[OFFICIAL W:ET] Discussion new rating method

[giriel]

That seems nice.
Too summarize:
- Add the team only comparison. -> Will need to insert some additional checks so that last minute team switchers aren't rewarded.
- Using those new xp values.
-> Light weapons will be :

WS_KNIFE, WS_LUGER,WS_COLT, WS_MP40, WS_THOMPSON, WS_STEN, WS_FG42, WS_GARAND, WS_K43.

These get 4 xp per kill, will lose 4 xp per death and will gain 1xp per headshot.
--> Syringe gets added with 4xp per hit.
-> Heavy weapons will be:

WS_PANZERFAUST, WS_FLAMETHROWER, WS_MORTAR, WS_AIRSTRIKE, WS_ARTILLERY, WS_SMOKE, WS_SATCHEL, WS_MG42

Get 3xp per kill.
-Take battle sense and engineering XP.
-Calculate the PPM's.

I have checked the ET source once more and they provide 5xp per headshot kill and 3 xp for a body shot kill with the so called light weapons.
The heavy weapons all get 3 except for the artillery(4xp) and satchel(5xp).

Therefor I don't think we'll need to change the battle sense or engineering xp since it will be of the same order of magnitude.
Although we subtract xp for deaths but I don't consider this as a major flaw for the moment.

If someone sees a major flaw or something that could be improved now is the time

[testforecho]

Seems balanced,

the only thing I can say is that with the old method deaths didn't give a penalty, so as it is now the light weapons part could gets less points if compared to the past, giving a higher weight to killing sprees (battle sense xp), which could also be obtained by camping, and heavy weapons + the rest.

But with 1 point per hs, 2hs per kill gives 6 points per frag, which is not bad and it could compensate for the deaths. You could add 0.5 or less points per (hits-hs) as well, but then there is the problem with adrenaline on (half damage) or off. I would say no but it could be an idea.

For the team switchers problems, can't the average score of team mates be taken on each sample, instead of all at once at the end?

I mean on every single sample you take, you store how many points a player did in that slice of time, and how many average points his team mates (those player who are in his team *at the time you checked*) have gained since the last check. Next check, those names could have changed if someone switched, but you are only taking a partial score, at the end you sum it all.

example (edited):

name(partial points gained since last check)

          Player | team mates average
sample1 | +10    | name1(+6), name2(+5), average +5.5 --> store (+10, avg. team +5.5)
sample2 | +15    | name1(+3), anotherguy(+20), avg. +11.5 --> (+15, avg. team +11.5)
---------------------------------------
total   | +25    | +17 average for his team mates

this way they could switch team at will, the only downside is that you can't show a clear endgame table with all this info for players to check the way their score was calculated

[testforecho]

If someone sees a major flaw or something that could be improved now is the time

Had to edit my previous post since I forgot to average team mates score, I was thinking about another potential problem,

Is the fact that scores can be negative ok? An average player who gets the same kills and deaths will get 0 points + headshots, someone who goes really negative in k/d could get a negative overall score, hopefully that won't be a problem when it comes to calculate score ratios...

Another possible critique of the system kills-deaths+hs is that it's based on the assumption that a score like 60/10 is worse than 100/30, which is not necessarily the case. That is, players are awarded for trying to fight as much as possible, quantity and not only quality (ratios).

A score like hs + fights*(k/d), that is heads + (kills+deaths)*(kills/deaths) would never be negative.

example:

     | kills deaths  hs   oldscore newmethod
pl 1 |  100   30    200     480     633
pl 2 |  60    10    120     320     540
pl 3 |  20    60    24     -136     51
pl 4 |  50    50    75      75      175

of course this new one has to be calculated at the end only, since you don't know the value of (kills/deaths) until the end... since I don't know what's 'behind the curtains' I can't say how much of a problem that could be in your code/method

[giriel]

that ain't a problem with the current implementation.
I would add the other xp values as well though.
Seems to come out nicely

[testforecho]

that ain't a problem with the current implementation.
I would add the other xp values as well though.
Seems to come out nicely

good, sounds promising

hope you are not putting first aid, signals and covert ops xp in the mix though, or ppl will keep 'whoring' again... if I were you I wouldn't put them in the score, not even reduced by a factor, otherwise ppl would still be tempted to round up their scores that way.

