Create an account


Thread Rating:
  • 3 Vote(s) - 5 Average
  • 1
  • 2
  • 3
  • 4
  • 5
What's new in XonStat

Nice feature, Antibody!
Does the matrix take teamkills into account?
Reply

The matrix counts teamkills, but doesn't currently show any indicator for them. I suppose another feature could be that if the fragger's team is the same as the victim's team, color the background of the cell with some color (light red?) to indicate that it isn't something to be proud of.

There's other oddities with the implementation right now. I consider them pretty mild, especially given that we won't use the matrix values anywhere else but on the game info page:
  • Frags in warmup are counted (we could zero those counters with a change to the QC code)
  • Frags on people who left the game are not shown (this is normal, and nothing will be done to stop that)
Cheers!
asyyy^ | are you releated to chuck norris?
Reply

The "top servers" ranking and corresponding index is now based on the cumulative hours played. Here is what that looks like:

[Image: NHvGzeY.png]

If you click the '+' sign you're taken to the index page for the category where you can see the whole list from top to bottom. You can also see the full time listing down to the minute since there's room for it. Here is what that looks like:

[Image: 7w9MZZe.png]

*Note that these images were taken on my local machine and that I've stretched out the ranking window greatly for testing purposes. 

Previously these two views were based on the number of game played for the week. While it is a close approximation of the current list, it didn't prevent two people from playing five-minute games all day to reach the number one spot.
asyyy^ | are you releated to chuck norris?
Reply

Is it hard to include total damage made and total damage taken into xonstat score tables (like in the in-game scoreboard)? It would be useful for me if they were there, like for comparing my performance with teammates in tdms or cas.
Reply

I know how to do it, but haven't yet dedicated the time to do so.
asyyy^ | are you releated to chuck norris?
Reply

How can this guy be 1st place on dm stats?
http://stats.xonotic.org/player/98473
He doesn't play often  and he loses most of the time. This is not logical for me?
Reply

It's not about winning games. It's about your score per second versus your opponents during games. I could have a win percentage of 0 and still have a huge Elo if I had a better score-per-second than the majority of people in the matches I play. 

I've mentioned it before, but there's also a "grinding" component to Elo. If you play a ton of games (1300 *is* a lot) you can get to a high spot due to sheer volume. This is like a person who has no expenses that eventually amasses a fortune by picking up peoples' dropped pennies!
asyyy^ | are you releated to chuck norris?
Reply

(01-03-2018, 04:48 PM)Antibody Wrote: It's not about winning games. It's about your score per second versus your opponents during games. I could have a win percentage of 0 and still have a huge Elo if I had a better score-per-second than the majority of people in the matches I play. 

I've mentioned it before, but there's also a "grinding" component to Elo. If you play a ton of games (1300 *is* a lot) you can get to a high spot due to sheer volume. This is like a person who has no expenses that eventually amasses a fortune by picking up peoples' dropped pennies!

Elo, as a system for approximating relative skill levels between players - elo grinding is an exploit in the rating system not part of the system itself. Same goes for "picking" matches and deduction dodging.

Quote:How can this guy be 1st place on dm stats?
http://stats.xonotic.org/player/98473
He doesn't play often and he loses most of the time. This is not logical for me?
If you ask me, Elo is poorly implemented in this game. Much of the ratings except maybe duel, because there's a very dedicated community for it, don't really indicate much, but even duel elo ratings are a bit off. I.e the relative skill difference between players doesn't work globally when there's completely separate playing populations (private server, which is that guy's case, or regions).
Reply

(01-04-2018, 01:19 AM)Antares* Wrote: If you ask me, Elo is poorly implemented in this game.  Much of the ratings except maybe duel, because there's a very dedicated community for it, don't really indicate much, but even duel elo ratings are a bit off. I.e the relative skill difference between players doesn't work globally when there's completely separate playing populations (private server, which is that guy's case, or regions).

It is generally not really helpful to just say that it is bad if you don't also deliver a clue how to make it better.

A player's skill is something that is really difficult to grasp in a single number, and I highly doubt it is even possible in a way that delivers a close estimate of a player's skill. This is just because ELO is just a single number representing a whole bunch of variables that depend on game mode and playing style of the opponent(s) and whatnot. I have made the experience that I play much stronger in duel than I do in other game modes. But other variables affect playing performance as well - opponent's behavior, your own adaptiveness, does your playing style work against the opponent's playing style... all this simply does not fit into a single number, even if we have a single number for each game mode (which we effectively have).

