wow, so someone actually plays the game?

also gj not using nosql nor mysql bullshit

pretty cool, good job!

tracked harder than the NSA

not bad for a dead game

More stats! These are uplifting

Re-install for a cake icon or not?

Hype for CakeCon 2017!

[img1]

A lot of the time when I search players, the same player has multiple entries, occurring again and again. Why does this happen? And does this inflate the number of tracked players?

Probably different names used by the same steamid. No it wouldn't inflate players.

^ this

How many deactivated? lel

i lite a blunt in this celebration festviety =]<o~~~

26 GB postgresql database

No wonder it had problems handling the traffic, periodically.

156 thousand players
Barely enough to fill 2 NA ffa servers and 1 CA server

OMG you just had the most amazing idea.... Quake MMO!!!!

Im emailing tim willits right fking now... there might still be time to make it happen!!!


or not....

(but the idea would till be dope)

Idea was great till you mentioned Willits.

with trump will be different

Would you consider making (a sample of) that database available publicly so that we can do some data analysis on it? E.g. I'd love to draw some graphs about weapon statistics and how it influences ranking. Tons of other possibilities. The data geek in me is salivating at just the thought of it. A good source to host it could be the Datasets section at Kaggle.com

Pred was willing to give me a pack of a few thousand JSON's, if you are into that kind of thing. Sample db or 30-day DB could be good, as long as it's not hindering main server performance, which is demanding considering the things he mentioned above.

For data privacy reasons I can't make it public, but I can give individual people access to it when they explain what they will do with it and agree that it must not be used for anything else.
The zipped version of the DB is about 5GB.

The raw zipped JSON data files are about 450MB/month. Those files are stored separately from the DB and not included in the 26GB.

I haven't played Quake for the last few years so I'm not very familiar with what stats are available to base calculations on, but these are some things I came up with:

- Player's glicko rating per map
- How does weapon accuracy relate to a player's ranking per map?
- How do weapon # shots relate to a player's ranking per map?
- How do weapon # shots relate to damage output per map?

I spoke briefly to one of the Kaggle admins. They can host datasets for free up to 500MB and can make exceptions up to 5 GB.

What's up with the data privacy? :). Publishing the dataset won't intrude people's privacy anymore than the qlstats website is doing right now? :)

Myself I'm a bit busy at the moment, maybe I can revisit this some time in next months. But I may not be the only person here interested in analysing this so let's not finish this conversation just yet :)

What's up with the data privacy?

QLStats allows to hide player's match history
Example: http://qlstats.net/player/14843
How to do it: http://qlstats.net:8081/login

It's a really dumb oversight having that feature, for the limited privacy control that's granted. Basically the site will become a cheat hiders paradise.

The stats need to be public so we can easily see who the aim-botting scumbags are.

It's a matter of not getting your ass sued, not an oversight. That said, it doesn't mean the one suing would win.

It would be nice to do that! I thought about doing graphs ranking x weapon accuracy in the past to see which curve approximates it the most, altough quake is so up to player interpretation I dont think graphs will highlight any precise trend.

Weapon accuracies depend a lot on playstyles, even one player abusing pads for shaft or spamming corners to have an edge will screw it up the accuracy - rank ratio. Ofc im expecting a linear (or logarithmic) trend but it will have so many singularities it will be worthless. Anyway it would be nice to do it just to see it confirmed.

Its the same reason we can't have an average rocket damage accuracy, it takes away the contest (spamming, missing, timing rockets just to hit directs).

Quake is so complex you'd have to set so many parameters to get reliable data, i think its not doable.

I think to be able to work with weap accuracies you should set players with same weapon time utilization per map, and then check their damage output over their rank.That way it should work, but its a lot of effort and time.

All we can do is get data about common purposes in a mode (win/lost rounds in ca, enemy flag captured/your flag captured in ctf) cause despite game interpretation, thats the common ground.

Even this current ca ELO is an approximation of what it should be.

Speaking about weapon spam influencing accuracy, I don't think it is that much of a challenge to filter that out.

E.g. on a certain map, filtering for 1v1, you could create bins for average # of weapon shots fired per weapon, and then graph the accuracies (possibly a box&whiskers plot). You can then check where you are on that graph and compare with other players who spam similarly like you on that map.

Further 'normalisation' of data should be possible.

I would actually be interested to team up with someone to do this kind of stuff. Could be fun :). I'm using Python Pandas and Matplotlib for these things.

Replying to the privacy thing: I'm not sure it would work out well for Evil or any other top player to hide their identity.

I'm a real 'privacy' aware person myself, but hiding a player's Quake stats - a person hidden behind a nickname anyway - doesn't feel very valuable to me.

E.g. on a certain map, filtering for 1v1, you could create bins for average # of weapon shots fired per weapon, and then graph the accuracies (possibly a box&whiskers plot). You can then check where you are on that graph and compare with other players who spam similarly like you on that map.

I dont know about that. You can then compare results only within players who spam the same way, not mentioning that its not easy to tell when someone is spamming or not.

It seems like a lot of work and unreliable results. You could achieve the same by giving everyone a standard mouseaim test and compare results with their rankings.