A2A Online Stats

[zorlac]

EDIT: Looking at the table more careffuly it seems to me that you implemented carreer ranks after all, but I think you made too many carreer ranks. I should have been in the same as Goya and Ice for example(I was actualy enjoying climbing the ladder). Its not fair if they cant attend as much as I do (again I dunno if this is riguorously true, but we should give the benifit of a doubt with a certain amount of frequency---->) 4 or 5 carreer ranks shuld be enough, I dunno if you should spread them linearly or in a logaritmic kinda scale. Because all we want is to prevent sporadic high scorers.

 

Career ranks? eh? I havent done that yet.... the K : D stats are sorted by the K : D ratio, and the "score" stats are sorted by the "score"... no other sorting is done

 

There is a HUGE difference between what a ratio should mean and a score. Ratio is an average over their playing time, and thats what is shows. A score is a running total since a predefined time. People who are good end up at the top of the K : D page, people who are good and play a lot end up at the top of both.

[Pilotasso]

The items on the previous page:

 

- Keeping the stats "current" with expiry on them

 

Still working on the best way to do this. Two options... well three actually.

 

1 - ignore EVERYTHING older than xx days. Pilots would need to be on the server quite regularly to keep up their stats, as the older stats "expire". Go away for a week or two on vacation, and you will plummet down the "score" table.

 

2 - for pilots who have flown in the last xx days, remember ALL their kills and deaths from the start of the log. This has the side effect of making it hard for new guys to climb the "score" table. I also have the problem of deaths of pilots who have "expired" from the stats. If Piltasso kills Numptie, and numptie never comes back, after xx days do I ignore Numpties kill in Pilotasso's stats? Or do I count it, making Numptie's stat sheet look terrible because I am ignoring his kills (time expired), but showing his deaths by pilots who havent time expired.

 

- Sorties

I am unable to count the number of "sessions" or sorties the pilots have. Do you mean the number of successful take-off/landing cycles here? (rather than take-off/kill/crash)

 

- K/D ratio vs "Score"

K/D sorted stats still exist as a seperate page, I havent replaced them with the a "score" based system. The "score" based sorted page is in addition, so both sets of ranks can be displayed.

 

 

Otpions 1 and 2 are a bit hard, when someone needs to be absent for vacations, and when they return all their hard work is gone. It is a bit demoralizing. If it means to be hard to climb the ladder, so be it. Its more fun anyway.

 

"Sorties" was a way of puting things simple. Are you able to determine when and how often the server was up and how many times the players apeared on it? (by dates) IMHO we should play with that. A good score is always a good score in any time frame as long as the pilot that did it had a credible carreer. His record should stay for others to try break it (like the olympic records LOL). This is another reason not to do options 1 and 2.

 

EDIT:

Career ranks? eh? I havent done that yet.... the K : D stats are sorted by the K : D ratio, and the "score" stats are sorted by the "score"... no other sorting is done

 

yeah a career is the best sorting system IMHO. It would give players their K/D deserved score ranking while rewarding primarily those who actualy have stable career but a sporadic apearence on the server.

[zorlac]

"Sorties" was a way of puting things simple. Are you able to determine when and how often the server was up and how many times the players apeared on it? (by dates) IMHO we should play with that. A good score is always a good score in any time frame as long as the pilot that did it had a credible carreer. His record should stay for other to try break it (like the olympic records LOL). This is another reason not to do options 1 and 2.

 

I can count how often an IP address joins the *SERVER*, but there is no way of being 100% accurate about tieing up that IP address with a player. Also, many people have a dynamic IP address, so that doesnt help. If ED implemented a system in the mp_log to tie a player name to an IP address, that would make life easier to count how many times a specific player joined the server.

 

The mp_log looks something like this:

 

accepting connection from 1.1.1.1/1609

append: client 1

accepting connection from 2.2.2.2/28001

append: client 2

spawning: client 1

spawning: client2

client[1] started

client[2] started

client[1] is ready to start

client[2] is ready to start

(game event) 01/13:00:51 RED "Pilot_1"(Su-25T) entered the mission

(game event) 01/13:00:54 RED "Pilot_2"(Su-25T) entered the mission

 

You would think this would be easy to work out, and you would assume that client 1 is "Pilot_1"...but often its not... with the differences in network speed and pilot's own PCs, the time diffrerence between hitting fly and being in-game (the "entered the mission" entry) can mean that sometimes its client 2 who appears in the game event log first. This happens a lot, especially when the server gets busy.

