Still, even if it's not 100% perfect, I'd consider it a very accurate measurement of skill.
I don't think individual skill is what's being measured really. It's just a representation of team performance over the past x games. Certainly the player's skill is a factor in that, but that's not quite the issue. What's being measured and what players want represented by their elo are distinct.
The modelling for elo based on the result of the sampling could be perfect and it'd still have 5-9 confounding variables in every sample, namely the 5-9 other players. And this is before you get into roles, teamcomps, and so on.
The only thing inaccurate about the concept of "elo hell" for example, is that you cannot get out of it. The lower the elo, the greater the effect these confounding variables have because of the ability of 1 player to throw a game for their team; the higher the elo, the more normal they get because players approach the baseline of skill in all areas generally necessary not to totally throw a game.
You can look at a player as having a set of "skills":
1) Map awareness. Being able to gauge what enemies might be doing when they're missing, as well as being aware of what it means when they're visible (jungler ganking top? We're safe for aggression for a minute at bot/I can steal his bot buff camp if it comes up).
2) Role awareness. Knowing what they're supposed to be doing (supports who don't realise they're supposed to ward and ward clear).
3) Role capability. Doing what they're supposed to be doing (tanks who don't initiate/softzone enemy damage dealers well, damage dealers with bad focus or scared of getting focused).
4) Champion building. Knowing how to build a champion, how to structure a build based on what's going on in your game, what runes/masteries/summoners to take against the champions you're facing.
5) Laning capabilities. A lot of people who are just bad at positioning and last hitting in lanes, always a step too far away to get a last hit in time, facechecking bushes, and so on.
6) Mistakes. Just plain ol' "how many mistakes will this player make"? Are they greedy, cowardly or blind?
And as you get higher in elo, most players will come to around the same level of "skill" at each of these things. But at lower elos, you'll find people with much different skillsets.
For example, the dudes who are brilliant in lane but have no map awareness and are way too brave in teamfights. Leave them to 1v1 and they'll dominate, but they'll always get ganked if the jungler/other lanes try it, and in teamfights they'll get focused because they're out in front too much. Or the dudes who know exactly what's going on where on the minimap, but play way too conservatively in lane and lose half of each wave's last hits while the enemy laner(s) have control of the lane and when to gank. Or the junglers who aren't really aware of their own buff spawn timers, thus ganking lanes at the wrong time and coming back to find the enemy jungler has stolen their buff because that jungler knows where they were at the time the buff would come back up at.
Hmm I might as well throw this up there. I have a replay of myself playing Lee Sin solo top vs Irelia at 1600 solo ranked. I got ganked early on and lost the lane advantage and then proceeded to tower hug and play only semi-safely while my team carried me hard.
I didn't teamfight or gank or anything. Maybe someone here has some input on what I could have done to make this matchup easier for myself, because I know I could have played that a lot better. Basically input for the first 10 minutes of laning.
We were doing some work in those games, though I think our first set of opponents was of a spectacular class of baddie. We ought to do it again sometime.
Hmm I might as well throw this up there. I have a replay of myself playing Lee Sin solo top vs Irelia at 1600 solo ranked. I got ganked early on and lost the lane advantage and then proceeded to tower hug and play only semi-safely while my team carried me hard.
I didn't teamfight or gank or anything. Maybe someone here has some input on what I could have done to make this matchup easier for myself, because I know I could have played that a lot better. Basically input for the first 10 minutes of laning.
I don't play a lot of Lee Sin, but WWQQE on the enemy tends to do wonders.
Irelia is a bit like Udyr when it comes to the laning phase; they're almost impossible to play poorly and they either dominate a match-up or are countered hard. I find LS pretty good at farming at the tower, though, so it shouldn't be too much of a problem.
It occurs to me that I know have champions I like to play on every role except AD carry. Well I guess support but that's a case of where I have the champs , I just dont want to play them. The only character that is an ad carry that I like is Twitch , im just bad with him. If I were to learn an AD carry who would you suggest? I have Ashe , Silvir , Corki ( came in the camp bundle never played him) and Twitch. I think I have all the necessary runes for them already so thankfully that wont be a problem .
Sivir is a very strong and, in my opinion, underrated champ. There are very few champs that beat her in lane (a much better player will still be better, of course) and her Q hits like a TRUCK if you can land it without hitting minions. It does merely good damage if it goes through a minion wave first. W is eh, it's great for pushing lanes later in the game and ok for random poking in lane. E saves your life and makes your ridiculous mana usage bearable. R is a great move and attack speed steroid that you can use to burn down towers, buff your whole team in team fights, or help your ally escape from a bad skirmish. It's what makes you able to split push in relative safety, since if you have boots2 and a PD you can move almost as fast as a sprinting Yi.
