Project ELO (Dataset released)

[MagnaofIllusion]

I'm guessing it's a curiosity thing mainly, which is more or less how I treated it.

Well I agree but even as a curiosity there are huge swaths of players who are not currently on the board simply because they don't play any significant number of Newbie or Normal games. So even as a curisoity it falls flat IMO.

Interesting exercise but really needs a massive number of people doing the data compiling.

[Realeo]

Also, realeo, did you end up adding in normal stats or is this just newbie games?

I add both normal stats

[Realeo]

288 players are eligible.
1852 players are not eligible due to "min.5 games rule"

Until more people are eligible than not posting "Top 25" lists is highly skewed. Not sure if the effort of going through enough games is worth it to you.

I disagree.

Top 25 is top 8%.

Top 8% is quite impressive.

[LicketyQuickety]

288 players are eligible.
1852 players are not eligible due to "min.5 games rule"

Until more people are eligible than not posting "Top 25" lists is highly skewed. Not sure if the effort of going through enough games is worth it to you.

I disagree.

Top 25 is top 8%.

Top 8% is quite impressive.

I guess I would just say (as we have seen in the top 25) that the further down the list you go, the more people you see closer to each other. Naturally like a bell curve.

[Realeo]

That is true.

[Realeo]

So Top 5 is better?

[mhsmith0]

Not especially, you just need to know there's a lot of noise and don't want to go nuts with your results being super representative or the like. Sample size and variance, ESPECIALLY for town games, is a big deal.

[Realeo]

I really want to change to Glicko rating system after I combined the Newbie Games. Glicko rating system is like ELo, but with a uncertainty level.

So instead of ELO 1520, you have ELO 1520

+

50 to address uncertainty. Which means your actual ELO is around 1470 ~ 1570.

[mhsmith0]

Well if you add in up to newbie 1755 that should help my rating either way

[MagnaofIllusion]

I disagree.

Top 25 is top 8%.

Top 8% is quite impressive.

Feel free to disagree but throwing stats that aren't meaningful isn't very helpful.

Your 25 players listed may not even be in the real Top 100 of all players. Or Top 200. or Top 500. There is no way to know given the small sample size you are using. As I said until you get at least half the player-base included in your pool of eligible players then your Top is just listing of the best ranked players in a small strata of the site's games.

I mean I understand the fun in doing it for you.

[Realeo]

Feel free to disagree but throwing stats that aren't meaningful isn't very helpful.

Your 25 players listed may not even be in the real Top 100 of all players. Or Top 200. or Top 500. There is no way to know given the small sample size you are using. As I said until you get at least half the player-base included in your pool of eligible players then your Top is just listing of the best ranked players in a small strata of the site's games.

I mean I understand the fun in doing it for you.

The biggest problem is the fact that a striking proportion of the 2000 players is just a one-timer or two-timer players.

I mean, people called Andy Murray the top tennis player because he has a consistent track record. Nobody said "Hey! I think this Person X is the best tennis player in the world because of this, despite he only play 1 game."

You are correct. My Top 25 players may not even be in the real Top 100 players. Very True.

There is no way to know the actual accuracy, but there is a way to increase the accuracy.

The track record is a very valuable proof to increase the accuracy of the list. One thing we have to consider is the possibility of

hot streak

or fluke. By having 5 games limit, it is expected that that lucky player would run out of luck and would finally lose. It is expected that over time, people's track record would converge to a certain point to show your actual skill.

Because of that, I apply (the seemingly) arbitrary 5 games limit to wait for the ELO to converge. If the ELO doesn't converge then it's a fallacy to imply A is better than B from non-converged ELO, which is why I don't include it.

Are 5 games enough for the ELO to converge? This is debatable and researchable. But the limit is important.

[Realeo]

There are 90 games yet to analyze~a combined players list of at least 1 000 players.

We haven't consider Open Games.

We won't consider Theme Games.

[Realeo]

To be fair, it's not a problem of whether we have enough

time

to dig. It's whether we have enough

game

to dig.

[MagnaofIllusion]

There are 90 games yet to analyze~a combined players list of at least 1 000 players.

We haven't consider Open Games.

We won't consider Theme Games.

So you aren't going to consider what in all likelyhood amounts to a majority of the games on site?

Well thanks for the clarity on that.

[Realeo]

For context, the Newbie Games data extends to the first game of Matrix-6 (2013) and Micro Normal data extends to 2012.

[Realeo]

So you aren't going to consider what in all likelyhood amounts to a majority of the games on site?

Theme Games is not necessarily designed with balance in mind (ie. Mod-intervention game). Some does, but we cannot generalize for the purpose of the leaderboard.

There are 2 ways to process Theme Games.

[1] Somehow extract data out of it.
[2] Ignore it.

Until I find a way to solve [1], [2] it is. The leaderboard is designed to tackle extreme player diversity, but not extreme game diversity.

[Realeo]

I just remember that there is Micro Open and Micro Normals.

Also, realeo, did you end up adding in normal stats or is this just newbie games?

Normal + Newbie

[Realeo]

I do have a proposal on how to process Large Theme, but for order, I will keep it for later.

[callforjudgement]

Even not including Theme games, it's still a reasonable question to say "who's the best player at Newbies, Normals and Opens". Someone might be a great player, but if they never play in the simpler games, they can't reasonably be called the best Newbie/Normal/Open player.

[Realeo]

I re-open the motion of finding "Champion over Different Queue" instead of simply "Generalized Champion of Everything"

After I saw that the data extends back to 2012 & 2013, I feel more confident at the system.

Although I still don't really get why.

[LicketyQuickety]

So Top 5 is better?

I would say rank everyone.

[Realeo]

The complete rank is avaliable.

[LicketyQuickety]

The complete rank is avaliable.

Where is it? I want to see how RC did.

[Realeo]

Update 2

Current Dataset:
Newbie 1370 ~ Newbie 1699
Mini Normal 1401 ~ Mini Normal 1789

We consider 501 games and 288 players had at least 5 games to be eligible for the elo system. The "1 game in the last 365 days" has been waived for this update until Newbie 17xx is updated.

The Elo can be found here

<---There! The complete rank Click it!

Things to do
[+] Integrate Newbie 17xx
[+] Find misspelled user (We have an algorithm in place to automatically detect it.. It's slow. I haven't coded the Dynamic Programming version.)

[LicketyQuickety]

Update 2

Current Dataset:
Newbie 1370 ~ Newbie 1699
Mini Normal 1401 ~ Mini Normal 1789

We consider 501 games and 288 players had at least 5 games to be eligible for the elo system. The "1 game in the last 365 days" has been waived for this update until Newbie 17xx is updated.

The Elo can be found here

<---There! The complete rank Click it!

Things to do
[+] Integrate Newbie 17xx
[+] Find misspelled user (We have an algorithm in place to automatically detect it.. It's slow. I haven't coded the Dynamic Programming version.)

Thanks, shocked by titus' score.

Do you happen to know the average Score?