Statistics

lump
81
Posts
499
Location
United Kingdom of Great Britain and Northern Ireland
Joined
7 May 2005
9 October 2009 - 04:32 CEST
#31
you'll need to include maths on positions and times of baiting, distractions, group decisions, individual decisions, etc, far too much
jiriki
176
old people
Admins
Provider & Webmaster
Posts
490
Location
Oulu, Finland
Joined
1 May 2009
9 October 2009 - 11:44 CEST
#32

Making useful formulas doesn't mean that you need to include everything. Many factors are atleast nearly normally distributed. Now there was actually a huge epistemological debate (called Methodenstreit) in the beginning of the 20th century in economics about using mathematics to measure human action, namely deduction and apriorism vs induction and positivism. But in the end, everything in the world uses induction. For example Newton's law are not accurate on subatom-level (or in singularity), because of the nature of quants, but for the sake of practical use of Newton's law, its totally irrelevant - it'd be same if there were little green men moving those atoms, because the formula is gives very accurate results on macro-level.

Microsoft's Trueskill which uses Bayesian probability mathematics, is able to predict match outcomes with very good accuracy (in Halo 3). Now I'd only have to apply the same for NS logs and see if it is equally as accurate.

Get to the spaceship!
Tane
91
HasBeens & NeverWas
Posts
57
Location
Jyväskylä, Finland
Joined
7 May 2005
9 October 2009 - 12:03 CEST
#33

I heard some people transformed everything to res in their head when doing tactics, e.g. every kill, every rt, players movement, players skill, every upgrade all of these has value in scale of res. I had always preferred to transform everything to time this helps a lot when you do tactics. I have always played against the clock rather than opponent. Obviously stats can't show full variation of this transform like I said before but frags are really big part of this transform. I'm arguing that you can see big part of person's skill and teamwork from stats if you know how to look at it. In the end teamwork and personal skill are just abilities which help you to win a round in ns.

And how you win round in ns? By destroying ips or hive which takes awful much of killing, destroying structures and building structures. You don't really need teamwork(at least not advanced teamwork) or personal skill in latter one before it turns to killing situations. So it seems that teamwork mostly lies in combat situation in which you kill and destroy. Couldn't we say that teamwork like personal skill is subordinated to killing and destroying? If this is truth stats would not only show personal skill but also teamwork because you can count killing. Sure player who is teamworker won't benefit from his ability but his team will which will make his team kill more and die less and this way he benefits. It's hard to believe that there would be a team where is one big teamplayer and rest are just fragwhores. You can see teamwork from team kills.

I would also like to say that you can't "fake" KPM as easily you could think. You get best KPM by being most effective which also interest of your team. You have find way were you are most aggressive but stay alive too. I would say that every "legendary" player has been as aggressive as possible. There is real problem in roles which mean KPM don't give you any information if you don't know how the subject plays. Anyway I liked to say that KPD is lot worse than KPM.

 

jiriki
176
old people
Admins
Provider & Webmaster
Posts
490
Location
Oulu, Finland
Joined
1 May 2009
9 October 2009 - 12:22 CEST
#34

Let me explain a bit how Microsoft's trueskill works. After a round it calculates the score of each player by taking into account two things:

  • The result of the match (win or loss)
  • Players of the teams (their score μ and distribution area σ)
Then it calculates the score and distribution area for each player using this information, and updates rankings. This means it doesn't include frags etc. just the list of players and result of the round, using then probability mathematics to calculate how much a player affects the game (be it frags, teamwork or commanding). This is why I find it a lot more useful than KPM or KPD.
Get to the spaceship!
lump
81
Posts
499
Location
United Kingdom of Great Britain and Northern Ireland
Joined
7 May 2005
9 October 2009 - 13:38 CEST
#35
playing just to res or just to time is almost as dumb as playing just for frags.. everything is important even when you include different playing styles.
lump
81
Posts
499
Location
United Kingdom of Great Britain and Northern Ireland
Joined
7 May 2005
9 October 2009 - 13:40 CEST
#36
I'm still not convinced in the slightest that any formulae can represent skill at all... especially teamwork.. i mean goddamnit it's just NOT a straightforward definition, never-mind scalable.
jiriki
176
old people
Admins
Provider & Webmaster
Posts
490
Location
Oulu, Finland
Joined
1 May 2009
9 October 2009 - 15:54 CEST
#37

They can still represent the probability of a victory for their team - mathematics mate.

pro tip: edit button

Get to the spaceship!
lump
81
Posts
499
Location
United Kingdom of Great Britain and Northern Ireland
Joined
7 May 2005
9 October 2009 - 18:48 CEST
#38

'represent' as in.. not show the real version but a  guess.... 

of the

'probability'.. as in.. the chances, a likelihood...

lol.. and tbh, I don't really believe that'd do that tbh, social effects have a much larger factor than previous wins..  You can argue for the case of anomalies but i'm pretty sure the anomalies should be less frequent than ALL THE TIME, lol..