I run ArPen marks, AD quints, flat AR seals, and MR/lvl glyphs (though I'm thinking about mixing in a few AD marks). For masteries I run 21/6/3, taking the left side of the offense tree (except crit) with 2 points in CDR and 3 in Havoc. Armor/mr and mana pool for the other points.
I start boots+3pots or boots/ward/pots and B sometime after I have 950g to buy 2 dblades. After that I like to get zeal>BT>finish PD>IE> then see what I need. If someone with a lot of armor is giving me fits then I'll build a LW as my 3 major purchase. 4th purchase is usually a defensive item against whoever is doing best on the enemy team, then finish off the build with some combination of PD, BT, and IE. Getting 2pd, 2BT, IE, LW is a thing of joy and beauty that I only managed to do once, but I could melt teams when I did.
In lane play safe until level 3 or so, taking q,e,q. Q enemies when you have a good shot at them with minimal creep in the way and auto attack them if they come into range. You are strongest against combo mages like LeBlanc and Morgana, and weakest against tanky AD types with gap closers like Jax and Xin. A good Teemo will just eat you, so ask for a lane swap. You must must MUST learn to be super quick with hitting your e, most champs have some sort of tell if they are about to hit you with an ability, and you need to be able to react to that instantly in order to gain mana. And you will be short on mana. In general, get ready mentally to hit e if a champ that has been hanging back suddenly walks toward you.
Later in the game you are a fantastic split pusher, a QE combo can clear out an entire minion wave. Watch the mini-map, when the enemies start vanishing then it's time to get the hell out of dodge. I like running TP so I can split push then tp to a different lane or to a team fight. It can also be used to escape, since it is faster than B and doesn't seem to be interrupted by damage. You can outrun everyone except Singed, Yi, Teemo, and sometimes Nid. A good Ashe can also ruin your day with the constant slow.
In team fights, stay behind the tougher people on your team. Skirmish with Q, focus on their squishy types when the team fight erupts. Save R for when the team fight involves everyone, don't lead off with it in the skirmish part.
Soon... soon I will install you, my precious.
0
Zen VulgarityWhat a lovely day for teaSecret British ThreadRegistered Userregular
3s transfer and 5s deathrays are mean in fights when you have 400+ ap, and his stun field is only about 6s as well.
Except lots of ap carries does that too, and better. I really really tried with Viktor but I can't find any appeal in him. If he was tankier or had longer range he might work. The only thing he has that no one else has much better is his stun.
Learn where and when to die. Learn to hit single targets with q.
Get Deathcap early for ult trolling. Build WotA when you want to be nice to your team, extra archangels when you just want to run around with defile on all day. Karthus charges a Tear incredibly fast if you're willing to just chuck out qs every few seconds.
Pair up with someone with strong cc to keep people harmless in the range of Defile. Leona is nice for it. And don't be afraid to flash offensively. Suddenly having a Karthus in the middle of a team is an unpleasant experience for everyone involved.
3s transfer and 5s deathrays are mean in fights when you have 400+ ap, and his stun field is only about 6s as well.
Except lots of ap carries does that too, and better. I really really tried with Viktor but I can't find any appeal in him. If he was tankier or had longer range he might work. The only thing he has that no one else has much better is his stun.
I dunno, he isn't top pick or anything but he can skirt around the outer edges of team fights firing off deathrays and dropping big slow/stun circles in the middle of enemy formations. You've got an aoe silence to drop that is also aoe damage and can follow anyone who drops low, and the damage and mitigation of power transfer really adds up with such a low cd.
I'd say he could be on par with Annie. Less initiation power but stronger over the long run.
Steam / Xbox Live: WSDX NNID: W-S-D-X 3DS FC: 2637-9461-8549
0
TehSlothHit Or MissI Guess They Never Miss, HuhRegistered Userregular
TehSlothHit Or MissI Guess They Never Miss, HuhRegistered Userregular
edited January 2012
Yay, Vayne was a boss. I didn't think she had a chance since there really isn't anything to condemn against mid (unless they're really bad) which hurts.
EDIT: First ranked game where my teammates were not only competent, but actually good. Things are lookin' up.
We were doing some work in those games, though I think our first set of opponents was of a spectacular class of baddie. We ought to do it again sometime.
