Looking to come to ravenholdt

05/02/2012 02:41 PMPosted by Tahlasar
I should mention they tend to be a bit more erotic in there role play activities. So if that is not your taste you might want to try another guild.

I know, they almost outclass my slash fics.

Seems like an interesting discussion in regards to arena rating vs. skill.

Firstly, while arena rating does equal skill in some cases, there are many cases where rating isn't as important as win/loss or what type of comp you are running/items you have.

For instance, i've seen 2.2 players in full current season gear, with over 400 games played who sit at their rating with 45% W/L ratio. This is an example of someone who simply got lucky one night and went on a winning streak vs. someone they countercomped. Then you have the people who are in honor gear who have 70-80% W/L with only 40-50 games played. While they are both 2200 rated, its clear that Team A had a much harder time coping in the bracket than Team B.

Now lets imagine i'm looking for a rogue to play RLS (top tier comp atm.)
I'm looking at two rogues, one is 2200 rated Human rogue and has legendaries and a heroic vial. The other rogue is a 2200 rated Undead and just uses T2 PvP gear. While rogue A certainly has the capability to do more damage than rogue B, its fair to assume rogue A was carried to his rating by his items, and that rogue B was more skilled and probably has a better grasp on his class.

Another bigger problem that you face on higher population servers, is carrying/piloting.
To be able to tell who paid for their rating you have to know what to expect out of somebody to be playing at that level. From a casters perspective you have to look at whether they know how to fake cast, correct use of escapes, use certain macros(like focus casting, mouseover and arena tar.), are using their mouse to turn, have a good sense of map awareness, are capable of coordinating with their team partners at a good level, know how to kite melee, peel for their teamates and aren't clicking anything. Melee is pretty much the same, but with a few differences like not getting faked, good use of gap closers, and more emphasis on map awareness.

Why am i using 2200 as a base point?
Well thats when you really start to notice a difference in players, 1500-2K you really don't see that big of a skillgap. Typically these are people who play for fun or play with friends, or who lack one or more of the characteristics listed above. You see a lot of melee in this bracket, with the casters being lumped up a little lower.

But how can I tell where i belong in the arena brackets?
When you start to get stressed out or overwhelmed by how the enemy team is playing, thats where your "true" rating is it. If everything seems completely easy to you, and you aren't really feeling pressured even when you are being focused, then you probably belong a bit higher than that.

The most important thing though, is how you interact with your team.
When you lose are you analyzing what happened rather than blaming each other? Are you attempting to change your play as a team so you don't lose to a certain comp in the future? Is everybody calling stuff out?(when you get CCd, when you fake an interrupt, when you want to pop your cooldowns and go for a kill, when you need peels, when you are CCing one of the members of the enemy team, when you are on DR for your CC and need someone else to take over, and what target you are going to be focusing or switching to.)

TLDR; Judging skill is complicated.
05/03/2012 09:19 PMPosted by Rimlol
I hope that kinda made sense outside of my brain.

It very much did! (And it's very refreshing to see an eloquent response; you're a good writer!)

I agree that the point when players stop making "mechanics" mistakes can be used as a "good" vs. "bad" separator, but unfortunately, I don't think it's that simple. I would argue that given a long enough period of time, every player will make a "mechanics" mistake, which puts you back at square one.

How do you decide how long of a period of time needs to pass without a mistake to classify a player as "good"? Is it one arena match? If an arena match passed without either team making a single "mechanics" mistake, would you put all players into the "good" category, even though they may make a mistake the following match? What about two matches without a single mistake? Twenty matches? Where do you draw the line and why is that line any better than any other place?

What about the other direction? Do I deserve to be placed into the "bad" category if I make one (but no more) mistake every arena match, even though an "average" (median) player makes 20?

Personally, I think there are very few dichotomies in the world, but as humans we like to think of things in the yes-no, good-bad, left-right, etc. terms for whatever reason, so we continue to fit a non-binary world into binary perception.

Why am i using 2200 as a base point? Well thats when you really start to notice a difference in players, 1500-2K you really don't see that big of a skillgap

Yep, looking for the largest gap in the distribution is a actually one of the better ways to break up a continuous measure (like skill) into two categories. The reason is that if you draw the line directly in the center of that gap, you will have the biggest distance to the players "closest" to that boundary, making it least sensitive to noise/variance (see below).