 

I can count how many times a player joins the *GAME*, but that would also include getting killed and rejoining, or leaving to change payload or aircraft and rejoining, even though they never actually disconnected from the server.

[Pilotasso]

Can you sweep the log files ^^^^ to determine if the player "x" apeared (true or false, doesnt matter exactly how many times) on a time frame established by you (like 24 hours) and then count how many "trues" it returns for sorting his rank on the table.

It would verify if the player was there in a day, and how many days he made a presence, since the absolute number of times per session doesnt exactly tell how much play he actualy did. Rather that he WAS THERE. I dont think much more is needed.

 

Aditionaly the value "true" should be only returned if he scored anything, to avoid ghost presence cheats.

[zorlac]

Can you sweep the log files ^^^^ to determine if the player "x" apeared ...

/quote]

 

mm... probably... /me looks at code again...

 

Aditionaly the value "true" should be only returned if he scored anything, to avoid ghost presence cheats.

 

oh man, you want blood dont you :) I`ll see what I can manage to cobble together, but dont hold your breath on this one...

[Pilotasso]

I had an alternative I dea to simplify all this. Hopefully it would cut the amount of time needed to figure out the above methods^^^^

It doenst track careers directly but rather indirectly.

 

for your algoritm:

1)check the player who has the most kills.

2)Who has the amount kills corresponding to at least 1/5th of that^^^^ (ratio between 1 and 1/5, or 20-100% of that) is placed at the top table

3) who has betwen one 1/5 and 1/8th is ranked to the second table below.

4) Who has between 1/8th and 1/10 on the next table.

5)last table comprising scores of 1/10 of the highest score to the last player, for the last ranking table.

6) rank all pilots in each table with K/D score.

7) refresh table in every 6 months so that no pilot becomes "untouchable" due to abnormaly high kill records.

8)keep a history table ranking all "pilots of the semester" (top 20 K/D of all times?).

 

This would push all players to keep up with each others careers within a minimum acceptable frequence. No longer it would matter when and how often they apear but simply how good could they match and update they scores. it gives plenty of freedom to apear when they like as long as they score well.

 

oh man, you want blood dont you :) I`ll see what I can manage to cobble together, but dont hold your breath on this one...

 

LOL not realy, Its just that we have to filter the current tables to figure out the real pilots measure such as ignoring all the perfect kills and skip to finite K/D directly to see whos the best. This is the way we could do that without cheating on how people had managed their scores untill now.

[Floyd]

I'm really curious which formula will finally determine any ranking. Looking at the

actual scoring one could say that a TK should weigh less for a pilot with - e.g. -

a 5:3 KD (a newbie) than a 150:20 (experienced, veteran pilot). Or why not

setting pilot's stats to zero with each TK? Or apply the LomacLeague ranking

system and consider who shot down whom (points depending on position in

the list).

 

Until then:

 

"People can come up with statistics to prove anything;

14 percent of people know that." - Bart Simpson

[Pilotasso]

I'm really curious which formula will finally determine any ranking. Looking at the

actual scoring one could say that a TK should weigh less for a pilot with - e.g. -

a 5:3 KD (a newbie) than a 150:20 (experienced, veteran pilot). Or why not

setting pilot's stats to zero with each TK? Or apply the LomacLeague ranking

system and consider who shot down whom (points depending on position in

the list).

 

Until then:

 

"People can come up with statistics to prove anything;

14 percent of people know that." - Bart Simpson

 

A newb naturaly is less skillfull than a veteran. There are veterans with as much as a 1/3rd only of the highest kill board. Since we are measuring skills by numbers why would we give free bonuses to whom havent the skills yet?

That way wou could see some noobs unrealisticaly overtoping some veterans. I would hate to see that.

