-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
Heroes of Newerth |OT| Free-To-Play [PC /Mac /Linux]
- Thread starter blitz64
- Start date
the bad thing about the trueskill algorithm is when people KNOW it's being used and they can't affect their rating as much, they tend to play less. It also fails to take into account changes of skill over the course of the rating period. If there were enough games being played to support shorter season (2 weeks to a month) i would like it more.
For the old Elo formula, the KFactor is at a constant 32.bzm said:
The Gaussian's KFactor is based on your rating. FIDE (World Chess Federation) changes their KFactor based on how many games you play and your rating. It should change based on your rating and how many games you play, but right now it just lowers itself if your rating is >2400. Probably need to add that, but I doubt it would change a person's rating by that much.
Not necessarily. It would be true if TrueSkill just uses the mean (μKapura said:
as the score. It doesn't. It uses a "conservative skill estimate" that uses the 1% lower quantile of the graph. A lower standard deviation (σ lowers your TrueSkill score since the formula is μ - 3 * σ.Yeah, so the graph becomes more certain. That means that if you started playing in the league, played 50 games and then learned how to ward, orbwalk, and land skillshots, you couldn't move up to players closer to your skill level. If we didn't have as many new palyers, or if the game didn't change so often, or if the seasons were shorter, I'd be more down to use trueskill.
I know growth isn't instantaneous, but some of the growing occurs outside of the data set. People playing like 2 games in tmm for every 1 game in league (conservative) will do a lot of growing outside of the league. I was talking to bzm, and I like trueskill in other situations but not heredelirium said:
I think Kapura thinks that sigma growth is linear or something. It's not. It's based on several factor (including the "fair" the match is) and is bounded.
EDIT: Apparently there is TrueSkill alternative that Microsoft published. TrueSkill Through Time (paper: http://research.microsoft.com/pubs/74417/NIPS2007_0931.pdf) factors later scores back into earlier calculators. I'm still reading through it, but it might address your problems with the algorithm Kapura.
EDIT: Apparently there is TrueSkill alternative that Microsoft published. TrueSkill Through Time (paper: http://research.microsoft.com/pubs/74417/NIPS2007_0931.pdf) factors later scores back into earlier calculators. I'm still reading through it, but it might address your problems with the algorithm Kapura.
kinda, yeah. It locks people down a lot. I really like it when it's being used to rate players of similar skill level, but not across broad swaths of the palyerbase. The alternative is using shorter periods of data collection, so people would be ranked where they play over a two-week period because let's face it, there can be a lot more variation in two months.delirium said:
Doesn't that make sense though? Skills just don't show up one day, a few good games doesn't mean that you're better, playing good steadily over a decent amount of games does.Kapura said:
I would rather compare the kpop wankfest group to the weeaboos if they were a high school clique.Halycon said:
Weeaboos are primarily 2D.delirium said:
Yeah true I guess, the groups overlap a lot. Just like how groups from both sides play HoN!
HoN is clearly a microcosm of high school life.
(I never said it wasn't, I just said it was nerdier.)
Again though, I feel a bunch of the players are going to improve significantly over the next few months. I mean, go ahead and keep tracking it but i don't think that a season of this length with this much variation in player skill (legit 1300s [bzm] to legit 1900s [thor]) is the right place to apply it. Maybe we can take another look at the numbers in a month and decide which feels the most right.delirium said:
StickSoldier
Member
Got to ask, how the fuck are you supposed to beat Amun Ra? I can't figure out how to win against a team that has him.
StickSoldier
Member
Kapura said:
orangecaramel said:
Man, it's like I'm not even the one dying to him at the start, he just gets fed everytime. I'll try these suggestions, thanks.
StickSoldier
Member
Well I just faced him 3 times in a row after winning 8 straight, and lost 2 of the 3 because of Ra. Luckily the last game the Ra wasn't any good.
PuppetYuber
Banned
either shut him down early game or get bkb / puzzle box 3.
puzzle box 3 is probably the best counter.
puzzle box 3 is probably the best counter.
Anything that can kite him wins, Slither, Hag, Frostwolf Skull.StickSoldier said:
Easiest way to counter is, pick hag, and make Frostwolf Skull, he won't be able to get close to your team unless they have a Behemoth, Tempest, or Keeper, also you'll force him to go PK in order to initiate.
Ra has two ways of putting you on the clock.
1) Ganks. His ganking power is amazing and with a dedicated healer he can do it nonstop
2) Farming. Once he gets certain items he can farm nonstop.
Your only two options, without resorting to counter picks, is basically to counter-gank him or try to finish the game quickly.
Both Magebane and Witchdoctor hard counter him. Soft counters include anyone with a very long disable, or someone who can kite.
1) Ganks. His ganking power is amazing and with a dedicated healer he can do it nonstop
2) Farming. Once he gets certain items he can farm nonstop.
Your only two options, without resorting to counter picks, is basically to counter-gank him or try to finish the game quickly.
Both Magebane and Witchdoctor hard counter him. Soft counters include anyone with a very long disable, or someone who can kite.
Good Example of a counter Ra Team, they have 3 slows, 1 root, 1 stun, 1 aoe hold ultimate, and 1 huge AoE stun. Even though I was at 470 gpm wasn't able to get close enough to do damage.
Procarbine
Forever Platinum
Void talisman isn't the best idea, shrunken head is much more beneficial. You just have to slow/disable him and leave him for last in team fights. His resurrect won't matter as long as you have some disable when he respawns.
PuppetYuber
Banned
hard to lose when you are playing against swag.
Lone_Prodigy
Member
I find that Ra plateaus really fast. Early-mid game he's pretty much unstoppable if he has sufficient farm, but once the other team gets defensive items (Bulwark/Barrier/Shrunken) and ignore him in team fights, his power drops off considerably.