Did maths work for psychology?  No... it doesn't, people are too prone to change, same for team sports when you have so many factors involved...

I mean shit, if getting strong probabilities for game outcomes is so simple, how can the stock market ever crash??  People would put far more time into the stock market than rts games :/ Why isn't there a no fail formulae for buying shares?  Because the numbers are TOO COMPLICATED TO REFLECT REALITY (especially when no one even knows the rules of the system!!).

jiriki
176
old people
Admins
Provider & Webmaster
Posts
490
Location
Oulu, Finland
Joined
1 May 2009
9 October 2009 - 20:51 CEST
#39

Firstly, learn to write your own language, without caps if possible.

Secondly, just because human action is involved, doesn't mean probability mathematics do not work, or are you telling you cannot predict (with distribution area ofc) how many car accidents are going to happen on some highway by looking at stats from previous years?

Stock markets are a whole different thing. I never said it was simple. You are making a fallacy of composition with your analogue anyway. After taking the Microsoft's Trueskill into use empirical evidence proved that Bayesian probability mathematics actually work. Experimental data tracking roughly 650,000 players over 2.8 million games support this claim: Not a single μ ever happened to be outside the range ± 4 times the initial σ and 99.99% of the μ's happen to be even within ± 3 times the initial σ.

There is no such thing as perfect information.

If you saw Quaxy losing twenty games to Saunamen, how could you even say Quaxy is worse than Saunamen without using probabilities? If you saw me dying 20 times to Kuitmo, how could you say I'm worse than him without using probabilities? We use induction in our daily life to produce information.

And guess what, you're likely going to respond to this post, because looking at your past behaviour, it is very likely. ;)

Get to the spaceship!
sublime
5
Posts
488
Location
United Kingdom of Great Britain and Northern Ireland
Joined
6 May 2005
9 October 2009 - 21:08 CEST
#40

I agree a lot of this should work Jiriki. But how much data do you actually have? For players who have played the fewest games, this iterative process to converge the skill rating towards the most likely outcome would surely be skewed? You would be effectively weighting practise and seriousness as the most important attributes. Furthermore these inactive players would represent a random function rather than a gaussian.

GibbZ
Noavatar
9L
Gather Moderators
NS1 Gather Mod
Posts
154
Location
Sierra Leone
Joined
7 May 2005
10 October 2009 - 00:57 CEST
#41

What about good players who decide to play for bad teams? How does it take the skill from that? I mean if a team loses most of it's games but they have the best player in NS, but the rest of the team is dogshit, can it still calculate an accurate skill?

I mean I feel sorry for people like Macpersil, who are actually decent, yet have lost many rounds playing with nibz. Playing with a bad team whatever your skill means you will underperform. Very rarely do you see one player evidentially carry, and the only times you really do are when it's a bad team vs a very bad team and a player is either fade or lerk, or in higher skilled games, maybe a shotgun carry. Mmmm

ZiGGY
815
Posts
66
Location
Earth
Joined
15 January 2006
10 October 2009 - 02:03 CEST
#42

it is quite common and effective for one player to deliberately perform an action that negatively effects their own personal value to the team in order to significantly increase the value of all the other players in the team. (Ex: baiting, tanking, pressuring etc)

though these tactical elements can be accounted for on a team level on an actual personal level it would be impossible to distinguish this variation between players, and all speculated models so far would not factor into account any proposed value of other units in the game.

thus the belief "stats are bs", stats are stats and will only tell you what they tell you but I believe the negative connotations should be quite clear from this surface comment on the elements of tactical/strategic awareness of players in game :P

jiriki
176
old people
Admins
Provider & Webmaster
Posts
490
Location
Oulu, Finland
Joined
1 May 2009
10 October 2009 - 02:16 CEST
#43

Sublime, actually Trueskill is a lot better than ELO in the number of games required to predict accurately. You can see from the page 7 of the specs the details.

Gibbz, it is specifcially designed to calculate probabilities for individual players from a team game. Like in NS, you always have some variation with whom you play with, which is enough to give information about your performance. Lots of similar questions are answers here.

