Xonotic Forums

(08-07-2013, 02:36 PM)Mirio Wrote: [ -> ]You only get into the stats if you joined the match until it reaches the "endmatch"-event, which is i.e. timelimit runs out, fraglimit reached, vcall endmatch, endmatch enforced.. etc.

Exactly, and that means you could either join a match, then idle until endmatch to make the match show up on xonstat for your playerid; OR drop out of the match (spec) just before endmatch to avoid the game from showing up. The latter is performed by way too many people who care too much for their elo, but there's not really much we can do about it...

Also note that a match only shows up for your playerid you only get an elo rating if you participated for at least half of the match. That effectively means you have to be there for at least the second half (until endmatch) to make the match count.

Ahh, okay, that's cool. Which actually brings me onto a suggestion - showing the length of time individual players spent in the match. Could that work?

And about the specing just before the endmatch thing, so you'e saying that even though you could have been in the match for 99% of the time, the very second you are specing while the match ends, you don't show up in the stats but still get an elo rating?

If that's the case then surely you can force it so that they still show up if they have played more than half.

(08-07-2013, 03:45 PM)rocknroll237 Wrote: [ -> ]Ahh, okay, that's cool. Which actually brings me onto a suggestion - showing the length of time individual players spent in the match. Could that work?

And about the specing just before the endmatch thing, so you'e saying that even though you could have been in the match for 99% of the time, the very second you are specing while the match ends, you don't show up in the stats but still get an elo rating?

If that's the case then surely you can force it so that they still show up if they have played more than half.

You mean like showing the time one played in the match (which we call alivetime) on the scoreboard to compare between players? I'm sure that would be possible because that value is already sent to xonstat and stored in the DB (it's also used for elo computation). I'm not sure if that info should go on the scoreboard though, might be too much information. But I will talk to Antibody and see what he thinks about it =)

As for the "spec before endmatch" problem, it can't be solved *that* easily. And it was discussed a few times before, you're by far not the first to notice..
To keep it short, the problem is that all the game info is only sent to xonstat once the match ends, and if a player specs, the data will be removed from the game-internal storage on the server. Therefore, xonstat will just not receive any data about players missing on endmatch. I guess to fully solve this problem, the code would need a larger rework.

Okay, that's cool. Thanks for the response.

In the past we used an absurdly conservative decay rate on Elo values for inactive players: 1 point per day past 30 days of inactivity. This rate led to a vast number of inactive players retaining their ranks long after they stopped playing a game type. Starting today, I'm accelerating this rate a bit.

The new rate will deduct points based on how many *weeks* past 30 days of inactivity you have. On your 1st week past the cutoff - starting on your 31st day of inactivity - you will have 1 point removed per day. On the second week past the cutoff, 2 points per day. This scheme continues decrementing your Elo totals in this fashion until you either play a recorded game or you reach the Elo floor of 100 points.

If you have any questions, do let me know! Happy fragging everyone.

(08-20-2013, 06:37 PM)Antibody Wrote: [ -> ]The new rate will deduct points based on how many *weeks* past 30 days of inactivity you have. On your 1st week past the cutoff - starting on your 31st day of inactivity - you will have 1 point removed per day. On the second week past the cutoff, 2 points per day. This scheme continues decrementing your Elo totals in this fashion until you either play a recorded game or you reach the Elo floor of 100 points.

Nice, it will be interesting to see how that works out (my guess is: it will work quite well). Players who haven't been active for a full year will use 52 points per day, that's pretty fast
But it's still applied to all gametypes in the same way, yeah? So if I play e.g. one CTF game after a few weeks, elo rot will also stop for DM and all the other gametypes?

No, elo rot only stops for the certain gametype which is alright.

Awesome news!

This sounds good to me. Now we need to make Elo a frikking lot better as well. Cause the elo system, in all honest opinion, is quite sucky. It only takes into account aggressive people, but if you play smart and on the defence, then you will lose elo not gain, even if you won. Thinking mostly on team games such as TDM. Not sure how it works on 1on1 as I rarely play that.

