Goldeneye Historical Record Holders / Visualisation

[GreasePig]

Hi,

I'm creating a data visualisation of historical goldeneye record holders.

Example screenshot:



In order to create this I'm aggregating all Goldeneye PRs and recalculating the expected point rankings on a given date according to the point rules.

It's easy for me to confirm that the my point calculations for the present day are correct, however there's no straightforward way (that I can see) to validate the historical point calculations as the-elite website doesn't show point rankings on past dates (only individual records on past dates), similarly for "GoldenEye World Record Database" site.

By checking the way-back machine, we can see how the page appeared in December 2011:
https://web.archive.org/web/20111230200622/http://rankings.the-elite.net:80/goldeneye
As you can see, Clemens had 5945 and I. had 5896.
My calculations are very close to these numbers but not completely identical.

Reasons for the discrepancy I can imagine:

A) the wayback machine happened mid-day while the numbers were still in flux for the day (unlikely)
B) new times were added later by other players with retroactive dates that affected these scores
C) old times were deleted by mods (invalid, no proof etc.)
D) some players deleted their own scores
Anything else?

If the reasons are A-C, then that's fine, I can trust my calculations as correct
If 'D' is possible, then I cannot trust my calculations.
Question 1: Can anyone confirm if some historical scores can/have been removed by players?

----------------

Question 2: What should be considered the correct start date of the rankings? I can see PR dates going back to 1998-07-26, however the data is sketchy and chaotic around then. I believe the point system was introduced in 1999? Anyone know the exact date? (Similar question for perfect dark, initial data is sketchy, do those players around then deserve leader status?)

----------------

Question 3: How are timezones handles with regards to the run dates? There seems to be no timezone related information related to the dates, always being in the user's local timezone. This creates a strange situation where run A may occur before run B, but due to timezone differences A may be displayed as the day after B. I've gone with how the dates are displayed on user's PR History pages with no consideration of timezones for the time being, is that the most correct approach? (This affects the 'number of days holding no. 1' statistic).

----------------

Question 4: According to my calculations there are no days where the top two players had the same number of points (though there is for perfect dark). In those cases are both players considered the leader and both their 'Day Count' should be incremented? (alternative is neither player)

----------------

Finally, if anyone has feedback on the visualisation itself that will be appreciated.

Cheers,

[Gelonidres]

Looks good!

In order to create this I'm aggregating all Goldeneye PRs and recalculating the expected point rankings on a given date according to the point rules.

It's easy for me to confirm that the my point calculations for the present day are correct, however there's no straightforward way (that I can see) to validate the historical point calculations as the-elite website doesn't show point rankings on past dates (only individual records on past dates), similarly for "GoldenEye World Record Database" site.
[...]
Question 2: What should be considered the correct start date of the rankings?

I don't know how you are collecting information on all PRs, but I doubt the information on very old PRs is correct, as that data just doesn't exist anymore. If you want to check how accurate very old data is, you can compare your point rankings of late 2000 here. You'll find some more historic rankings here. Comparing your All Time Leaders to the Goldeneye Champions page on speedrunwiki, you'll find rather big discrepancies between the two, mainly for old players.
Thus, it might make sense to have your tool start displaying the points as shown in the screenshot you provided only from a point of time where the PR info is fairly complete and use the speedrunwiki data for the "All Time Leader" day count on the years before that. I guess that would probably be something around 2011, but more knowledgeable eliters will be able to give better guesses than me.

If the reasons are A-C, then that's fine, I can trust my calculations as correct
If 'D' is possible, then I cannot trust my calculations.
Question 1: Can anyone confirm if some historical scores can/have been removed by players?

B-D have certainly happened. Times that come to mind for D are Frigate A 23 and Caverns A 1:01 by Marc. Veteran players can probably name more. I guess that B would cause bigger discrepancies than D though.


By the way, if you collected all PR information on the Elite anyways, it would be cool if you could dump all that info in a txt file somewhere. 

[flukey lukey]

lovin the graphics .. so far !

[AZ]

  Andrew Kent - 7 Days

[Wouter Jansen]

The data is in the live database, deleted times are still in there too

[渚カヲル]

  Andrew Kent - 7 Days

[vitorr]

How are timezones handles with regards to the run dates?  -> The Date Achieved is just a plain Date type (different from Date Posted, which is a unix timestamp in UTC), so you're correct. The default Date shown when posting a new record uses the current time (unix timestamp in UTC too), so if you're UTC-2 and post at 23:30, the default value for you is going to be the "next" day (which would be 01:30 UTC), but you can change it to your local current day anyway. I haven't checked this myself though, I'm using our source code as a reference (the default value is the value of date() PHP function). And when posting old times (e.g Kapap) people will most likely use their local times, so it can get messy.

For example: dsx matey got silo SA 108 on 2017-06-01 which was posted after Kapap 3. He lives in Perth (UTC+8), so say he got it at 04:00 (4AM) - that's 2017-05-31 20:00 in UTC time. In UTC it would be a time achieved in May 2017, not June 2017. Probably NOT this specific case as I suppose he got it 2017-06-01 somewhere between noon and midnight so it wouldn't make a difference. But you get the idea  At the same time if I had gotten silo SA 108 on 2017-05-31 @ 23:30 (my timezone is -2), technically it was 2017-06-01.

