NeoGAF

Wok · 09:47 AM

Trying to learn a ELO score based on a match history of length 20 sounds so dumb.

Honestly, let the guy apply the algorithm to another game for which:

the Elo score is known,
the outcome of the match depends on a player, not on a team with players being assigned different roles,

such as Chess or Scrolls.

This heuristic method would dramatically fail in both cases, which are easier to validate and easier to predict than the ELO score of Dota 2.

Just imagine classifying chess games between pro and novice based on the game duration, the number of pieces left at the end, which pieces were taken, etc, plus a game history of length 20...
I would be more confident in features such as at what time points in a game do the kills happen, and the duration between each kill.
Anyway, the clustering would fail, and hence the predicted pseudo-ELO score would be meaningless.

Cyborgmatt · (Today, 10:00 AM)

Originally Posted by Drkirby

BZM, is the guy who made Mydotaskill trolling us? Does the maths he is talking about actually do the things he says they should? It sounds like he is talking technobabble and not real mathematics.

This is why I said this site is just fluff. If he doesn't want to talk about his formulas at all then it's pretty obvious that he isn't confident of how they would hold up to scrutiny.

ElyrionX · (Today, 10:19 AM)

Originally Posted by Wok

Trying to learn a ELO score based on a match history of length 20 sounds so dumb.

Honestly, let the guy apply the algorithm to another game for which:
the Elo score is known,

the outcome of the match depends on a player, not on a team with players being assigned different roles,

such as Chess or Scrolls.

This heuristic method would dramatically fail in both cases, which are easier to validate and easier to predict than the ELO score of Dota 2.

Just imagine classifying chess games between pro and novice based on the game duration, the number of pieces left at the end, which pieces were taken, etc, plus a game history of length 20...
I would be more confident in features such as at what time points in a game do the kills happen, and the duration between each kill.
Anyway, the clustering would fail, and hence the predicted pseudo-ELO score would be meaningless.

Except Dota *is* a team game and the guy is trying to account for the team aspects of it in his model.

Nirvana · (Today, 10:21 AM)

Originally Posted by Divvy

The only thing I want from Valve for Xmas:

༼ つ ◕_◕ ༽つ GIVE LANGUAGE RESTRICTED SERVERS

'Please recite this brief passage from Macbeth, this will allow you to join this server'

snack · (Today, 10:23 AM)

I'm watching arteezys twitch stream and he has Korean announcers. How'd he do it?

Spookie · (Today, 10:25 AM)

Originally Posted by Cyborgmatt

This is why I said this site is just fluff. If he doesn't want to talk about his formulas at all then it's pretty obvious that he isn't confident of how they would hold up to scrutiny.

Not been to his site, but I'm sure that fluff might be making him a bit of dough via ad views surely?

ViviOggi · (Today, 10:25 AM)

Originally Posted by snack

I'm watching arteezys twitch stream and he has Korean announcers. How'd he do it?

You just have to change the game's language to Korean in Steam. This only affects the standard announcer and hero voiceovers, everything else stays in English.

Puck is best Korean voice

snack · (Today, 10:26 AM)

Originally Posted by ViviOggi

You just have to change the game's language to Korean in Steam. This only affects the standard announcer and hero voiceovers, everything else stays in English.

Sweet. Many thanks kind sir!

Haly · (Today, 10:27 AM)

He describes his models as "hyperdimensional", which I take to mean it's more than just a line plot and he's factoring game time as well.

I agree with Cyborg in that he is clearly not confident in his method enough to reveal the formulae (although he might want to preserve it for commercial use), and the constant changes support this as well. This is very much a work in progress and DOTA2 players are the guinea pigs.

Anbokr · 10:48 AM

Originally Posted by Haly

He describes his models as "hyperdimensional", which I take to mean it's more than just a line plot and he's factoring game time as well.

I agree with Cyborg in that he is clearly not confident in his method enough to reveal the formulae (although he might want to preserve it for commercial use), and the constant changes support this as well. This is very much a work in progress and DOTA2 players are the guinea pigs.