Mid against Vayne can give you a false sense of security. Yeah the walls to be condemned against aren't a big deal but simply taking the shot+shot+roll+shot harass can add up. Don't forget about Vayne's passive.
I tried jungle Mundo using pages I already have for junglers. Doing the regular path of Wolves-Blue-Wraiths-Wolves-Red-Golems and starting Cloth+5 potions. Skillwise I went E, W, E, Q. Potentially, you could start with W instead but it adds like 10+ seconds to the Wolves opener. Maybe if you start W and go Wraiths then Wolves... hmm...
#1: ArmPen/Armor/CdR/AD runes using 21/9/0 = Got lucky with some crits on Blue but Red nearly killed me. Level 4 by 4:50 is not good.
#2: AS/Armor/AS/ArmPen runes using 9/21/0 = So much easier! Got down to 130hp on the golems but I was able to clear the path by 4:05.
Masochism makes up for not having much AD early on so there's no need for the runes. I think it's possible to even start with regrowth pendant instead of cloth+5.
BLACK FLAGS ON THE HORIZON
GREAT THUNDER AND BURNING RAIN
HIS WARSHIP IS COMING
I HEARD THE LAST PICKS SCREAM HIS NAME
I WATCHED AS HE RALLIED
TO THE DRAGONS WHO DIED THAT DAY
HE HELD UP HIS SABER HIGH
AND BLUE PILLED SO HE COULD BUY
GANGPLANK, THE MIGHTY
GANGPLANK, THE BRAVE
CRUSH THE NIDALEES IN YOUR WAY
BY YOUR PARRRLEY LET NONE BE SAVED
LIVE TO CRIT IN THE PIRATE WAY
SORAKAS, WARWICKS AND SHENS
WE'LL BARON TOGETHER
IN THE END
GOD OF FARMING, GOD OF RANKS
PUB STOMPER WHO FEARS NO GANKS
THE POWERHEAD OF THE METAGAME
NOW ULT FOR NEVER ENDING BLAME
SHOOT YOUR PISTOL TO CRACK THE SKY
LIFT YOUR CAPE SO THAT YOU MIGHT FLY
BACK TO PORT FOR THE WHORES ON HIGH
AND CURE YOUR SCURVY WELL
Also, your writing correctly predicts that SC2 did indeed use an uncertainty subsystem and would track individual player performance variance. (I use quotes in "Elo" above. SC2 doesn't actually use the Elo system, but a variant like Glicko or Trueskill. Still they're all commonly called ELO and attempt to achieve the same goal anyhow.) Here's a link to the SC2 ELO thread.
Starcraft does not use a Bayesian system, they likely use something that approximates a Bayesian system. Every system is bayesian in that its an updating system built on Bayes rule that attempts to generate a normal distribution. Trueskill is similarly not Bayesian but rather algorithmic. This is not a bad thing, its just not strict distribution updating. But strict distribution updating probably isn't possible anyway on the scale we're talking about. Any system that updates your rating based on whether or not you won or lost a game is Bayesian. Elo was just one of the first sorting algorithms(and was relatively simple. E.G. if you look at true-skill it reduces to Elo when players have the same std)
Adding the sigma doesn't really give us any more information once you have played a sufficient number of games. This is because, so long as we believe that players can get better or worse there is a hard minimum cap on your standard deviation as measured by the game. We can assume this hard minimum cap both by simply starting with the same prior distribution for players who have not played a game and running the algorithm for the game number cut off that we believe is no longer representative, but also because the assumption of a moving rating means that we must have a minimum uncertainty about your current rating.
Having a sigma is useful only really for putting people into leagues, so people don't bounce between leagues as easily and so that we don't put people in leagues above where they are until we really believe they can handle it.
I don't think individual skill is what's being measured really. It's just a representation of team performance over the past x games. Certainly the player's skill is a factor in that, but that's not quite the issue. What's being measured and what players want represented by their elo are distinct.
The modelling for elo based on the result of the sampling could be perfect and it'd still have 5-9 confounding variables in every sample, namely the 5-9 other players. And this is before you get into roles, teamcomps, and so on.
So lets get some things straight
1) Before champion selection LoL is symmetric
2) Winning/Losing takes into account all of those factors
3) player skills, while sometimes disparate across areas, tend to be highly correlated. I.E. it is possible to be a good laner and a bad jungler, but the better you lane the better a jungler we expect you to be.
