Vicious Infected Versus - TrueSkill Experiment Poll

I agree with that too a point, some people just dont know when to stop or go way too far with it. At that point, its not really fun for the majority.

 

For those of you that are not sure or want to vote "no", please explain why after I've rebuttled my viewpoint. Thanks! After marvel approves this, you will notice this in action on VI.

 

While I have no objections to the mod, I'm a little confused as to what the overall effect (other than statistic collection) will be for players on the VI servers. Admittedly, I have skimmed some of the longer posts but I don't see much, if anything, about any effects (that affect gameplay or otherwise) of the skill ratings. Could you fill me in on this (or inform me of how blind I am, if I missed what I'm looking for ).

However, I do think Zsoka brought up some valid points in post #25 (even though you have responded to them). It eludes me as to how this mod, which claims to measure a player's skill seems to be based solely on the team's overall performance - if my understanding is correct, a poor player on a team that is fairly good overall would be rated the same as his/her teammates?
I wouldn't go as far as suggesting that Common Infected have any influence on the rating but contributions like saving your teammates from SI would, IMO, constitute teamwork and could be a factor in judging a player's skill.

?Also, Zsoka, you seem to have forgotten my name on that list... :p

 

A poor player would affect the overall sigma and mu summation, and thus affect the distribution curve and probability density curves. Trust me, using the individual player stats to determine skill has been tried time and again, and there isn't a single one of them that has succeeded.

The algorithm I am using has been designed at the Microsoft Research laboratory by well-known mathematicians and used all throughout Xbox Live. It's been tested to work on a scale that spans millions of players, and it'll work for L4D2 as well. However, for it to work, I need to modify some base constants.

Consider the functions:




The "u" is pronounced "mu". The "o" is pronounced "sigma". The mu of the player is mu_final (the mean changes that take place over time) - mu_initial (the base constant that I determine). Same goes for sigma. The "B" is pronounced "beta", and that's one constant (along with epsilon, the "E") that I have to determine. Small changes in the base constant has a drastic effect on the overall skill that is determined. However, every game is unique, and the base constant must be determined on an individual basis. Halo uses a scale that ranges from 0-50 and GoW uses a scale of 0-100, and I'm going to start with the base constant B=25/6 and E=1.0 and use the 0-50 scale and tweak things while maintaining that scale. I need a server with lots of players and a lot of competition, and there's no better way to get the statistical data I need than to install this experimental plugin on a live server.

How does this plugin affect your gameplay?

It doesn't, unless you play PR, in which case you will begin to see brand new skills and upgrades that I will begin to copy from PRev into PR. Remember, TrueSkill is all about "skill", and not what you would refer to as connection time or game time ("experience"). I could connect with 0 days connection time and reach a TrueSkill of 50 in less than 12 hours of connection time if I was really good. Then, I'd be able to use upgrades and skills that noobs cannot use and thus level up faster. This will give newcoming players a chance to level up and use upgrades/skills that will also give them an edge vs. high level players. It will also give high level players something new to strive for than just a higher level.

From what I've noticed, a lot of high level players are noobs. These people have enormous connection times but when it comes to pure L4D2 skill, they have none. Yet, they walk around like they're on top of the world, like they own the world. Now it's their chance to prove it. (Put a high level PR player in VI and they will fail epically, e.g. me.) So people who are really good but just don't have the connection time will have some upgrades and skills to aid them.

A warning perhaps, TrueSkill is known to make games extremely competitive. Let's see if the same holds true for L4D2.

 

Does this plugin take into consideration a survivor helping or saving a teammate? Or sharing items u have like pills or kit with a teammate? What about when someone who is incapped, do they get any type of credit from this for using their pills to save(revive) a teammate?

 

I'm down with whatever ranking system you wanna use, so long as it doesn't directly affect game play, but lemme say this...

You can use a much simpler algorithm to get a true trueskill ranking. Just pick an arbitrary max value (say 50) and then subtract for each time a player bitches about teamstacking, and for each time they call a vote scramble, and for each time they demand teams be evened up, and for each time they rage quit, and for each time they are a girl.

Simple.

 

Thank you Mr Positive ain:

 

None of those, and for a good reason. If you're curious about how it works, I highly recommend you read this article. To summarize that article, trueskill looks at only two things: 1) did your team win? 2) who were you versing? That's it. Saving your teammates helps your team win, and that's what counts in the end.

That system has been tried in other games and have failed. Trust me, TrueSkill is the way to go. It's the algorithm being used by a multi-billion dollar gaming industry and continuously tested by millions of gamers.

IMO, if we were using Bonekeep's system, I would program it so:

Called vote scramble? -1 skill
Demanded even teams? -1 skill
Rage quit? -2 skill
Is a girl? +100000000 skill.

 

I have lost all confidence in your programming skills and common sense.

 

So in a nutshell, this in no way actually uses the players individual skillz to calculate this 0 to 50 skill level?

 

Individual skills? In a team-based game, individual skills don't matter. Though that statement doesn't make sense considering how TrueSkill measures individual skills, in all actuality, individual skills is determined through your sigma and mu value and its impact on your team's overall performance, the sigma/mu summation of your team vs. another team and the final outcome.

TrueSkill is like dice-- you never know what the outcome will be (unless the mu difference between the two teams is huge), but you can use probability to your advantage.

 

Hmmm :juggle: . Had I not already voted, I'd prolly go with no. Honestly not seeing the point since the current stats system takes into consideration the abilities of the player. Its not great, but its something :juggle: .