Of course, I mean he's one of the first to attempt a ranking system since Valve crushed any hope of an elo based one with the whole dotabuff fiasco. I'm sure he's constantly refining the algorithm as he finds better ways to attempt to quantify skill amidst all the restrictions and private data. Like green biker dude said, you can't tell the difference between a 1500 and a 1600, or hell, probably even a 1200 and a 1600; but the system doesn't seem bad for general tiers. People are (generally) ranked around where you'd expect them to be, and something like DSR (granted when it stops getting changed so frequently) could be a nice temporary solution for people to be able track their progress on the legendary quest to escape the trench.

Also he could simply be hiding his formulas/method in order to protect it. Valve crushed DBR fairly quickly and if he was completely transparent with how he calculates DSR, Valve could attempt to find a way to quash that or throw a kink in there.

I mean all this DSR crap just reaffirms that people want a legit ranking system and it brings to the forefront, once again, something that DOTA 2 is sorely lacking. People are so desperate for some sort of rating that they'll flood some random dude's site without knowing jack shit about his methods/formulas, it speaks volumes that we're even having this debate and that DSR (an unknown ranking from an unknown author on a previously unknown site) has already blown up in just a few days time.

DOTA 2 is missing just 3 features to be the ultimate game IMO (in order of importance):

1) Parity
2) Ranked matchmaking
3) Custom maps

Finish up those three and I honestly can't see myself playing very many other games for the next 10 years. #1 is coming along sluggishly (though we'll get there eventually), and no word on #2 or #3. Hopefully the next couple big updates give us something in these departments.

Hylian7 · (Today, 10:43 AM)

Originally Posted by Drkirby

BZM, is the guy who made Mydotaskill trolling us? Does the maths he is talking about actually do the things he says they should? It sounds like he is talking technobabble and not real mathematics.

He seems to imply that he was actually putting some stock in the "VOLVO FORCES 50% WINRATE!!!" argument.

This guy has to be trolling.

Haly · 11:23 AM

Originally Posted by Hylian7

He seems to imply that he was actually putting some stock in the "VOLVO FORCES 50% WINRATE!!!" argument.

This guy has to be trolling.

English is clearly not his first language:

In one way or another, they force a 50% winrate by sometimes making you take really unfair matches. I highly doubt it.

He contradicts himself in this same sentence so it should be taken with a grain of salt.

However he is correct that, speaking in the most technical sense, Valve is forcing a 50% win rate by creating games where both teams have an even chance of winning. Over a long enough period of time, or if Valve creates the perfect DOTA2 matchmaking system, all win rates of all players will eventually tend toward ~50%, barring outliers such as smurfing or people intentionally losing.
Retracted again, because I can't decide if this is a random walk.

I'm reading the convo between him and other math nerds on reddit and it's pretty nerdy.

sephi22 · (Today, 11:12 AM)

HNNNNGGG....
Preorders uncancelled

bounchfx · (Today, 11:20 AM)

ok that's pretty damn awesome

Nirvana · (Today, 11:35 AM)

Any Australian gaffers around? Need some more dotoers to queue with.

M.D · (Today, 12:17 PM)

You know, solo que is not all that bad!
I refused to solo que till now, but last week I finally gave in after no one was playing, and I've been doing it ever since!

Some people actually support when picking a support hero, and not everyone picks a carry and call you names and shit

ElyrionX · (Today, 12:49 PM)

I just started incorporating shift-queueing into my game and I only just realized how much auto-attacking can screw up your queued commands.

Say, I'm using Puck and want to blink into a group of enemies and immediately follow this up with a Waning Rift. If I don't shift-queue the Waning Rift quickly enough, the hero would actually start attacking the enemies around and wouldn't cast Waning Rift until all visible enemies are dead. This is because auto-attack actually queues the attack command before my Waning Rift enters the command queue. Is my understanding of this correct? If so, this must be a major reason why some people turn off auto attacks.