Because of this fact, we absolutely can go backwards from "did you win or lose" to "how good a player are you".
And unless you've played a very low amount of games you're going to be placed with a reasonable accuracy.
0
BethrynUnhappiness is MandatoryRegistered Userregular
2) Winning/Losing takes into account all of those factors
No, it makes a fundamental attribution error, in considering all of those variables as things that the player has control of and factor into their skill.
3) player skills, while sometimes disparate across areas, tend to be highly correlated. I.E. it is possible to be a good laner and a bad jungler, but the better you lane the better a jungler we expect you to be.
This is exactly what I was objecting to. The idea that you can lump in every aspect of the skills needed to be a perfect player to determine whether someone is "good" or "bad" as if it accurately presents their skill.
Skill in this game is not a one-dimensional attribute. It is multi-dimensional, and there's plenty of interplay between those dimensions.
If you've seen a top player play a role they're very unfamiliar with (saint did quite a fun experiment while his account was banned, which was to play a role he was unfamiliar with from starting elo; you could easily see he wasn't performing at the level he does on a jungler) you can see where elo starts to fall down when you're trying to assert that it's an accurate representation of a player's individual skill.
No, it makes a fundamental attribution error, in considering all of those variables as things that the player has control of and factor into their skill.
No, this is not a fundamental attribution error. Winning is a stochastic function of the players skills. Given a rating we can expect high correlation between skills and given similar ratings the same.
The fundamental attribution error is a psychological issue, wherein we can see our own context, but not the context of others. When you play the game and someone in your game "feeds", that is you making the fundamental attribution error. It could be that they got counter-picked some bad luck and did the best they could. It could be they had connection errors. It could be any number of things. But you don't see them so that you fundamentally misattribute their failure to them wherein you would attribute your failure in that situation to the situation.
What we are dealing with here is induction. To argue against it we pretty much have to argue that induction doesn't work. Cue failure of the backing of science. Basically we are asking the question "if we have observed what you have done in the past, can we reasonably take it to mean how you will perform in the future". Since before the game the game is symmetric if it turns out that the team has two junglers losing due to that is indicative of your skill in the game, because the game required you to play something else and you could not do it.
Skill in this game is not a one-dimensional attribute. It is multi-dimensional, and there's plenty of interplay between those dimensions.
Yes, which is why i described it as multi-dimensional. However those dimensions are correlated. Which is to say that while there may be differences in skill level between one person as a jungler and the same person as a laner this means nothing for our statical inference which still works when considering overall skill.
Madpandasuburbs west of chicagoRegistered Userregular
edited January 2012
I don't think I am jungling right. I've tried it with Rammus and Udyr most recently, and a few others. The problem I seem to be having is unless I am spending 100% of my time going from mob farm to mob farm, I get outleveled/golded by team-mates. Even if I dont drop into the lanes now and then to assist I just seem to be progressing really slowly and would be better off in a lane.
This is in co-op vs ai matches.
I am taking smite and using jungling builds from solomid. I usually try jungling because the majority of builds for a character I want to try are for jungling so I figure that is their strong point.
I don't think I am jungling right. I've tried it with Rammus and Udyr most recently, and a few others. The problem I seem to be having is unless I am spending 100% of my time going from mob farm to mob farm, I get outleveled/golded by team-mates. Even if I dont drop into the lanes now and then to assist I just seem to be progressing really slowly and would be better off in a lane.
This is in co-op vs ai matches.
Jungling in coop vs Ai is almost always going to lag behind the lanes. In Coop vs Ai your lanes are going to be killing the enemy and gaining exp/gold faster than they would typically in a normal game. Also you won't typically have many ganking opportunities (since the lanes are in farm champion mode) and you will naturally fall behind.
Use Coop vs Ai or custom games to get jungle routes down, but aside from that you won't really get a good feel for jungling unless you do normal/ranked games.
NNID: delphinidaes Official PA Forums FFXIV:ARR Free Company <GHOST> gitl.enjin.com Join us on Sargatanas!
I don't think I am jungling right. I've tried it with Rammus and Udyr most recently, and a few others. The problem I seem to be having is unless I am spending 100% of my time going from mob farm to mob farm, I get outleveled/golded by team-mates. Even if I dont drop into the lanes now and then to assist I just seem to be progressing really slowly and would be better off in a lane.
This is in co-op vs ai matches.
I am taking smite and using jungling builds from solomid. I usually try jungling because the majority of builds for a character I want to try are for jungling so I figure that is their strong point.