Would be cool to have at least Date + Hour (full timestamp is not really necessary imo), and some kind of local time to UTC converter when posting (detecting your timezone beforehand) to avoid these inconsistencies. But probably we'll have to learn to live with it.

[GreasePig]

Thank you for the responses all, greatly appreciated.

I don't know how you are collecting information on all PRs, but I doubt the information on very old PRs is correct, as that data just doesn't exist anymore. If you want to check how accurate very old data is, you can compare your point rankings of late 2000 here. You'll find some more historic rankings here. Comparing your All Time Leaders to the Goldeneye Champions page on speedrunwiki, you'll find rather big discrepancies between the two, mainly for old players.
Thus, it might make sense to have your tool start displaying the points as shown in the screenshot you provided only from a point of time where the PR info is fairly complete and use the speedrunwiki data for the "All Time Leader" day count on the years before that. I guess that would probably be something around 2011, but more knowledgeable eliters will be able to give better guesses than me.

Thanks for those links, those ancient scores are much better than the ones I've been calculating.

The "Goldeneye_Champions" wiki page is great, though there seem to be a few anomalies:

1. Wiki says "Wouter, 2001.9.1 - 2003.12.24 (844 days)"
However the snapshots on following dates have Randy as the leader:
2002-04-18 - https://web.archive.org/web/20020418021456/http://www.the-elite.net:80/GE/elite.htm
2002-06-07 - https://web.archive.org/web/20020607135739/http://www.the-elite.net:80/GE/elite.htm

2. Wiki says "Illu, 2007.1.10 - 2007.4.12 (92 days)"
However the snapshots on following dates have Bryan as the leader:
2007-02-08 - https://web.archive.org/web/20070208064051/http://www.the-elite.net:80/GE/elite.htm
2007-02-19 - https://web.archive.org/web/20070219055235/http://www.the-elite.net:80/GE/elite.htm
2007-03-21 - https://web.archive.org/web/20070321152240/http://www.the-elite.net:80/GE/elite.htm

Were some runs of Randy and Bryan later disgraced / proved false and thus the real point leader was retroactively changed? If so anyone know how much I should drop their point scores and during what rough period?

Separate to these, I can see some other players have had drops in scores over time, so this wouldn't be unusual, eg. Ryan and David have massive plummets for short times with returns greater than before, smaller dip for Bryan (see pic below), presumably these are due to runs being removed then better times being posted shortly after (if anyone has a different theory please let me know). I could easily plug these gaps by interpolating the scores in-between, or keep them as-is, depends on the reason for the drops.

Edit: Another strange sudden change in score is Karl Jobst from 10th in the 2004-06-09 WayBack snapshot to completely non-existent in the next one, only to return to the charts again in 2006-02-17 - did Karl ask to be removed from the pages or something? This one is less important as he's not in the top few ranked players during that time, but still curious.
Edit2: Karl confirms he disappeared due to range-quitting during that time.



Fortunately, the point in time where I transition in the visualisation from the old wayback scores (smooth curves in this pic) to the new calculated scores (jagged curves in first pic) is quite seamless, giving me high confidence that the retroactively calculated scores for the modern era are quite accurate.

-------

For the super-ancient period between 1997-08-25 and 2000-04-22 I'm just going to display the three players Sterling, Martin and Zwartjes (with no Y-Axis numbers) since the only data I have for that period is the relative ranking of these three players. If anyone knows anything else about this period, that will be useful for the visualisation.

By the way, if you collected all PR information on the Elite anyways, it would be cool if you could dump all that info in a txt file somewhere. 

I'll happily share some data if the mods are fine with it.

The data is in the live database, deleted times are still in there too

Is this live database available anywhere? I'm currently inferring some deleted times from the WayBack machine but it's not completely
 accurate as the page snapshots are only available sporadically there. Having the actual database info will improve the accuracy of the chart slightly, accurate created_at and deleted_at timestamps of all PR history would make the chart 100% accurate.

----

Thanks for the timezone explanation vitorr, appreciated. I think I'll just live with the time postings being accurate to +- 1 day, it's impossible to discern a single day from the scale of the chart anyway.

[Wouter Jansen]

I think Scott or Vitor can make you a copy? I'm not sure who all have access

[vitorr]

Nah I don't have a copy neither access to production DB (wish I had, though, Scott denied all my requests ).

Seriously though, maybe Scott / Thingy / TheFlash / <insert alt nickname here> could help you with this. Maybe he could generate a temporary database for you with only the deleted times to make things easier. A full DB dump is way too big and getting access to live production DB is a little more complicated.

[GreasePig]

Mamma mia, this took way more time than I had expected. The time is finally here to watch...