ElyrionX · (Today, 12:56 PM)

Originally Posted by Anbokr

DOTA 2 is missing just 3 features to be the ultimate game IMO (in order of importance):

1) Parity
2) Ranked matchmaking
3) Custom maps

Finish up those three and I honestly can't see myself playing very many other games for the next 10 years. #1 is coming along sluggishly (though we'll get there eventually), and no word on #2 or #3. Hopefully the next couple big updates give us something in these departments.

2 and 3, yes. But I seriously don't think parity is important at all. The game has more than a hundred heroes as it is. If they release a few new heroes every year and ignore the heroes from Dota 1 that have yet to make it into Dota 2, I honestly couldn't care less. And this is coming from someone who played a ton of Techies back in the Dota 1 days.

Sanjay · (Today, 01:04 PM)

Originally Posted by sephi22

HNNNNGGG....
Preorders uncancelled

That is bloody good.

DarkKyo · (Today, 01:06 PM)

Originally Posted by sephi22

HNNNNGGG....
Preorders uncancelled

What are these from??

Arkanius · (Today, 01:17 PM)

Originally Posted by DarkKyo

What are these from??

WETA items

mfaex · (Today, 01:26 PM)

That Vanguard is extremely cool. Best statue too imo.

Trasher · (Today, 01:32 PM)

Originally Posted by Anbokr

Of course, I mean he's one of the first to attempt a ranking system since Valve crushed any hope of an elo based one with the whole dotabuff fiasco. I'm sure he's constantly refining the algorithm as he finds better ways to attempt to quantify skill amidst all the restrictions and private data. Like green biker dude said, you can't tell the difference between a 1500 and a 1600, or hell, probably even a 1200 and a 1600; but the system doesn't seem bad for general tiers. People are (generally) ranked around where you'd expect them to be, and something like DSR (granted when it stops getting changed so frequently) could be a nice temporary solution for people to be able track their progress on the legendary quest to escape the trench.

Also he could simply be hiding his formulas/method in order to protect it. Valve crushed DBR fairly quickly and if he was completely transparent with how he calculates DSR, Valve could attempt to find a way to quash that or throw a kink in there.

I mean all this DSR crap just reaffirms that people want a legit ranking system and it brings to the forefront, once again, something that DOTA 2 is sorely lacking. People are so desperate for some sort of rating that they'll flood some random dude's site without knowing jack shit about his methods/formulas, it speaks volumes that we're even having this debate and that DSR (an unknown ranking from an unknown author on a previously unknown site) has already blown up in just a few days time.

DOTA 2 is missing just 3 features to be the ultimate game IMO (in order of importance):

1) Parity
2) Ranked matchmaking
3) Custom maps

Finish up those three and I honestly can't see myself playing very many other games for the next 10 years. #1 is coming along sluggishly (though we'll get there eventually), and no word on #2 or #3. Hopefully the next couple big updates give us something in these departments.

The problem with DotA is that there is no good way to quantify a rating. There are far too many factors and variables going on all at once. The one place where a rating does make sense is where it already exists in the game: TMM. At least in there you have a somewhat controlled sense of setup in having two teams actually working together. That isn't a given outside of TMM in normal MM. In regular MM you can have random mixed parties, people trying to teach their friends the game who are way below their skill level, people training on Meepo, people griefing their carries while playing support, and so on. There's no way to account for all the crap that goes on there.

The differences between TMM and regular MM are huge though. It's too bad that this is the state of the game, but the main issue is that the vast majority of players don't want to play in a TMM setting. Not only is it hard to even setup a team, but the pool of players to play against is too small for all of the different skill tiers. And I think that's what scares most of the lower skill-tiered players away from even giving TMM a shot. The TMM style of play is probably the best way to actually learn the game and get better though. Knowing how to actually play as a team is what most people are lacking in DotA. It is far too rare to actually find a good game in regular MM where you either A) don't have that selfish person/persons on your team that bitch and refuse to work with the team, or B) have someone on your team that virtually functions as cement shoes because they have no idea what's going on.

Anbokr isn't aware of these issues because he never tries to play with people below his skill tier. He only enjoys playing in try-hard 4-5 stacks where milkman carries his ass to victory.