It's unlikely you will keep up with people in bot games. I find on Shyvanba I tend to be right behind mid/top so long as I am focused on rapidly attaining six. Then I go gank/steal their jungle. It's easy to fall behind as jungler even when you're not trying it in a game where every lane feeds on kills. In actual games the lanes don't level as fast unless they are doing well, and you can often find yourself getting higher than your team if you are generating kills and they aren't doing so hot when you are not around.
Next time I am playing I need to experiment more with thieving some CS from middle lane or new paths. Shyvana tneds to be one camp clear from six right about the time she needs to run back to golems with stonewall's path. A creep or two middle lane will do the same thing. A gank middle lane before six, hit six, steal blue, gank bottom sounds nice.
The goal they stated was that if a jungler spent his time clearing camps, then he would be somewhere inbetween the solo lane and the duo lane in terms of experience. Remember in an AI game that you have two jungles worth of experience to roam and that the buffs are where the majority of that experience lies.
0
BethrynUnhappiness is MandatoryRegistered Userregular
No, it makes a fundamental attribution error, in considering all of those variables as things that the player has control of and factor into their skill.
No, this is not a fundamental attribution error.
No, it's the very definition of a FAE. You're attributing an error to the player that cannot be attributed to them. Sorry, but this has bugger all to do with induction; it's an FAE plain and simple.
Winning is a stochastic function of the players skills. Given a rating we can expect high correlation between skills and given similar ratings the same.
Unless I'm misunderstanding you here, this prediction also makes zero sense at lower elos (it holds true for higher, more consistent elos).
The fundamental attribution error is a psychological issue, wherein we can see our own context, but not the context of others. When you play the game and someone in your game "feeds", that is you making the fundamental attribution error. It could be that they got counter-picked some bad luck and did the best they could. It could be they had connection errors. It could be any number of things. But you don't see them so that you fundamentally misattribute their failure to them wherein you would attribute your failure in that situation to the situation.
No, that's not what I'm talking about. I'm talking about the attribution of the system that because the team failed, the player must have failed.
Yes, which is why i described it as multi-dimensional. However those dimensions are correlated. Which is to say that while there may be differences in skill level between one person as a jungler and the same person as a laner this means nothing for our statical inference which still works when considering overall skill.
Except that if an excellent jungler only plays as a jungler, you're no longer measuring "overall skill", are you?
3s transfer and 5s deathrays are mean in fights when you have 400+ ap, and his stun field is only about 6s as well.
Except lots of ap carries does that too, and better. I really really tried with Viktor but I can't find any appeal in him. If he was tankier or had longer range he might work. The only thing he has that no one else has much better is his stun.
I dunno, he isn't top pick or anything but he can skirt around the outer edges of team fights firing off deathrays and dropping big slow/stun circles in the middle of enemy formations. You've got an aoe silence to drop that is also aoe damage and can follow anyone who drops low, and the damage and mitigation of power transfer really adds up with such a low cd.
I'd say he could be on par with Annie. Less initiation power but stronger over the long run.
People complain about Viktor because he needs to get kind of close
To which I say- Augment your W, not your E. The range on that thing gets downright brutal. Wait for the right time, drop that after the tank initiates and you will most likely catch the entire team in the stun, then you are free to walk up and silence whoever you want while softening them up with death ray and power transfer
Also build tanky AP items. I start with a RoA as soon as possible and I usually become really hard to kill after that thing ramps up.
I still do say they need to do something about his Q. My best idea is lower the damage a bit, increase the range a bit, and then let it shield any allies it hits on the return path
Posts
The modelling for elo based on the result of the sampling could be perfect and it'd still have 5-9 confounding variables in every sample, namely the 5-9 other players. And this is before you get into roles, teamcomps, and so on.
The only thing inaccurate about the concept of "elo hell" for example, is that you cannot get out of it. The lower the elo, the greater the effect these confounding variables have because of the ability of 1 player to throw a game for their team; the higher the elo, the more normal they get because players approach the baseline of skill in all areas generally necessary not to totally throw a game.
You can look at a player as having a set of "skills":
1) Map awareness. Being able to gauge what enemies might be doing when they're missing, as well as being aware of what it means when they're visible (jungler ganking top? We're safe for aggression for a minute at bot/I can steal his bot buff camp if it comes up).
2) Role awareness. Knowing what they're supposed to be doing (supports who don't realise they're supposed to ward and ward clear).