Speedrun Champions: GoldenEye 1997 - 2018
http://www.youtube.com/watch?v=XpounXuqFoo

Some notes:
- very old historical periods have no y-axis labels as the points are not clear initially, only the leader of the time from the wiki is known
- when data conflicted with the wiki, I have modified it to fit the known leader of the time, the most obvious cases were noted in my previous post, though most transitions of leadership required some tweaking to get the points & dates right
- the top-10 at yearly snapshots has been cross-checked and phantom data introduced to make sure it matches
- since lots of interpolation and tweaking is required, the note on the left of the champion's points is prefixed with "~" and rounded to the nearest 10 for most of the video
- some of the strange gaps in people's curves have been smoothed over, making the assumption that the data of the time had issues
- it seems the further back in time one goes, the more work was required to get things right, PR history only seems right for the most recent dates
- deleted some cheating player's data
- the Wiki tallied Marc's days as leader incorrectly, so I've edited the wiki with the correction
- lots of other tweaks and fixes to make sure the zoomout at the end looks legit

If anyone has any other questions let me know, hopefully no egregious errors that require a repost.

Off topic: during this exercise I noticed the elite leader boards have the level "Egyptian" mislabelled as "Egypt". All other levels are named correctly (it's not Archives is shortened to Archive). Doesn't this trigger anyone else!?

[T+]

Speedrun Champions: GoldenEye 1997 - 2018
https://www.youtube.com/edit?o=U&video_id=XpounXuqFoo

Nice link

Great vid btw

[GreasePig]

Haha whoops, fixed, thanks.

[OHMSS]

GREAT video mate!! That vivid speedrunner sperm.

Wtf happened at the start of 2000? Probably the beginning of good data records. Or many people getting internet access to celebrate the millenium; or speedrunning GE to prepare for the PD release.

The rise of Clemens is insanely visible, very cool. Wodahs as well, but that niggu was hoarding skill And the stagnant pre-twitch years ... flat lines all the way. The 2016 skirmish for the champ spot was also great to see.

[50]

that vid is fucking tight

[flukey lukey]

GREAT video mate!! That vivid speedrunner sperm.

Wtf happened at the start of 2000? Probably the beginning of good data records. Or many people getting internet access to celebrate the millenium; or speedrunning GE to prepare for the PD release.

The rise of Clemens is insanely visible, very cool. Wodahs as well, but that niggu was hoarding skill And the stagnant pre-twitch years ... flat lines all the way. The 2016 skirmish for the champ spot was also great to see.

[Icy]

10/10 video, that was absolutely fantastic! Great work!

[GreasePig]

Cheers brothers! Sweated my arse off on this one so happy to hear it's received well.

I don't think interest in the game necessarily took off in the year 2000, just that before that time I only was able to find out who was no. 1, not the rest of the top 10 or actual point values.

For anyone interested, this was rendered in JavaScript with canvas, exported to WebM with "CCapture" library, with music and sound effects added in iMovie. Was hoping to make it 4K 60fps but iMovie only supports 4K or 60fps, not both  - Render time about 6 hours (which I had to do 5 times cause each time I noticed small problems).

[Gelonidres]

This video is top-notch. I can actually see it being linked on the past records page as a substitute for not having past player rankings there. Maybe someone with the power to update pages will consider it!

Again, good stuff. It's easy to see that much work was put into this, and it really paid off. There is one frame where the lower half looks bugged at 2:16 though. 

[FOrK]

Brilliant. Guaranteed prop in all new speedlore vids

[Huzi]

Absolutely amazing!

[Shadow]

Heh, 2016 was a crazy year.

Loved seeing how "dominating" each champ was, from Sterling, to Wouter, to Bosshardt, Clemens, and Ace, there were definitely periods of clear domination.

At one point Bosshardt had nearly all the records, I don't remember, which ones was he missing?

Brilliant video, thanks!

[JDBlack21]

This video is absolutely fantastic. I've seen some visualization videos like this before, but this is among the best.

[deletedprofile.u]

Absolutely brilliant!

Much respect.

[eastwood]

great vid mate

[Smit]

Super interesting. The clash in July 2016 was insane , that fight for the top spot

[RWG]

This is obviously mega top tier.

Would it be cool if I used this in a video where I showed the progression + every time a new champion took over, I paused the progression and explained what happened at the time & showed the run that gave the champion first place?  I'd obviously attribute you on screen at all times and link to whatever channels you'd like.

"The 'WR' Progression of Goldeneye Champions" can become a real thing now.

[AZ]

One of the best videos ever posted on these forums.

[GreasePig]

Cheers y'all.

This is obviously mega top tier.

Would it be cool if I used this in a video where I showed the progression + every time a new champion took over, I paused the progression and explained what happened at the time & showed the run that gave the champion first place?  I'd obviously attribute you on screen at all times and link to whatever channels you'd like.

"The 'WR' Progression of Goldeneye Champions" can become a real thing now.

Hey Goose, sure mate, love your series. Happy for you to use this in your video as long as it's as you say, and not enormous chunks at once, so people still have a reason to go check out the original, please mention and link to the original vid and make a shoutout to "Puffrey Chambers"! Feel free to get in touch if you have any questions about the data etc.

Looking forward to your show.

[Scrambler Fanny]

*Heavy breathing*

Instant classic, vid!  Great job!

[OHMSS]

^ Told you you were 6th. And here's the evidence!

[AForgottenEvent]

Epic video is epic, holy shit