Public Statistics

[Static Burn]

the stats haven't been updated for a few days it seems, looks like it stopped on december 5th as there's no "new players" since then.

[Kontrolz]

So what's going on with the public statistics? Is it being dropped or worked on? Are people just too busy at the moment working on other things to get around to it?

I see the page is still up but the statistics haven't been updated in like 2 weeks. Just wondering if we're going to lose this great idea.

[roxxkatt]

it is a good idea, and would be even better with usage rankings =(
lets hope it doesnt die.

[Maverick]

I'm sorry the public statistics have been down for some time.
It went bad one time and spammed our bot logs until qan turned it off. Ever since it hasn't been turned because of the fear of it happening again.

Unfortunately my computer (harddisk) crashed at the end of 2008 so I was unable to fix the public statistics until now.
I just fixed the public statistics and reset the scorereset period. Everything should be back up and running now.

[Kontrolz]

Um, awesome :turned:

[Crown of Thorns]

Good job.
Thank you

[Maverick]

I had the formula in a ?log

Kontrolz> what is the formula for rating anyway?
BH822> let me check
Kontrolz> ok
BH822> R = (points from kills x 10 + (wins - losses) x 100) / (wins + 100) kills

That's almost the same formula as I'm using for the pubstats (apart from the "kills" at the end).
It's obvious that the rating formula is off as "24" is at the top now with 0 - 516 in score.

I need a better formula then the following which I'm using:
Rating = (points from kills x 10 + (wins - losses) x 100) / (wins + 100)

... or maybe someone with a negative score difference just needs to have Rating = 0

[Zingy]

That's almost the same formula as I'm using for the pubstats (apart from the "kills" at the end).
It's obvious that the rating formula is off as "24" is at the top now with 0 - 516 in score.

I need a better formula then the following which I'm using:
Rating = (points from kills x 10 + (wins - losses) x 100) / (wins + 100)

... or maybe someone with a negative score difference just needs to have Rating = 0

Hey mmav, i think you're right. I did some experimenting and it does seem that negative values = 0.

I have adjusted the formula a bit and the numbers I am getting are now matching the rating in the game. Here is my modified formula:

davinci's modified rating formula:
Rating = ((<points from kills, not flags>*10+(<wins>-<losses>)*100)/((<wins>+100)*100))*100
Where negative ratings = 0

The part I have adjusted in my formula is the last part, with the two extra *100

I think the reason pubstats is calculating 24's rating wrong is because it shows 28400 as his "kill points" when in reality he has gotten no kills, and should have no kill points. So probably pubstats made a mistake and labeled his flag points as kill points?

If 24's rec is 0-516 then he has no kill points, which, using my formula, would result in:
((0*10+(0-516)*100)/((0+100)*100))*100 = -516
We should assume that rating can not be negative, so 0 is the lowest ratings can be in TW.

I believe this modified formula is correct and what is causing the rating issue is a labeling mistake between "kill points" and "flag points."

I played in pub on a private freq so that i would not get flag points and used my modified formula, and the rating I calculated matched what I had in game.

For example, here are the numbers I used from my experiment in pub:
Wins: 3
Losses: 3
Points from kills: 441
Points from flags: 0 (i was on a private freq and did not get flag)
Rating = ((441*10+(3-3)*100)/((3+100)*100))*100
My formula calculated 42.8
In game rating: 42 (no rounding)

[Maverick]

Thanks for your work, Zingy. Maybe 24 didn't have the correct flagpoints / killpoints what made things go wrong.

However I don't understand why your formula works better then the one I'm already using? I'm checking both now and both seem to generate the correct Rating (apart from negative Ratings). I'm sorry if I'm missing the point though.

[Maverick]

I finally found out what is wrong with the pubstats!

A small error was made in the protocol - the packet that's supposed to update the bot (Server to Client) of the latest scores (including flagpoints and killpoints) of players. The protocol made a mistake with this packet and reversed the flagpoints and killpoints. TWCore thus thought flag points were kill points and vica versa.
I checked the source of Mervbot where this mistake wasn't made.

On the next grand bot update this patch will be applied.

Unfortunately I have to decide that all previous score reset periods and achievements will be wiped.

[Maverick X+]

ah well, there goes my rec of being first for just about everything on lev then lol

[blackrazor]

Nice job! Thanks for putting the work into this project.

[Maverick]

Some statistics of previous scoreresets were removed:

- Highest rating
- Highest average

The Most Flagpoints and Most Killpoints results have been swapped.

[Maverick]

I noticed that the rating in the public statistics still wasn't matching the rating in-game. I feared that the flagpoints and killpoints was swapped at another packet aswell.
When checking I indeed found another packet that had the flagpoints and killpoints swapped - this was the packet that's received when a player enters the arena.

I repaired this bug and now it seems that the rating in the stats is matching the rating in-game. Finally we can see who is the best pilot! ^-^

[Kontrolz]

Nice work mav, way to stay on top of it and get this thing working.