But I think that was already agreed, reading last posts. Only engineering and battle sense, right?

edit:
so, which method will you use? the 4xp per kill -4xp per death one? I think that is the best because the last formula I've put can give funny results if deaths are too few, not to mention divisions by 0

[.PG|monk]

As i am reading this topic i see a lot of good idea's to improve the current ranking system. But none of you haven't thought about the issue between the difference in mods. So if you want to do it good, you should add a factor as well based on the mod that is running on that particular server. In this case im referring to the unbalanced rating system between mods like Jaymod & NQ for example. Because we all know that top Jaymod players will not have such high rate when they would play NQ instead. This "could" be fixed by implementing a Factor for the mod.

Theoretical;

Score in Jaymod score 403 points (22.39 ppm)	18.45 min * 0,8 = 17,47 (ppm)
Score in NQ 432 (17.38 ppm)	24.85 min
This is a score from player number 1 in ranking vs a top player in nq with a rate around 20. (you won't have them a lot higher)

On this way you can approach a more equal rating system. Of course the factor is the point of discussion, thats why i said theoretical.
I would like to share opinions with you if we should add a factor and what factor.

Edit1
Some other solution which is possible: make a ranking system for the different mods. Like a ranking list for Jaymod (based on scores at Jaymod servers only) and a ranking list for NQ (based on scores at NQ servers only)

Kind regards,
monk

[giriel]

Yes well you already presented the problem
We would need an elaborate collection of differences between the 2 mods which would require either the creators to help us out or the source being available.
I am afraid neither of those options are available at the moment.

For your second suggestion : that would be hard on the database.
We don't have the room to do it separately.

EDIT: testforecho to make my final decision about those formula's I'll create a matlab script to see how they will perform on whatever values I throw at them. But haven't got the time yet to do that

[testforecho]

It's hard to have a rating system which takes server differences into account. One possible way would be to have a server rating factor, and assume that if 2 ppl with the same colored name are playing in 2 different servers, then it's the same person, so their rating must be the same in both servers, for example:

EtplayerX plays on jaymod and he owns making scores 2 times better than the rest, then he plays with the same name in some etpro server and he gets his behind kicked, then the factor for that server would go up and the other down, with players getting a lower rating there. Eventually his points will become equal in both servers due to that factor.

I guess there would be lots of whining then plus you can't be sure if 2 different people are playing with the same name...

Anyway the difference in difficulty between different servers doesn't lie much in the mod used, but mainly in the players who are regulars there, and server settings which suit the taste of beginners or more experienced players, plus some unwritten rules in some servers (like, if one is a total noob he will be sometimes voted for kick, then the average server skill goes obviously up).

In my opinion it's all about this: you set the rules and the score for a game, then players will obviously figure out how they can get a higher score, and the solution could even be playing against people who are worse.
I think that stopping xpwhores and comparing points vs teammates only would be already a huge step forward. For the rest, it's a game, who is able to play better *that* game will get more points. That doesn't mean they can do better at other 'games' but if it takes skill, skill is still skill.

@giriel:
well for the other formula, (the one with kills/deats) the problem is that scores can skyrocket if someone manages to have only 1 or 2 deaths, let alone 0, it would be a division by 0.
If I collect some more data with /statsall I could try other formulas myself, but I am sure that you know what you're doing

[giriel]

Ok I have ran some tests and actually got the values that I expected.
If you are interested I can show you the graphs

So indeed your formula rises exponentially, that is because if you work a bit with that formula:
heads + (kills+deaths)*(kills/deaths) you get this :
heads + (kills²/deaths + kills)
That square is annoying so I removed it.
As a result I get : heads + (kills/deaths + kills)
And this does give me nice results and it doesn't have the problem of going negative(which is risky).
Perhaps not fully as you had intended but now we have no problems with mathematical behavior.

now we have to make some changes to the database and I'll start implementing this.

[carnage]

Ok I have ran some tests and actually got the values that I expected.
If you are interested I can show you the graphs

So indeed your formula rises exponentially, that is because if you work a bit with that formula:
heads + (kills+deaths)*(kills/deaths) you get this :
heads + (kills²/deaths + kills)
That square is annoying so I removed it.
As a result I get : heads + (kills/deaths + kills)
And this does give me nice results and it doesn't have the problem of going negative(which is risky).
Perhaps not fully as you had intended but now we have no problems with mathematical behavior.

now we have to make some changes to the database and I'll start implementing this.

Wait, so this new system is going to be mainly headshot-based? Not that I can complain (my hs:kill-rate is ~2.5:1 on Jay/ETpub ), but I am just wondering. Looks cool.

Aand, btw, I hope you implement this as soon as possible. gj