I obviously haven't played at the 2200-rated level, but thinking about it, it may go in line with Rim's discussion about not making mistakes versus playing better. It could very well be that players below 2200 make a mistake every match, while players above 2200 make a mistake every twenty matches, giving you that perceptible "skill gap".

Of course, 2200 is also one of the points when you can slightly upgrade your gear, which may be increasing the "gap" between sub-2200 and over-2200 players, but it's gear-related, not skill-related.

Unfortunately, unless you're Blizzard, you have no way of collecting statistics on the number of mistakes made by players around the 2200 threshold, so we can't test that hypothesis. :P

05/03/2012 11:29 PMPosted by Uglygoblin
TLDR; Judging skill is complicated.

The 400 games at 45% W/L ratio vs. 40 games at 70% W/L ratio example you gave is a classical undersampling issue. While both teams are showing 2200, I would argue that the "true skill" of the second team (the 70% W/L ratio one) is higher, and 2200 is simply a worse approximation of that "true skill" due to the fact that they only played 40 games. When they play more games, I'm willing to bet that their rating will continue climbing until it stabilizes (which will happen either when they hit 50% W/L ratio, or they reach the very top of the ladder where they beat almost every team they play).

If you want to get into the gritty technicalities of it, there are two theorems in probably theory that explain what's happening.

1) The Law of Large Numbers, which basically means that the more samples you take the closer your sample mean will be to the "true" mean. This means that as a particular team continues to play games, their rating will move closer and closer to what their "true" (not directly observable) rating is. (Technically, every game gives you a rating increase or decrease, rather than the rating value itself, but this wonkiness will smooth out over a large enough number of games).

2) The Central Limit Theorem, which explains how far the sample mean is from the "true" mean. (There's more to it, but I don't want to go into the details.) Basically, it lets you make claims like "The 'true' (unobservable) team rating is between 2150 and 2250 with a 95% probability". That range will grow smaller and smaller with more games, allowing you to "zero in" on the "true" skill.

All this is assuming that players remain at their "true skill" level rather than get better, as they actually would simply by the virtue of gaining more experience. This also assumes that the environment is static (same teams, same players, same game mechanics), which isn't in reality.

So yes, judging skill is complicated.

EDIT: Also, terribly sorry to the OP for hijacking the thread. If you have more questions, feel free to butt in and I'll shut up.
ITT, Cayreth is a nerd. But he's OUR nerd, and we loves him.
05/04/2012 09:34 AMPosted by Oolaki
ITT, Cayreth is a nerd.

It's true. I spend more time thinking about game mechanics than actually playing the game.
Cayreth awesome guildmaster, too. Him quite likeable.

But him still now allowed to ride on WarPig's back.
My two cents, you can obtain the 1v1 skill of a 2200+ play if you duel enough. Duels build skill, trufax.

Only reason i get annoyed with the rating statement is because i am very skilled have been 2200 many times but simply don't care for it any more. I am still a good PvPer just don't feel like spending hours rating climbing each night, that and i don't really have the time for it.

I feel like there is an Evil Robot Overlord Nenai joke in there somewhere...I can't quite find it. xD

Edit: Had to correct grammar before the grammar !@#$ caught up with me.
05/04/2012 02:23 PMPosted by Tahlasar
Only reason i get annoyed with the rating statement is because i am very skilled have been 2200 many times but simply don't care for it any more.

I used to think this way, but then again I realize that if I were to start arenas again, I'd be so stupid rusty that I'd be sub-1700 for a while. Why? I am a competent player, but part of winning MMO PvP is playing a lot.

Just like in SC2, while I know everything zerg macro entails, know my build order, etc. without playing regularly I'm going to have abysmal APM and get squashed because half of being efficient is having the speed to pull it off. Having the practice in WoW arena in order to be able to react quickly is just as essential as being a competent player.
Cayreth awesome guildmaster, too. Him quite likeable.

But him still now allowed to ride on WarPig's back.

You are too kind, Krahmm.

I think.
anyone who thinks Ravenholdt is a good server your dumb

You spelt (or spelled; depending on your nation & preference) YOU'RE wrong.

Join the Conversation

Return to Forum