TWD Rating System info

[Shadowmere]

Our Dev (Qan) and math master (Ogron) have worked tirelessly along other projects for over 3 months now.
They started to revise the current TWD statistics in December, and have come a long way.


Result:


We now have formulas to rate each player based on their performance and team's skill level versus the skill level of the opponent they are facing.
Your rating will now also take into account "confidence". What this means is the more you play, the more accurate your rating will be.
If you get subbed out or subbed in, the amount of time you played will result in whether you will gain or lose any rating at all.
By this you should maximize your contribution to your team. Do whatever you can to be the best you can.


In Javelin Division all this also incorporates teamkills which will naturally lower your rating.


Basing Division is the most complex in terms of rating calculations.
It will incorporate Kills / 30 minutes, as well as depending on the ship, how long your team has held the flag, which ships you kill, how good your RPD is and so on...


What this means:


We will prepare ourselves to inputting all the relevant data into TWD.
TWD will require downtime to execute this change.
Hopefully we can manage the downtime at quieter times of the day, but there may be need for hotfixes at urgent times as well.
Please don't be surprised if that happens.

[Shadowmere]

Once TWD Ratings System is in it will require testing!!


The more squads participate in this change, the better.
As your personal ratings change rely heavily on who you play against and how you do against them, the more you play, the better.
Note: You will not gain huge ratings continuously playing against worse players than yourself!

[rothe]

So K/D and time in game will have more of a impact on ratings? Along with rating of the player your playing against?
Only question I have is does the more you play the better take into account the more you fly. IE. the more your in Matches/Public the higher your rating or just being in game counts? All and all looks like a interesting rehash better players vs better players is always more fun. Never found it right top players used friends to get top spots so this hopefully will fix that.
Thank you Qan and Ogron along with anyone else who helped. I guess Shadowmere as well = P

[qan]

So K/D and time in game will have more of a impact on ratings?

Loosely speaking. The formulas haven't been tweaked in years. Ogron's crunched the numbers with the intent of having them be more representative of skill.

Along with rating of the player your playing against?

That's the biggest change, yes. I did a quick calculation with some basic numbers and it can be as much as a 30 point difference, if a group of top players are playing a group of low-rated players. The rating of the squad doesn't play into this, though. So you can be on a #1 squad and play your weaker players to offset this difference.

One thing that I think may be a little confusing is that there will be 3 ratings for a player, 2 of which are long-term/held over all games.

Game rating: How you're rated for 1 particular game. This takes into account the Star Rating of your opponents (see below).
Player Ladder rating: The average of all your game ratings multiplied by confidence. Confidence increases with each game you play. You can't just be there, and need to play a majority of the game. If you only play 3 games in a season, you'll be near the bottom of the player ladder, even if you play perfect games. Elim is using this system as of fairly recently.
Star rating: A rating that is sort of an expectation of how well you'll do. It's what's used to calculate expected opponent strength, and that gets factored into the game rating at the end. Star rating changes slightly with every game, based on game rating, unlike ladder rating, which can change a bit more. The data from TWDT will be fed into this as a starting point. Anyone not rated on TWDT will start with a 6 rating.

There will undoubtedly be some problems with this so we'll need to have a decent testing phase. If you can see anything that might cause problems ahead of time, please bring it up ... don't even care if you need to roll your eyes like a bratty teenager and say something like, "GOD, staff are SO dumb, why couldn't they see this?" -- let's just get it figured out, eh? There are already a few issues coming to mind we're going to have to nail down...

[Turban]

first of all, thank you for your hard work.

the introduction of player ratings will definitely spice up the competition and thus will be a great addition to the league. however, in the end all of this boils down to the execution of the project. it is more than likely that the algorithms used need to be tweaked further in order to balance it out, and knowing how unexpected events tend to appear out of nowhere, several adjustments and changes may be required before fully implementing this system to the league structure. anyway, that is what the testing phase is for.

a minor concern of mine is about the player rating formula. it would not hurt to add some way of catching the players who are trying to inflate their skill rating with the help of their friends. it would not be too surprising to see some people try to cheat the system, and even though the confidence factor should help, it could still end up being a potential problem in the future unless you want to rely on player reports.

anyhow, keep us updated on the progress.

