Introversion

jelco wrote:I've decided to follow TomCat39's suggestion and have changed the ranking system to only rank registered users.

There was me thinking my jump from #200 to #46 was through merit

Greeba wrote:

jelco wrote:I've decided to follow TomCat39's suggestion and have changed the ranking system to only rank registered users.

There was me thinking my jump from #200 to #46 was through merit

LoL, that would be something.

jelco wrote:I've decided to follow TomCat39's suggestion and have changed the ranking system to only rank registered users.

Sounds kewl. Now I just hope I'm correct in thinking it will be motivation to register......

/me runs from all the angry mobs as they chant, "We want TomCat39's head!"

jelco wrote:I've decided to follow TomCat39's suggestion and have changed the ranking system to only rank registered users. They still get assigned scores but will not have ranks and will not show up on the Leaderboard. Quoting my news post: "I remind you that this is experimental, so any change is still possible." Make of that what you will.

Jelco

Are clans possible?

Couple of quick questions/comments. I've been doing a little reading on Glicko.

It's showing "rating period: 2". Does that mean we are no longer essentially static?

I think your terminology "score" would be better off called "rating" to avoid confusion since score implies you are awarded points (or lose points) based on individual games, whereas glicko computes a rating over the entire rating period (of one week in this case) after reviewing who you have played against and how you have performed. On that note, and even though I know you calculate daily to give a semblance of, hmm, movement (for want of a better word), I think it would be better to calculate only in conjunction with the established rating period, as that is what glicko is designed to do.

Also on that subject, since the data is being crunched daily, yet the rating period is weekly, is the entire dataset recrunched from scratch on the sunday rating period end, or is it iteratively crunched from the daily data? Because if it is the latter it will introduce very large compounding errors. Personally, I wouldn't do daily updates, if the system is not designed that way. Most people should be sensible enough to understand why it is that way if the glicko system is explained well enough, maybe in a future faq update.

I've read that glicko works best when people play 5-15 games per rating period. My suggestion on this would be to adjust this period after you have sufficient player game data (based mainly on the top 50 or so players) to make it a compromise on time versus accuracy , if you get my drift. You may be able to drop the rating period to a few days and still maintain accuracy for the majority of players.

I also think it wouldn't hurt to put the Reliability Measurement in another column for the nerd factor.

Cheers.

Bwahahahahaha!

So, I started a game earlier today, but Multiwinia crashed just after it started (actually, I think the poor lappy overheated, and tried to shutdown, causing Multiwinia to crash). However, I managed to still win the game. Anywho, that brings me to my question: I disconnected and still won. How does that affect my rating?

xander

Man, you do have some good hax xander!

Also , just then noticed on the board, 3 people on 1500, yet I was rated 33 the other two were rated 31 and I have the most wins. Not sorting properly or a hang up from removing the randoms?

Probably because the score recalculations happen periodically.

or because you beat someone who was higher/lower etc up the ladder than they did.

By the way, is there a way to see global statistics? For example, what is the most played game mode, map, etc?
I only see specific server statistics, but i was curious about the whole.
From my experience, tho most popular game type seems to be KOTH, since it's the one I see hosted more.

Phelanpt wrote:By the way, is there a way to see global statistics? For example, what is the most played game mode, map, etc?

I don't think you need anything of the sort to know that KOTH is the most played. It's played almost as much as everything else put together.

As for most popular map, that'll be Pursuit of Victory, played even more often than every Assault and Blitzkrieg put together, and also even more than Rocket Riot.


Although, I would like it to be possible to easily tell where the others place.

RabidZombie wrote:Probably because the score recalculations happen periodically.

The GoldFish wrote:or because you beat someone who was higher/lower etc up the ladder than they did.

As I understand it, glicko doesn't or shouldn't operate that way. If the score or rating is equal, then the ladder should check who has won the most games, if that is equal, it should then check who has lost the least games, if that is equal, it should check ties, but since afaik ties are impossible due to sudden death, then this last column is superfluous. The only thing I can think of is that there are decimal points that are not being shown, or it's a bit borked. It's not a biggy, but you know, for the sake of completeness. That the scores are being calculated within a ratings period, could have something to do with it, but that is beyond me.

Yeah ok I have no idea how whatever I was on about has anything to do with what you're saying. You're quite right, you should have been at least tied/ahead of them.

Also, you can turn off sudden death.

Mas Tnega wrote:

Phelanpt wrote:By the way, is there a way to see global statistics? For example, what is the most played game mode, map, etc?

I don't think you need anything of the sort to know that KOTH is the most played. It's played almost as much as everything else put together.

As for most popular map, that'll be Pursuit of Victory, played even more often than every Assault and Blitzkrieg put together, and also even more than Rocket Riot.


Although, I would like it to be possible to easily tell where the others place.

Yep, I'd like that too.
And I'm curious about how much this bias is because of the demo maps, since NeoThermic's demo servers seem to be the most played.

Thanks jelco, those are interesting!

Thanks Jelco, that's excellent. Explains it all.

Might I suggest then you leave the rating period, but possibly have the daily score recalculations given a separate title and numbered as say 2.1 for monday 2.2 for tues etc (I know it's not decimal, but it could be like a cricket overs score. At 2.7 it clicks over to 3.

I still think the reliability measurement would be sweet. Can you even turn off sudden death on ranked servers? Taht would get rid of the "ties" column.

Anyway thanks again for explaining it all.