3) Role capability. Doing what they're supposed to be doing (tanks who don't initiate/softzone enemy damage dealers well, damage dealers with bad focus or scared of getting focused).
4) Champion building. Knowing how to build a champion, how to structure a build based on what's going on in your game, what runes/masteries/summoners to take against the champions you're facing.
5) Laning capabilities. A lot of people who are just bad at positioning and last hitting in lanes, always a step too far away to get a last hit in time, facechecking bushes, and so on.
6) Mistakes. Just plain ol' "how many mistakes will this player make"? Are they greedy, cowardly or blind?
And as you get higher in elo, most players will come to around the same level of "skill" at each of these things. But at lower elos, you'll find people with much different skillsets.
For example, the dudes who are brilliant in lane but have no map awareness and are way too brave in teamfights. Leave them to 1v1 and they'll dominate, but they'll always get ganked if the jungler/other lanes try it, and in teamfights they'll get focused because they're out in front too much. Or the dudes who know exactly what's going on where on the minimap, but play way too conservatively in lane and lose half of each wave's last hits while the enemy laner(s) have control of the lane and when to gank. Or the junglers who aren't really aware of their own buff spawn timers, thus ganking lanes at the wrong time and coming back to find the enemy jungler has stolen their buff because that jungler knows where they were at the time the buff would come back up at.
And so on.
:^:
http://www.leaguereplays.com/replays/match/144855/
I didn't teamfight or gank or anything. Maybe someone here has some input on what I could have done to make this matchup easier for myself, because I know I could have played that a lot better. Basically input for the first 10 minutes of laning.
We were doing some work in those games, though I think our first set of opponents was of a spectacular class of baddie. We ought to do it again sometime.
I don't play a lot of Lee Sin, but WWQQE on the enemy tends to do wonders.
Irelia is a bit like Udyr when it comes to the laning phase; they're almost impossible to play poorly and they either dominate a match-up or are countered hard. I find LS pretty good at farming at the tower, though, so it shouldn't be too much of a problem.
spoilered for wall o' text
I run ArPen marks, AD quints, flat AR seals, and MR/lvl glyphs (though I'm thinking about mixing in a few AD marks). For masteries I run 21/6/3, taking the left side of the offense tree (except crit) with 2 points in CDR and 3 in Havoc. Armor/mr and mana pool for the other points.
I start boots+3pots or boots/ward/pots and B sometime after I have 950g to buy 2 dblades. After that I like to get zeal>BT>finish PD>IE> then see what I need. If someone with a lot of armor is giving me fits then I'll build a LW as my 3 major purchase. 4th purchase is usually a defensive item against whoever is doing best on the enemy team, then finish off the build with some combination of PD, BT, and IE. Getting 2pd, 2BT, IE, LW is a thing of joy and beauty that I only managed to do once, but I could melt teams when I did.
In lane play safe until level 3 or so, taking q,e,q. Q enemies when you have a good shot at them with minimal creep in the way and auto attack them if they come into range. You are strongest against combo mages like LeBlanc and Morgana, and weakest against tanky AD types with gap closers like Jax and Xin. A good Teemo will just eat you, so ask for a lane swap. You must must MUST learn to be super quick with hitting your e, most champs have some sort of tell if they are about to hit you with an ability, and you need to be able to react to that instantly in order to gain mana. And you will be short on mana. In general, get ready mentally to hit e if a champ that has been hanging back suddenly walks toward you.
Later in the game you are a fantastic split pusher, a QE combo can clear out an entire minion wave. Watch the mini-map, when the enemies start vanishing then it's time to get the hell out of dodge. I like running TP so I can split push then tp to a different lane or to a team fight. It can also be used to escape, since it is faster than B and doesn't seem to be interrupted by damage. You can outrun everyone except Singed, Yi, Teemo, and sometimes Nid. A good Ashe can also ruin your day with the constant slow.
In team fights, stay behind the tougher people on your team. Skirmish with Q, focus on their squishy types when the team fight erupts. Save R for when the team fight involves everyone, don't lead off with it in the skirmish part.
3s transfer and 5s deathrays are mean in fights when you have 400+ ap, and his stun field is only about 6s as well.
Except lots of ap carries does that too, and better. I really really tried with Viktor but I can't find any appeal in him. If he was tankier or had longer range he might work. The only thing he has that no one else has much better is his stun.
Learn where and when to die. Learn to hit single targets with q.
Get Deathcap early for ult trolling. Build WotA when you want to be nice to your team, extra archangels when you just want to run around with defile on all day. Karthus charges a Tear incredibly fast if you're willing to just chuck out qs every few seconds.