 

I have yet another crazy idea. Zorlac, if you realy lack the time just make an average of table position number between the 2 ranking systems you have and make a another table ranking that average number (ommitting it of course).

[zorlac]

I have yet another crazy idea. Zorlac, if you realy lack the time just make an average of table position number between the 2 ranking systems you have and make a another table ranking that average number (ommitting it of course).

 

do you mean:

 

( K : D ratio + score) / 2

 

or position rank down the table

 

( K : D rank + score rank ) / 2

 

It really doesnt matter to be honest... no matter which way i do it, or how the stats are calculated, or how long we discuss complex ways of making it fair, someone somewhere will lose out. There is only so much time I have to do this, and I have another project coming up soon that I want to work on quite urgently....

[Pilotasso]

Of course someone will loose out, the single perfect kills will go down the table. I guess all other positions will be more or less the same.

 

I meant like this: (position number on table 1+position number on table 2)/2= average

 

Then Rank the pilot based on that number (i.e "average" is an abstract average of the 2) the higher the better the position in the final table.

 

This is because on the first table who wins is usualy those who have scored the kills and no losses (regardless if its just 1), while on the second it only counts how much kills you got (i.e how long you were online to get that score) Having an average of the 2 is a balanced ranking of the K/D VS time playing.

 

This would save you alot of time. Calculations are very simple.

I made a test with this in excell. I chosent the first 9 pilots of the AA kills and ranked them with the wieghted average, heres the results:

 

 

Table:.....|Kills|K/D|average

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

Pilotasso.|1....|46.|23.5

Goon......|2....|47.|24.5

Ice........|3....|39.|21

eagle.....|4....|40.|22

frostie....|5....|71.|38

Goya......|6....|38.|22

Mustang.|7....|57.|32

Breakshot|8....|92.|50

kevlon....|9....|88.|48

 

New ranking:

Ice

Eagle and goya

Pilotasso

Goon

Mustang

frostie

Kevlon

breakshot

 

Note: I ommited some names from the K/D table because I only choosen 9 of the Kills table in that order, so there are more people between the averaged rank above. On the other side all the single perfect kill pilots are so down the Kills table that the weighted average is much bigger than any of the above. What hapened is that they got filtered out of those who have proven themselves (wich again some are not listed above)

[S77th-RYKE]

Using "weighted" calculations is just lying to yourself IMO , if those guys that showed up once got a perfect K/D ratio so be it , they earned it , they weren't killed were they ? It was the truth as it happened . Can't get any better than the truth .

 

Looking at those stats we can make our own opinions of the real meaning , let those guys show up again and try to maintain it .

 

I so hate the banks that use "weighted calculations " to screw you over interest on mortgages ... ohh BTW the firts 10 years you only pay interest and your capital doesn't go down ... now bend over ;)

[504Goon]

Well.. i personally wouldn't like to see a pilot with a "perfect" K/D ratio done by a single a2a victory above someone with 100+ kills and lets say 15 kills, for example.. Of course, we can all do our own calculations based on the stats displayed, and in the end, that might be the best solution, as you can never satisfy everyone on how the info is displayed...;)

[Pilotasso]

Using "weighted" calculations is just lying to yourself IMO , if those guys that showed up once got a perfect K/D ratio so be it , they earned it , they weren't killed were they ? It was the truth as it happened . Can't get any better than the truth .

 

Looking at those stats we can make our own opinions of the real meaning , let those guys show up again and try to maintain it .

 

I so hate the banks that use "weighted calculations " to screw you over interest on mortgages ... ohh BTW the firts 10 years you only pay interest and your capital doesn't go down ... now bend over ;)

 

If a noob hapens to kill one guy (even if its an A-10 or -Su-25) and never returns, who do you think is kidding himsel by believing hes any better than a pilot with over 40 kills and a K/D of 8 or 9?

 

Me and many others will figure out by then, that particular single kill pilot will surely get creamed VS any serious oponent. At the same time we have to filter out the "clutter" in that table to know the real extent of the ladder to climb. What it will hapen is that the higher scorers will tend to ignore those single kill pilots to measure each other. This is what is hapening with at least 20 pilots on top of the K/D table (respectfully saying of course).