I love being more defensive in team games but if I am I lose elo. But if I am aggressive I lose cause I don't think of protection just run into enemy fire, being afraid I will lose elo if I don't hunt people down. So yeah, I more or less don't care about elo, if it was nicer and smarter system then it would be much more fun.

Your comments should be directed at the scoring system rather than Elo. What you are really complaining about is the lack of rewards (read: points) for defenders in the game. Since Elo is based on score per second, those activities that help your team but aren't rewarded directly with points will have negative impacts Elo-wise.

This is nothing new. Players aren't rewarded for:

- Fragging an opponent near a certain radius of their own flag (note: waiting for someone to take the flag and then fragging them gets you more points, but is completely counterintuitive and also risky)
- Fragging opponents near a certain radius of their flag carrier
- Fragging an opposing flag carrier in their base, before they are about to cap
- Being a part of a flag carrying "chain" (last person gets all or at least majority of the points)

Now, outside of CTF, why are you being defensive? You probably shouldn't be defensive in any of the other major game modes, as it is rather counterproductive.

(09-03-2013, 07:42 AM)Antibody Wrote: [ -> ]Now, outside of CTF, why are you being defensive? You probably shouldn't be defensive in any of the other major game modes, as it is rather counterproductive.

It's quite important in TDM (which he talks about), because the Net (kills-deaths) matters more for the result in the end than just the amount of kills.
You can still get a load of kills (say 40) but die 50 times at the same time, which is actually worse than a player with 20 kills and 10 deaths. However the system would reward the players with 40 kills more, cause it checks the score. But the value for the team is higher by 20/10 player sorta.
There are other components though, where dying is legit so to say.

ai's point is that less deaths does not really get awarded with Elo, but the rush-mania (giving easy kills to the other team) does.

There could be sort of bonus points for least deaths or something. ;o

Okay, I understand that. My point still stands, however: it's the scoring system, not the Elo system.

I just wanted to throw it out there. I bet people already knew this but maybe not all. Shouldn't the elo system be based upon the scoring system, which means is should also take into account least deaths and so on.

(09-03-2013, 07:42 AM)Antibody Wrote: [ -> ]This is nothing new. Players aren't rewarded for:

(09-03-2013, 07:42 AM)Antibody Wrote: [ -> ]- Fragging an opponent near a certain radius of their own flag (note: waiting for someone to take the flag and then fragging them gets you more points, but is completely counterintuitive and also risky)

Whileyour statement is true, spamming electro from a safe position on the flag if you predict someone is coming, or if you follow the enemy and are not able to attack before he are near your flag you will get less points, which seems very unfair. This is probably bad for pickup CTF and cup CTF matches, even though in pub games your point seems to apply pretty well.

(09-03-2013, 07:42 AM)Antibody Wrote: [ -> ]- Fragging opponents near a certain radius of their flag carrier

So if you are able to frag the FC's defenders you get less points? o.o

(09-03-2013, 07:42 AM)Antibody Wrote: [ -> ]- Fragging an opposing flag carrier in their base, before they are about to cap

So if you are able to frag the enemy FC just before he captures the flag you get less points? o.o

(09-03-2013, 07:42 AM)Antibody Wrote: [ -> ]- Being a part of a flag carrying "chain" (last person gets all or at least majority of the points)

Can you explain this a bit more, I don't quite understand?

(09-03-2013, 07:42 AM)Antibody Wrote: [ -> ]Now, outside of CTF, why are you being defensive? You probably shouldn't be defensive in any of the other major game modes, as it is rather counterproductive.

Kojn is playing defensivily quite successfully and it's not counterproductive. In fact, lowering damage on SG and raise spawn times for weapons seems to be something many good TDM players want which basically means they want slower more tactical gameplay.

