as Friend
Challenge to a Debate
Send a Message
6/5/2009 7:48:15 AM
Posted: 7 years ago
No. I think it is fine as it is.

With the system you propose, a new debater with only one win would be placed above LM, RR, RL, etc. which is just absurd.
When Reach fell, I came.
Xer
Posts: 7,776
Add as Friend
Challenge to a Debate
Send a Message
6/5/2009 12:37:55 PM
Posted: 7 years ago
TheSkeptic: "Win ratio is too flimsy to base it on. If my record is 1-0, then I have a perfect 100% record. So basically, any newcomer can come onto this site, win in a ridiculously easy debate, and trump me."

theitalianstallion: "With the system you propose, a new debater with only one win would be placed above LM, RR, RL, etc. which is just absurd."

No. I said, "There should be more emphasis on win ratio instead of wins." I didn't say that it should be based solely on win ratio. So, no, a person who is 1-0 would not be in the 100th percentile.
mongeese
Posts: 5,387
Add as Friend
Challenge to a Debate
Send a Message
6/5/2009 1:38:16 PM
Posted: 7 years ago
Well, we can either base it on total wins, win ratio, or some mix. What kind of mix do you propose?

And it still doesn't matter, because the masses are still easily fooled.
Volkov
Posts: 9,765
Add as Friend
Challenge to a Debate
Send a Message
6/5/2009 1:39:33 PM
Posted: 7 years ago
Here's what annoys me about the current rating system.

User Republican95 has been participated in 5 debates, and lost them all. His percentile is 47.86. ( http://www.debate.org...)

User komachi has participated in 1 debate, and won. His percentile is 47.86. (http://www.debate.org...)

What is that!?

I don't understand why we have the percentile system anyways. Why not a points system? I mean, for example:

Republican95: 0 W, 5 L, 0 T, 5 TOT
Using the formula W + (TOT - T) - L, we get 0.

komachi: 1 W, 0 L, 0 T, 1 TOT
1 + (1 - 0) - 0 = 2

For the top three we get:

beem0r: 119 W, 18 L, 1 T, 138 TOT
119 + (138 - 1) - 18 = 238

L-M: 115 W, 26 L, 3 T, 144 TOT
115 + (144 - 3) - 26 = 230

Kleptin: 110 W, 8 L, 2 T, 120 TOT
110 + (120 - 2) - 8 = 220

For someone that is currently in the Top 10, has a lot of debates, but has a lower win ratio, they get:
brian_eggleston: 81 W, 54 L, 7 T, 142 TOY
81 + (142 - 7) - 54 = 162

For myself, I get:

Volkov: 5 W, 1 L, 1 T, 7 TOT
5 + (7 - 1) - 1 = 10

For someone with the same amount as debates as me, but different W-L numbers, they get:

Nik: 2 W, 3 L, 2 T, 7 TOT
2 + (7 - 2) - 3 = 4

With this system, you are noted for participating in many debates, for which you should be rightly recognized for, but if your win-to-loss ratio is lower than those that are near your around your total debates score, you are penalized as such.
If your win ratio is 100% but you haven't participated in as many debates, you won't get such a high score as others.

Now I'm not advocating this system, but it is just an idea, because I have to agree - this does need major reform.
mongoose
Posts: 3,500
Add as Friend
Challenge to a Debate
Send a Message
6/5/2009 1:45:19 PM
Posted: 7 years ago
I think it should total wins x win ratio.
It is odd when one's capacity for compassion is measured not in what he is willing to do by his own time, effort, and property, but what he will force others to do with their own property instead.
I-am-a-panda
Posts: 15,380
Add as Friend
Challenge to a Debate
Send a Message
6/5/2009 1:49:07 PM
Posted: 7 years ago
At 6/5/2009 1:45:19 PM, mongoose wrote:
I think it should total wins x win ratio.

Encourages vote bombing.
Pizza. I have enormous respect for Pizza.
Xer
Posts: 7,776
Add as Friend
Challenge to a Debate
Send a Message
6/5/2009 1:49:38 PM
Posted: 7 years ago
At 6/5/2009 1:45:19 PM, mongoose wrote:
I think it should total wins x win ratio.

That wouldn't work. 100-100 would be the same as 50-0.
mongeese
Posts: 5,387
Add as Friend
Challenge to a Debate
Send a Message
6/5/2009 1:56:27 PM
Posted: 7 years ago
I think that 100-100 and 50-0 being the same is kind of decent, as it still encourages debating.

Of course, the problem lies where someone who just lost five debates would suddenly want to start an entirely new account so that those five debates don't drag his record down for the rest of his life.