Pair up with someone with strong cc to keep people harmless in the range of Defile. Leona is nice for it. And don't be afraid to flash offensively. Suddenly having a Karthus in the middle of a team is an unpleasant experience for everyone involved.
I dunno, he isn't top pick or anything but he can skirt around the outer edges of team fights firing off deathrays and dropping big slow/stun circles in the middle of enemy formations. You've got an aoe silence to drop that is also aoe damage and can follow anyone who drops low, and the damage and mitigation of power transfer really adds up with such a low cd.
I'd say he could be on par with Annie. Less initiation power but stronger over the long run.
I'm trying to be hopeful. Is hard.
twitch.tv/tehsloth
FFBE: 898,311,440
Youtube: http://www.youtube.com/dElementalor
Unless their Orianna is simply bad, that's not a good lane for your team.
EDIT: First ranked game where my teammates were not only competent, but actually good. Things are lookin' up.
twitch.tv/tehsloth
For sure. I'm really happy I got those GP5 seals.
Juking skillshots is pretty important if you want to go mid. Tumble does wonders for that.
Edit: Condemning someone on a wall isn't a necessity for vayne, although if you do get one it's just about a guaranteed kill.
FFBE: 898,311,440
Youtube: http://www.youtube.com/dElementalor
3rd game, my internet lags out. Guess who I random? Mundo!
Any tips on Jungle Mundo?
#1: ArmPen/Armor/CdR/AD runes using 21/9/0 = Got lucky with some crits on Blue but Red nearly killed me. Level 4 by 4:50 is not good.
#2: AS/Armor/AS/ArmPen runes using 9/21/0 = So much easier! Got down to 130hp on the golems but I was able to clear the path by 4:05.
Masochism makes up for not having much AD early on so there's no need for the runes. I think it's possible to even start with regrowth pendant instead of cloth+5.
BLACK FLAGS ON THE HORIZON
GREAT THUNDER AND BURNING RAIN
HIS WARSHIP IS COMING
I HEARD THE LAST PICKS SCREAM HIS NAME
I WATCHED AS HE RALLIED
TO THE DRAGONS WHO DIED THAT DAY
HE HELD UP HIS SABER HIGH
AND BLUE PILLED SO HE COULD BUY
GANGPLANK, THE MIGHTY
GANGPLANK, THE BRAVE
CRUSH THE NIDALEES IN YOUR WAY
BY YOUR PARRRLEY LET NONE BE SAVED
LIVE TO CRIT IN THE PIRATE WAY
SORAKAS, WARWICKS AND SHENS
WE'LL BARON TOGETHER
IN THE END
GOD OF FARMING, GOD OF RANKS
PUB STOMPER WHO FEARS NO GANKS
THE POWERHEAD OF THE METAGAME
NOW ULT FOR NEVER ENDING BLAME
SHOOT YOUR PISTOL TO CRACK THE SKY
LIFT YOUR CAPE SO THAT YOU MIGHT FLY
BACK TO PORT FOR THE WHORES ON HIGH
AND CURE YOUR SCURVY WELL
Not even 200 CS and we get all that?
Starcraft does not use a Bayesian system, they likely use something that approximates a Bayesian system. Every system is bayesian in that its an updating system built on Bayes rule that attempts to generate a normal distribution. Trueskill is similarly not Bayesian but rather algorithmic. This is not a bad thing, its just not strict distribution updating. But strict distribution updating probably isn't possible anyway on the scale we're talking about. Any system that updates your rating based on whether or not you won or lost a game is Bayesian. Elo was just one of the first sorting algorithms(and was relatively simple. E.G. if you look at true-skill it reduces to Elo when players have the same std)
Adding the sigma doesn't really give us any more information once you have played a sufficient number of games. This is because, so long as we believe that players can get better or worse there is a hard minimum cap on your standard deviation as measured by the game. We can assume this hard minimum cap both by simply starting with the same prior distribution for players who have not played a game and running the algorithm for the game number cut off that we believe is no longer representative, but also because the assumption of a moving rating means that we must have a minimum uncertainty about your current rating.
Having a sigma is useful only really for putting people into leagues, so people don't bounce between leagues as easily and so that we don't put people in leagues above where they are until we really believe they can handle it.
So lets get some things straight
1) Before champion selection LoL is symmetric
2) Winning/Losing takes into account all of those factors
3) player skills, while sometimes disparate across areas, tend to be highly correlated. I.E. it is possible to be a good laner and a bad jungler, but the better you lane the better a jungler we expect you to be.