This means that of course ELO ratings are no accurate skill values but only rough estimates that are by definition mostly off. And unless someone finds some sort of magic solution this is all ELO will ever be. Needless to say, volunteers to develop a magic solution are welcome.
My Xonstats Profile
Latest track on soundcloud: Farewell - to a better Place (piano improvisation)
New to Xonotic? Check out the Newbie Corner!

Reply

Rankings on XonStat are now disabled. I've removed references to them on the front page and the player info pages. JSON views will continue to operate as before, but with the ranks attribute yielding empty objects. The in-game view of this information will show up as "unranked" until it too can be removed.

I'm hoping that this change will eliminate a lot of the negative behavior that we've been seeing/talking about lately. Most of the complaints centered around someone "cheating" their way above other players for one reason or another. Since such players no longer are rewarded for that behavior, hopefully it will decrease. Since there's no explicit pecking order based on Elo and thus no advantage to artificially boosting it, the cheating/dishonest/ragequitting/ragespeccing players are left cheating only themselves.

Do note that Elo calculations are still enabled for use in various places (the pickup bot, for example). This means that you can still do comparisons manually if you want, but I'd really encourage people to heed Halogene's advice in that regard: treat those numbers as rough estimates.
asyyy^ | are you releated to chuck norris?
Reply

Sad that you removed it. I found it interessting to see who are the top players in their category. Maybe some people, who are good in math and analysis can make suggestions how it could be calculated better. Where can i read how it is calculated?
Reply

Woha. I surely didn't expect the ranking to be disabled. Was there so much obsession of players with their ELO values that the public ranking had a noticably adverse effect on the gaming experience? I didn't notice that myself, but then again I am not that often on the servers anymore, sadly. If ELO obsession has become a problem that leads to games being ruined by ELO whores, then removing the ranking display is probably a good decision.

* Halogene thinks a bit more about this

I very much value Xonotic being a game that does not utilize the usual addictive techniques (such as progressive achievements that take longer and longer to complete) and instead focusses on an enjoyable gaming experience with the only progressive development being a player's personal skill. ELO ranking indeed provides the option to farm ELO points in order to progress in the ranking, which is a totally different incentive to play the game than what Xonotic originally was made for (which is to have fun and to compete with other people).

So all in all, I think removing the ranking emphasizes how Xonotic aims at delivering a fun gaming experience and at providing a competitive platform, as opposed to satisfy people's desire to harvest virtual achievements. I hereby declare that I fully support this decision. Heart
My Xonstats Profile
Latest track on soundcloud: Farewell - to a better Place (piano improvisation)
New to Xonotic? Check out the Newbie Corner!

Reply

I understand your decision, Antibody.
Anyway, with all the different mods and communities in this game, the rankings didn't really mean a lot. I saw players on the Top 10 that I never met in-game (probably because they're American or Australian or insta players or whatever).

If people want some competition, they can still participate in the tournaments that are organized by the community.
Reply

It was a tough decision, but in the end I think it was the right one for our ecosystem right now. As mentioned, the Elo numbers are still present so people can still get an idea of another player's skill. It just won't be along the lines of "this person is in the Nth percentile of all players". With multiple mods, mutators, regions, and player cliques that one number became less and less reliable in an "overall" sense anyway. It still remains useful for comparisons within those cliques, and for that I decided to keep it.

None of this prevents implementing ranking using a better system in the future. I'd be more than happy to do that when certain things are addressed. Off the top of my head those things are as follows:

- Account for quitters. Those who leave the game after a certain percentage of it has been played should still be subjected to points changes based on their performance.
- Account for spectators under similar (but the not same) circumstances as the above.
- Identify separate config categories so we can compare players with higher confidence (e.g. don't mix insta skill numbers with vanilla or overkill).
- A points system that gives immediate feedback (i.e. after matches concludes) but also takes multiple observations into account. Something similar to how OGS does it.
- A whitelist-centric approach where participation in ratings is opt-in at the server level.
asyyy^ | are you releated to chuck norris?
Reply

I also support this change. Yes, it's a pity that we lost the leader board, which was one of Xonotic's features. However, I do think the benefit that we won't have players gaming the scoring system outweights it.