I am understanding trueskill only tracks wins an losses an distributes points solely on that criteria, correct? Kinda seems like a waste of a cpu process, but that is JUST MY OPINION and does not reflect the opinions of ANYONE else here :sorry:

 

See, that's exactly where you're wrong

Read the algorithm first. It's so damn accurate. In fact, the stats system we have doesn't measure skill at all. It measures experience. How many headshots you have, how many infected you killed, and your gameplay time and thus the calculated points all have to do with how long you have played on MG servers.

TrueSkill is different-- it doesn't care about any of that stuff. It solely cares about your skill and how you contribute to the team. Although it doesn't consider any individual stats per say (technically it does because it looks at your sigma and mu values), it's so accurate. It's been tested time and again and it works.

 

No, man, read the algorithm. It's not as simple as winning or losing. It's also matter of your skill and how your skill contributes to the team's overall skill and who you're versing.

Every ranking system thus far on MG was experience based. This one is skill based, and it works, and it already has been tested to work. I'm just bringing the same algorithm to L4D2.

EDIT: lol, I posted the above, and it didn't show up, so I posted this post, and then the above shows up.

 

I don't really agree... L4D2 is still a FPS and being able to place headshots or not, for example, is a player's skill, you can't bundle that with the amount of time you've played on the server.

In fact it'll be the same with your TrueSkill. the more you play the more accurate it will get and the longer you play determines if it goes up or down, if you get better not. The fact that a team wins or not is important in determining team based skills, but not all. It's because each individual in a team is good + can play with his teammates that people win in VS... Neglecting individuals is neglecting the team itself lol.

The algorithm is very complete, for what it does: calculate global stats based on team scores. I just don't see it measuring a player's skill to their full extent, when the basic gameplay itself is neglected (headshots, damage, helper, weapons used, round time, defibs used, pills given, etc...)

 

Just let him try it out and lets see what happens. If Crotchfire places anywhere other than the bottom then you'll know it's a piece crap.

 

Read this: http://www.bungie.net/Forums/posts.aspx?postID=14414600

Why HLStats does not measure "True" Skill:

If you read through some of the pages in the above link, everyone thinks like you. "TrueSkill is b.s.! Your K/D does matter! The number of headshots you get do matter! The longer you play, the more experienced you are, and thus the more skillful! How well I play should be determined on an individual basis, so if I do well but my team fails, I shouldn't be penalized!"

No, all wrong. I could be a L4D2 MLG pro (Major League Gaming pro) right now, and I could join the MG server and be classified as a noob for not having 7 day connection time. Experience is not equal to skill. You guys don't seem to get the difference between skill and experience.

The number of headshots I rack up depends on how long I play, not how skillful I am. The more I play using HLStats doesn't make it more accurate. The more I noobily get biled, the more infected I kill. HLStats will consider my kills going up and thus I must have more skill. How many weapons I used, the round time, the number of assists in that round, etc all depend on individual factors that affected the gameplay within the round, and not skill itself, as each round is different. If I'm a noob and I get stuck behind, I will naturally kill more infected and use more weapons and assist more noobs. In the end, all that determines your "skill" is whether or not you win and survive, and how well your team did in comparison to you and the other team.

There are many people that have your ideology (at least 99% of the population I would imagine), but you need to understand that it has never worked. They all determine experience, not skill. Around the time of Half-life, people went a step ahead and determined the number of headshots / min. If you must have more headshots/min and more kills/min and less deaths/min and more assists/min then you must be really skillfull, correct? All wrong again.

While it's true that the longer you play the more skillful you become, any ranked system we used thus far only measured experience, not skill. You need to try to understand what trueskill is before you start putting some lowly HLStats algorithm above it. If you are really curious on why TrueSkill works, then please, read the algorithm I hope I made sense here, if not, feel free to rebuttle. So to summarize, HLStats is a time-based experience calculator, not skill calculator.

 

I read the article, and although I still don't agree with your point of view Luke, I understood the system lol. I'm all for TrueSkill, so thanks for transposing it into L4D2.

I still can't see how well it would actually work in a survival FPS like L4D2 when the two teams aren't equal in 1 round (surv vs inf then switch over) knowing the winning factors aren't the same for each and completely different from the usual FPS games where we play to shoot each other on equal grounds

 

When I used to play Halo 3, I got stuck at 45, and I was so pissed. I was sick of carrying the team, and I was sick of always ranking down when I always used to do really well. My team always brought me down, and I always used to yell on the mic that everyone sucked lol. I hated the TrueSkill system because the game never let me cross 45. Not until I actually started programming the algorithm in the last few months did I realize how effective it was.

I understand where you're coming from, and I agree with your point of view that the system needs to take individual abilities instead of the outcome to determine skill, but trust me, that has never worked, never will, and never can. Unfortunately, the probability distribution would be near impossible to calculate, and because of the random chance that anything can happen when you are keeping track of many variables, there is no efficient determination for real skill.

I hope that example of the MLG pro made sense to everybody on why HLStats isn't a skill calculator.

 

I've got to agree with Luke for the most part, from experience trueskill does tend to provide a better balance, I always found Halo games challenging because of the balanced teams (unless I joined a game with mates, then it was a cake walk ). How well it translates to L4D... well we will just have to experiment, see if it works long term, I doubt it's going to be any worse than the bizzare srambles we get at the moment.