(09-03-2013, 11:52 AM)machine! Wrote: [ -> ]

(09-03-2013, 07:42 AM)Antibody Wrote: [ -> ]- Being a part of a flag carrying "chain" (last person gets all or at least majority of the points)

Can you explain this a bit more, I don't quite understand?

kuniu picks the enemy flag, runs through the whole map, unfortunately dies in front of his own flag, his teammate Mirio bravely picks it up and finishes the capture.
Result: Mirio gets much more points than kuniu.

i think that's what Antibody meant by a "flag carrying chain"

Aa, okey thanks!

machine, I'm not quite sure what point you are trying to convey. All of the things I listed provide no additional score points other than what you normally get for a frag. I was merely pointing out some things that people have wished for in the past - things that people consider to be of more value than just one point (i.e. one frag).

I'm very happy to show you what I've been working on for quite a while now: new weapon charts for XonStat! I've just pushed them to the server now, so let me provide a brief overview of what they look like and what you can do with them.

First off, here is what they look like:



For starters, you'll notice that there's a lot of weapons on the accuracy chart now. All the weapons you've used in up to the past twenty games, in fact! You can hover over certain points in the lines shown to see your accuracy for the given game. Unfortunately something is wonky with the tooltips, so it will say something like "48% at 12345." The "12345" part is the game number. I'll have to figure something out to fix those tooltips, but for now I think this is useable enough.

The next thing to note is the weapon colors along the top right of the chart. If you single-click one of them, you'll hide the line for that particular weapon. If you double-click one, you'll hide the lines for *all other* weapons. In this way you can slice and dice your accuracy data until you can see exactly what you want.

Best of all, this chart updates along with the game type you choose along the top of the page. If you click the "duel" game type, only accuracy values from duels are shown! Clicking the overall game type brings you back to the default display, of course.

The damage chart can be found just below the accuracy chart. This one shows all of the weapons you've used in up to the past twenty games as well, but the value along the Y-axis is the damage you've dealt instead of your accuracy percentage. It shows this in a stacked fashion, so the highest point on the Y-axis you touch for a given game represents the total amount of damage you've dealt. Hovering over a point on one of the bands will show you exactly how much damage you've dealt with a particular weapon.

This chart also has the same type of weapon legend in the upper-right. Single-clicking will remove a weapon's "band" from the visualization, while double-clicking will show only that weapon's band. I find this toggling extremely useful for analyzing how I've done with weapons in any given match. Furthermore, I can see some potential insight to be had by examining the weapon usage of your opponents in similar fashion .

The damage chart has an additional control panel in the upper left that let's you skew the visualization in three ways. I've found the default (stacked) view provides the most information at a glance, but the "expanded" view is also useful. This view expands the data into percentages instead of absolute damage values for the Y-axis, so it can show you how reliant you are on certain weapons from game to game.

Last but not least, the damage graph also updates its visualization when you click on the different game types at the top of the page. Woohoo!

Anyway, I've rambled on enough about these charts for tonight. You'll have to check them out yourself to see how fun they really are! Happy fragging.

WOWOWOW!!!!!

That's very impressive, amazing, I think xonotic has the best stats system by far now

Now I can watch in misery how my acc drops alongside my isp's connection quality(((((

Nice job again mate!

Thanks guys! I also want to mention that these are still in progress. While they are using a library (nvd3.js and d3.js if you're interested), they are malleable to a certain extent. I intend to give better numbers for the splash damage weapons as well as fix the tooltips.

Brilliant!

Wow that looks very good! Great job Antibody.


Howether, I have found something in the damage chart: minsta damage shadows all other damage, and hides it. Look at the image.

Finally I understand the acuracy graph! Thank you for this update. I love Xonstat

I am considering decreasing minstanex damage numbers by a constant factor so it doesn't skew everything else. I believe a hit minstanex shot is ~1000 HP, so that's why it is so incredibly skewed.