There are certain players and servers I avoid, because some players on those servers always quit whenever they lost. By doing that, such players not only avoid negative scores, they also deprive others of their winning rewards. That kind of behavior really poisons our culture.

I personally don't care about my scores, but I have little interest playing with cheaters.
Reply

@antibody, In Frag Matrix, I think we do have a meaningful value for the diagonal line -- the times when a player kills oneself. Can we put that number in, too?
Reply

@BFG: We do not have that information on a player-per-player basis. Sorry! If someone were to add it, I'd certainly make the corresonding changes to show it.
asyyy^ | are you releated to chuck norris?
Reply

(01-21-2018, 01:30 PM)Antibody Wrote: @BFG: We do not have that information on a player-per-player basis. Sorry! If someone were to add it, I'd certainly make the corresonding changes to show it.

When a server submits the scores of a game, does it not include suicide counts (which is shown in the in-game scoreboard)?
Reply

Ah, you're totally right. I don't know what I was thinking. We are indeed getting that value, but it's not part of the "frag matrix" per se. It's sent separately. I can see about injecting it into the frag matrix data structure, though.
asyyy^ | are you releated to chuck norris?
Reply

But it is already shown in the scoreboard anyway?
Reply

It is indeed!
asyyy^ | are you releated to chuck norris?
Reply

Hi folks,

Talking with folks on IRC I realized how half-hearted my JSON responses were in XonStat. During this year's holiday break I addressed the situation. The result is now a much more consistent approach to JSON for all your data-grabbing needs. For those who prefer content negotiation via HTTP headers, take the regular URL you see in the browser and request it with the "Accept: application/json" header. If you're in a browser or who don't want to use headers, simply append ".json" to the URL to get a JSON-formatted response. The following views now have this functionality enabled:
  • /players
  • /player/<player id>
  • /player/<player id>/games
  • /player/<player id>/captimes
  • /games
  • /game/<game id>
  • /servers
  • /server/<server id>
  • /server/<server id>/topmaps
  • /server/<server id>/topactive
  • /server/<server id>/topscorers
  • /maps
  • /map/<map id>
  • /map/<map id>/topscorers
  • /map/<map id>/topactive
  • /map/<map id>/topservers
  • /summary
  • /search
Please note that not all of these are fully implemented, but those should be easy to identify
with their "not implemented" JSON responses. I hope to address those in due time. I'm sure some other "ergonomic" features can (and should) be added here as well. I look forward to the conversation.

Happy holidays, community!
asyyy^ | are you releated to chuck norris?
Reply

What sort of algorithms and formulas does XonStat uses for calculating elo? Is there a legible documentation for it somehwere? I would like to know them so that i ccould implement a ranking system between my friends. (Yes im ranking my friends using maths)


EDIT: Ive found this : https://xonotic.org/posts/2012/much-ado-about-elo/
[Image: 0_e8735_c58a251e_orig]
Reply

There's also another blog post: https://xonotic.org/posts/2012/much-ado-...part-deux/ and the code itself: https://gitlab.com/xonotic/xonstat/blob/...tat/elo.py (300 lines of pretty readable python)
[Image: 30381.jpg]

<packer> when i see martin-t's name my blood pressure increases

<[BOT]Hоtdоg> anyone here lives near martin?
<[BOT]Hоtdоg> will pay monies for shooting him
Reply

@aa - for ranking you could just grab all of  your friends' info via their player_id and sort them by their elo values in descending order. Here is an example URL for fetching that information for a random player I clicked on (Kurtz). No need to re-implement the algorithm if you don't want to. If you want to for educational purposes, that's fine too Big Grin
asyyy^ | are you releated to chuck norris?
Reply



Possibly Related Threads…
Thread Author Replies Views Last Post
  XonStat turns 1! Antibody 13 12,028 10-25-2012, 03:36 AM
Last Post: harry_ftw
  XonStat FAQ Antibody 2 4,743 04-20-2012, 10:30 PM
Last Post: Antibody

Forum Jump:


Users browsing this thread:
1 Guest(s)

Forum software by © MyBB original theme © iAndrew 2016, remixed by -z-