Also, no matter what the leaderboard says, as long as wins are sought after, vote bombing will be encouraged.
Volkov
Posts: 9,765
Add as Friend
Challenge to a Debate
Send a Message
6/5/2009 2:49:43 PM
Posted: 7 years ago
Hey, quick question: do we know what the current formula for the system is? Because I think I just figured it out, or at least most of it.
Puck
Posts: 6,457
Add as Friend
Challenge to a Debate
Send a Message
6/5/2009 4:02:50 PM
Posted: 7 years ago
At 6/5/2009 1:39:33 PM, Volkov wrote:
Here's what annoys me about the current rating system.

User Republican95 has been participated in 5 debates, and lost them all. His percentile is 47.86. ( http://www.debate.org...)

User komachi has participated in 1 debate, and won. His percentile is 47.86. (http://www.debate.org...)

What is that!?

The nature of the statisitcal percentile rank.
Komachi has a percentile of 72

http://upload.wikimedia.org...

As you can see the differences between between scores is not uniform - differences between a 50 and 70th percentile is roughly the same as 95-99. A score of 40 isn't bad, it sits just below the mean. That being said a percentile doesn't differentiate between a 'good' and 'bad' result - one can achieve a 99th percentile and still fail a test for example.

I don't understand why we have the percentile system anyways. Why not a points system? I mean, for example:

You can easily have both. A points system gives a score that score is then ranked relative to others. I'm not sure what formula Phil uses, but it's not an accurate percentile rank one - there is no 100% in a percentile calculation due to ones own score is always included - Number of scores below X/Total number of scores x 100
I-am-a-panda
Posts: 15,380
Add as Friend
Challenge to a Debate
Send a Message
6/6/2009 4:21:22 AM
Posted: 7 years ago
At 6/5/2009 2:49:43 PM, Volkov wrote:
Hey, quick question: do we know what the current formula for the system is? Because I think I just figured it out, or at least most of it.

We have no idea of it. Please enlighten us.
Pizza. I have enormous respect for Pizza.
mongoose
Posts: 3,500
Add as Friend
Challenge to a Debate
Send a Message
6/6/2009 6:38:04 AM
Posted: 7 years ago
Current formula:

First, check wins. People with more wins are placed above people with less wins.
Second, check ties. People with equal wins and more ties are placed above people with equal wins and less ties.
Then, if there are no wins or ties, check if there are any losses. If so, people with losses are placed above people with no debates.

This is then put into a percentile system, based on what percent of the population is below somebody.
It is odd when one's capacity for compassion is measured not in what he is willing to do by his own time, effort, and property, but what he will force others to do with their own property instead.
I-am-a-panda
Posts: 15,380
Add as Friend
Challenge to a Debate
Send a Message
6/6/2009 6:48:20 AM
Posted: 7 years ago
A person with 1 debate and 1 loss has a percentile of 47.92%. This means there's a constant in the formula.
Pizza. I have enormous respect for Pizza.
Volkov
Posts: 9,765
Add as Friend
Challenge to a Debate
Send a Message
6/6/2009 11:47:04 PM
Posted: 7 years ago
Current formula used:
Legend: W - Wins; L - Losses; T - Ties; TOT - Total Debates; WR - Win ratio; PNT - Points;
PCT - Percentile


(TOT - T) x WR = PNT ---> PCT

I found this out while there was nothing on television. I have yet to find a situation on the leaderboard where it isn't compatable. Let me demonstrate.

Top 3:

beem0r: 119 W, 18 L, 1 T, 138 TOT, 86.86% WR, 100.00 PCT
(138 - 1) x 0.8686 = 118.9982 PNT

L-M: 115 W, 24 L, 5 T, 144 TOT, 82.73% WR, 99.97 PCT
(144 - 5) x 0.8273 = 114.9947 PNT

Kleptin: 110 W, 8 L, 2 T, 120 TOT, 93.22% WR, 99.93 PCT
(120 - 2) x 0.9322 = 109.9996

Same Percentile, Different Win Ratio & Number of debates completed:


Sweatingjojo: 16 W, 16 L, 2 T, 34 TOT, 50.00% WR, 98.21 PCT
(34 - 2) x 0.5000 = 16

Daxitarian: 16 W, 4 L, 1 T, 21 TOT, 80.00% WR, 98.21 PCT
(21 - 1) x 0.8000 = 16

Percentiles just above and below 98.21 (16 PNT):


Above
rwebberc: 16 W, 2 L, 0 T, 18 TOT, 88.89% WR, 98.24 PCT
(18 - 0) x 0.8889 = 16.0002 PNT

Below
Pricetag: 16 W, 3 L, 1 T, 20 TOT, 84.21% WR, 98.11 PCT
(16 - 1) x 0.8421 = 12.6315

Minutest difference possible between two percentiles:


I-am-a-panada: 16 W, 20 L, 4 T, 40 TOT, 44.44% WR, 98.04 PCT
(40 - 4) x 0.4444 = 15.9984 PNT

zakkuchan: 15 W, 2 L, 0 T, 17 TOT, 88.24% WR, 98.01 PCT
(15 - 0) x 0.8824 = 15.0008 PNT

All you do is translate the PNT into a percentile score, and there you have it. That is how it works, and you can sample any random segment of the leader board and the formula will work.
Volkov
Posts: 9,765
Add as Friend
Challenge to a Debate
Send a Message
6/6/2009 11:48:29 PM
Posted: 7 years ago
At 6/6/2009 11:47:04 PM, Volkov wrote:
zakkuchan: 15 W, 2 L, 0 T, 17 TOT, 88.24% WR, 98.01 PCT
(17* - 0) x 0.8824 = 15.0008 PNT

*
I messed that up by accident, its meant to be 17, not 15 originally.
MTGandP
Posts: 702
Add as Friend
Challenge to a Debate
Send a Message
6/19/2009 9:37:55 PM
Posted: 7 years ago
Volkov's formula is the equivalent of:

_______ wins (wins + losses)
score = -------------------------
____________ losses

This is a good system. Unfortunately, its non-linearity makes it difficult to analyze. I'll do what I can.

Imagine person A and person B. If person B has twice as many wins and twice as many losses as person A, then person B has twice the score. That sounds fair.

If person B has twice as many wins and the same number of losses, the score increase varies based on the proportion of wins and losses. But if we assume that person A has the same number of wins and losses, then person B will have 3 times the score of person B. If person A has three times as many wins as losses (which isn't so rare on DDO), person B will have 2.5 times the score.

SR = person A's score over person B's score
WR = person A's win ratio

lim (WR ==> infinity) SR = 2

In other words, as person A's win ratio increases, person B's advantage over person A gets closer and closer to 2.

Another formula:

lim (losses ==> infinity) SR = wins

In other words, as losses continue to increase, the score gets closer and closer to the number of wins but never quite reaches it or goes below it.

***

Conclusion

It is advantageous to increase total number of debates, but it is more advantageous to increase number of wins. But if person A already has a lot of wins, person B doesn't gain much of an advantage by further increasing the number of wins.

I see this as a very fair system. Matches are good. Wins are good, but the more you have, the less important they start to become.

========
System 2
========

This is a system which is linear instead of quadratic. But as I analyze it, I realize that it is actually about as complicated.

score = 2*wins - losses

If person B has double the number of wins and losses as person A, then person B will have double the score. So this aspect is the same for both systems.

If person B has double the number of wins as person A, then the advantage is dependent on the number of losses. If there are zero losses, it will strictly double the score. If there are as many losses as wins, then the score will go up by 3 times. In this system as well, more losses equate to more wins, but the rate of change is different. Interestingly enough, if there are twice as many losses as wins, the score will go up infinitely (since twice as many wins as losses equates to a score of zero). And if there are more than twice as many losses than wins, things get a little crazy.

System 2, while simpler on the surface, leads to some problems when the number of losses is greater than twice the number of wins. I don't think this is much of a problem as long as you can handle negative scores, though.

========
Comparison
========

Which system is better? These two seem to be the best systems. Let's rate them.

Simplicity: System 2 wins easily.
Accuracy: This is much harder to judge. I think I'd give this to system 1. Although the systems are similar, I think that system 1 seems more elegant since the score never increases without bound and never goes negative.

========
Conclusion
========

System 1 is a more elegant system, and I call for the immediate implementation of Volkov's system!
MTGandP
Posts: 702
Add as Friend
Challenge to a Debate
Send a Message
6/24/2009 4:29:54 PM
Posted: 7 years ago
I have noticed a major flaw in Volkov's system. If the person has zero losses, the score will be undefined. There needs to be something besides losses in the denominator. A simple "losses + 1" will suffice.
Volkov
Posts: 9,765
Add as Friend
Challenge to a Debate
Send a Message
6/24/2009 4:36:27 PM
Posted: 7 years ago
At 6/24/2009 4:29:54 PM, MTGandP wrote:
I have noticed a major flaw in Volkov's system. If the person has zero losses, the score will be undefined. There needs to be something besides losses in the denominator. A simple "losses + 1" will suffice.

I'm not sure what you mean. If the person has zero losses, but 100% win?

1 x 1.00 = 1 point -> Percentile. It still makes sense.
MTGandP
Posts: 702
Add as Friend
Challenge to a Debate
Send a Message
6/24/2009 4:42:34 PM
Posted: 7 years ago
At 6/24/2009 4:36:27 PM, Volkov wrote:
I'm not sure what you mean. If the person has zero losses, but 100% win?

1 x 1.00 = 1 point -> Percentile. It still makes sense.

The win ratio is not 1. It is undefined. Win ratio is wins / losses, which requires division by zero.