 

If a single player achieves only 1 AA victory and never wants to return again, then hes indirectly admiting hes not THAT good.

[Pilotasso]

Well.. i personally wouldn't like to see a pilot with a "perfect" K/D ratio done by a single a2a victory above someone with 100+ kills and lets say 15 kills, for example.. Of course, we can all do our own calculations based on the stats displayed, and in the end, that might be the best solution, as you can never satisfy everyone on how the info is displayed...;)

 

If a pilot gets 9-nill kill ratio on one flight, its alot different from a guy with only ONE. While its not quite representative, it does count for something though. A weighted techninc as I demosntarted will put him in a very good position but not as good as a guy with more than 40 kills and with 8-9 k/D.

[S77th-GOYA]

Why not have a table for "Active Pilots" that have a kill or death within the past 7( or whatever) days. That will weed out the one kill wonders, promote the (504) server, and if someone is innactive and drops from that table, all they have to do to get back on it is fly and fight.

[Pilotasso]

As it has been said before, imagine you go on vacations for a couple of weeks, what hapens to your scores? ;)

[S77th-GOYA]

Nothing. It just moves you off the Active Pilots list. When you get back, your first kill/death puts you back on the Active Pilot list.

[S77th-GOYA]

I hate to say it, but warping has become a problem on the (504) server. Do you think that perhaps the number of clients is set too high? It doesn't seem to a problem of pings.

[Kevlon]

I think its good enough as it stands ;D no need to make more work on the stats , as zorlac has done a great job allready!!

[zorlac]

I hate to say it, but warping has become a problem on the (504) server. Do you think that perhaps the number of clients is set too high? It doesn't seem to a problem of pings.

 

Maybe the server needs a kicking, but doubtful... when did the warping start? any ideas?

 

CPU 0 use sometimes hits 100%, but not that often:

http://inc-mon.syn-ack.com/mrtg/vvs504-lomac-cpu.html

 

Its a dual CPU box (2 x 1.7GHz Xeon), with Lomac set to run on only one of them. Its the highest spec "spare" box I have that will allow me to fit a decent gfx card, all my high spec dual and quad CPU boxes are rack mount, with onboard gfx, and Lomac refuses to work on them (PLEASE GIVE US AN OPTION TO "HOST ONLY" SO WE CAN HOST ON A BOX WITHOUT THE NEED FOR A HIGH END GFX CARD!!!!!!!!)

[Pilotasso]

I think its good enough as it stands ;D no need to make more work on the stats , as zorlac has done a great job allready!!

 

Of course, his efforts are highly commendable ( I wish I could do the same), I dont want my suggestions to be misunderstood as demands. I Spilled out all what I wrote just because the question was risen by him. This is what I call braistorm, and its always positive. I even came up with work friendly alternatives. It is up to him to opt.

 

BTW rep inbound! (I dont usualy give it any importance, but Its always a gesture of recognition) :)

[Pilotasso]

Nothing. It just moves you off the Active Pilots list. When you get back, your first kill/death puts you back on the Active Pilot list.

 

While your idea is as good as any exposed here, think this:

 

Ice goes on vacations to my country for a a month and his score migrates to another table. Since your currently disputing the top dog award with him, you would have to compare both tables to keep climbing the ladder. It can get confusing and doesnt make much sense (at least not in my point of view), to guess about his position or anybody else's from 1 table to the other.

[Floyd]

Maybe the server needs a kicking, but doubtful... when did the warping start? any ideas?

IMHO it started when pilots began to fly for the stats :cool:

[Pilotasso]

Not realy, It has always happened. People just noticed it more since its their reputation at stake. ;)

[zorlac]

I`ll be leaving the code for the stats alone for a while, as I need to take a break from that project. I will (probably) return to it at a later date.... no ETA for resumption of work on the stats.

 

I also have some hardware arriving some time next week that requires a lot of coding for a project I have been planning for a few months, so that will take up some of my spare time for the foreseeable future, but not as much as the stats... hell, i might even get to fly on my own server occasionally.