[ogron]

So for TWDD and TWJD players, this means kills, KDA, margin of victory, time alive, strength of your team/strength of opposition, knockouts, and TKs will all be factored in to your rating.

For TWBD players, this also means kills per minute, margin of victory, teks, RPD, as well as more minute things like valuing shark kills slightly less. Also, sharks can now MVP : )

While no numbers can fully capture and determine good play, these formulas are pretty complex and should give a much better approximation of performance, and hopefully incentivize more TWD play as people vie for higher rankings on the top players list.

A 3.0 RPD, 5 tek, 20 kill performance in shark against a tough opponent and winning by 7 minutes is likely to produce an MVP, while a 10-7 game as an underdog with 2 knockouts in a 50-30 win will be more valued than a 11-6 performance with 2 knockouts against bad players. A 10-10 dying out halfway through the game will be much lower rating than a 10-10 if you died out 2 seconds before the end of the game.

[qan]

several adjustments and changes may be required before fully implementing this system to the league structure. anyway, that is what the testing phase is for.

Yep, definitely. It will be particularly challenging because I don't play TWD. Will be relying on those who do to not only observe what is an issue, but also to suggest how things should be changed to fix it.

a minor concern of mine is about the player rating formula. it would not hurt to add some way of catching the players who are trying to inflate their skill rating with the help of their friends. it would not be too surprising to see some people try to cheat the system, and even though the confidence factor should help, it could still end up being a potential problem in the future unless you want to rely on player reports.

A big concern of mine as well. It wouldn't hurt to add such a system but at the moment I have no idea what that would involve. Usually such algorithms are pretty complex, and also prone to error. I have absolutely no idea what the numbers look like when players are doing this (and maybe nobody will under the new system until we see it), so at the moment I don't have any proposal for how to stop it. I do know the star system should prevent some abuse, though. If you can provide some info on how some players do this currently, it might provide some insight. Though it's a little surprising anyone would do it, as it effectively makes their rating meaningless, and purely a lie they can brag about to others. Then again, I sometimes am naive and underestimate the depth of human foolishness, especially when impressing people on the internet is at stake.

[qan]

I've begun mapping this out in more detail, going through what will need changing in each part of the system, and here are some snags I've run into so far.


EDIT: Worked out these issues so I'm removing them here.

I've now mapped out almost everything needed for this project. Just have to go through the process of putting it together in code.

[2pac]

I've begun mapping this out in more detail, going through what will need changing in each part of the system, and here are some snags I've run into so far.


EDIT: Worked out these issues so I'm removing them here.

I've now mapped out almost everything needed for this project. Just have to go through the process of putting it together in code.

good looks bro u need help just let me know

[Attacks]

Looks good at first glance and keep up the good work!, can we have a quick look at player star ratings again before just copying them over from twdt to twd, im not sure how perminant they will be when posted on twd. Absent, 24,afri and myself better than 7 in warbird, Trasher better than 7 in jav etc, a few players need to be brought down also needs a skim through.
Also anyway of punishing consistant squad hoppers by a rating penalty just like changing names on twel?

[qan]

Jesus, this is a hell of a lot of work...

I spent about 8 hours on this today and have the larger part of it done. Basically all the stats are loaded where they need to be, and I've entered the stat crunch for dueling modes, but need to do the crunch of base (about an hour of work). Also have some odds and ends to wrap up, 4-5 hours tops, but it's mostly in the bag. Then bug-testing for awhile.



Re: Star rating, it will only exist in dd/jd, and it will change depending on your performance. Having a high star rating is actually not a great thing, except obviously for bragging rights, because you need to play equal to your level consistently or else your round rating and thus your ladder will suffer. So star rating and ladder rating should kinda balance out eventually to what you actually are. I apologize for the confusion about what the difference is between star rating and ladder rating ... it's a bit hard to grasp. Basically star rating is your perceived value at the start of the personal rating system.

Personally I worry that it might just make a bunch of people register alts to play TWD. But, eh, that's not really my problem, in the sense that I don't league, I know little about TWD, and I'm just following a blueprint on this one...

[qan]

Basing ratings finished, just have to do some exciting PHP work now...

[Markmrw]

Can't wait to see the result of all your hard work

[Crescent Seal]

*edit title

TWD Rating System - COMING IN 2037

[2pac]

*edit title

TWD Rating System - COMING IN 2037

Done for you, yw