Because of this fact, we absolutely can go backwards from "did you win or lose" to "how good a player are you".
And unless you've played a very low amount of games you're going to be placed with a reasonable accuracy.
This is exactly what I was objecting to. The idea that you can lump in every aspect of the skills needed to be a perfect player to determine whether someone is "good" or "bad" as if it accurately presents their skill.
Skill in this game is not a one-dimensional attribute. It is multi-dimensional, and there's plenty of interplay between those dimensions.
If you've seen a top player play a role they're very unfamiliar with (saint did quite a fun experiment while his account was banned, which was to play a role he was unfamiliar with from starting elo; you could easily see he wasn't performing at the level he does on a jungler) you can see where elo starts to fall down when you're trying to assert that it's an accurate representation of a player's individual skill.
No, this is not a fundamental attribution error. Winning is a stochastic function of the players skills. Given a rating we can expect high correlation between skills and given similar ratings the same.
The fundamental attribution error is a psychological issue, wherein we can see our own context, but not the context of others. When you play the game and someone in your game "feeds", that is you making the fundamental attribution error. It could be that they got counter-picked some bad luck and did the best they could. It could be they had connection errors. It could be any number of things. But you don't see them so that you fundamentally misattribute their failure to them wherein you would attribute your failure in that situation to the situation.
What we are dealing with here is induction. To argue against it we pretty much have to argue that induction doesn't work. Cue failure of the backing of science. Basically we are asking the question "if we have observed what you have done in the past, can we reasonably take it to mean how you will perform in the future". Since before the game the game is symmetric if it turns out that the team has two junglers losing due to that is indicative of your skill in the game, because the game required you to play something else and you could not do it.
Yes, which is why i described it as multi-dimensional. However those dimensions are correlated. Which is to say that while there may be differences in skill level between one person as a jungler and the same person as a laner this means nothing for our statical inference which still works when considering overall skill.
to win, don't lose
This is in co-op vs ai matches.
I am taking smite and using jungling builds from solomid. I usually try jungling because the majority of builds for a character I want to try are for jungling so I figure that is their strong point.
Steam/PSN/XBL/Minecraft / LoL / - Benevicious | WoW - Duckwood - Rajhek
Just giving you a hard time since you're never around the PA chat. :P
Jungling in coop vs Ai is almost always going to lag behind the lanes. In Coop vs Ai your lanes are going to be killing the enemy and gaining exp/gold faster than they would typically in a normal game. Also you won't typically have many ganking opportunities (since the lanes are in farm champion mode) and you will naturally fall behind.
Use Coop vs Ai or custom games to get jungle routes down, but aside from that you won't really get a good feel for jungling unless you do normal/ranked games.
Official PA Forums FFXIV:ARR Free Company <GHOST> gitl.enjin.com Join us on Sargatanas!
It's unlikely you will keep up with people in bot games. I find on Shyvanba I tend to be right behind mid/top so long as I am focused on rapidly attaining six. Then I go gank/steal their jungle. It's easy to fall behind as jungler even when you're not trying it in a game where every lane feeds on kills. In actual games the lanes don't level as fast unless they are doing well, and you can often find yourself getting higher than your team if you are generating kills and they aren't doing so hot when you are not around.
Next time I am playing I need to experiment more with thieving some CS from middle lane or new paths. Shyvana tneds to be one camp clear from six right about the time she needs to run back to golems with stonewall's path. A creep or two middle lane will do the same thing. A gank middle lane before six, hit six, steal blue, gank bottom sounds nice.
Unless I'm misunderstanding you here, this prediction also makes zero sense at lower elos (it holds true for higher, more consistent elos).
No, that's not what I'm talking about. I'm talking about the attribution of the system that because the team failed, the player must have failed.
Except that if an excellent jungler only plays as a jungler, you're no longer measuring "overall skill", are you?
People complain about Viktor because he needs to get kind of close
To which I say- Augment your W, not your E. The range on that thing gets downright brutal. Wait for the right time, drop that after the tank initiates and you will most likely catch the entire team in the stun, then you are free to walk up and silence whoever you want while softening them up with death ray and power transfer
Also build tanky AP items. I start with a RoA as soon as possible and I usually become really hard to kill after that thing ramps up.
I still do say they need to do something about his Q. My best idea is lower the damage a bit, increase the range a bit, and then let it shield any allies it hits on the return path