Ziggy that is true for KPD and KPM, but when it comes to Trueskill, its not, because it calculates whether the round was won or not, not frags.

Get to the spaceship!
sublime
5
Posts
488
Location
United Kingdom of Great Britain and Northern Ireland
Joined
6 May 2005
10 October 2009 - 02:27 CEST
#44

Thats quite impressive Jiriki, especially the learning rate.

 

Guys this is a MODEL. It predicts more accurately than chance who will win, and better than any other model by the sound of it. It will never account for things which cannot be measured but if these actions truly change the bearing of the game they will be quantified in the results.

GibbZ
Noavatar
9L
Gather Moderators
NS1 Gather Mod
Posts
154
Location
Sierra Leone
Joined
7 May 2005
10 October 2009 - 02:53 CEST
#45

Also I found alien commanding makes my KPD go right down mainly as I'm concentrating on 6 players instead of just 1. :\

lump
81
Posts
499
Location
United Kingdom of Great Britain and Northern Ireland
Joined
7 May 2005
10 October 2009 - 03:04 CEST
#46

firstly, don't knock my writing of language on a random gaming forum, WHY WOULD I CARE HERE?  I put effort into my language at work, writing reports, in my study, this is a gaming forum and I'll write in prose for the hell of it without the use of gramma because I WANT TO.  If I feel the need to stress a point with caps, why whine?  It's not like you've never wrote in caps...  Stop trying to defend your point with irrelevant attacks, it's a sign of weakness.

Yes, the stats may tell you the likelyhood of a car crash ON ONE ROAD, with everyone driving CARS and no COMPETITION involved but they still don't tell you who is responsible for them..  NS is more complicated than driving down a road.  Believe it or not.

Can you please justify your arguments about stock markets being completely different instead of just randomly making statements please.  As far as I'm aware, stock markets are a competitive situation with a number of variables on a variety of playing fields and highly dependant on time and economy...  More similarities there than with cars driving down a road I feel.

Quaxy vs saunamen?.. ey?  We're not talking about team performances, we're talking about player skill.

jiriki vs Kuitmo?.. eh?  We're not talking about fragging, we're talking about player skill.

lump
81
Posts
499
Location
United Kingdom of Great Britain and Northern Ireland
Joined
7 May 2005
10 October 2009 - 03:16 CEST
#47
jiriki wrote:

Firstly, learn to write your own language, without caps if possible.

And guess what, you're likely going to respond to this post, because looking at your past behaviour, it is very likely. ;)

[facetious twat post]. let me help you lead by example.

 

  • don't start a sentence with 'And'.  it's bad PRAcTice

 

  • I'm likely?  Thankyou.  Don't use the same adjective or adverb twice in a sentence, it's bad practice.

 

[/facetious twat post]

And omg, of course I'm going to respond to this post, it's something I care about and I want my post count back.

Fana
Noavatar
Archaea
Donors
Movie Donator
Posts
291
Location
Oslo, Norway
Joined
6 May 2005
10 October 2009 - 15:15 CEST
#48

Why even reply to lump's posts? Do you enjoy wasting your time?

#archaea @ irc.quakenet.org

jiriki
176
old people
Admins
Provider & Webmaster
Posts
490
Location
Oulu, Finland
Joined
1 May 2009
10 October 2009 - 15:32 CEST
#49

Nope, that's why I'm not responding anymore. Besides, there's not really anything to add on substance level.

Get to the spaceship!
lump
81
Posts
499
Location
United Kingdom of Great Britain and Northern Ireland
Joined
7 May 2005
10 October 2009 - 15:43 CEST
#50

i like it; so you make an argument and i reply to the points directly; then i put an argument foward of my own.. but you ignore it and just turn it into another attack at me. GEGE

jiriki
176
old people
Admins
Provider & Webmaster
Posts
490
Location
Oulu, Finland
Joined
1 May 2009
10 October 2009 - 18:49 CEST
#51

Well, there's nothing ad hominem about ignoring posts that are waste of my time. I already gave the links if he wants to know more.

Get to the spaceship!
lump
81
Posts
499
Location
United Kingdom of Great Britain and Northern Ireland
Joined
7 May 2005
10 October 2009 - 19:44 CEST
#52

Stating that someone is a waste of time is very ad hominem actually, learn some respect. You ignored my arguments with a shitty unjustified judgement on them, kinda lame tbh.

I've read the links, they don't justify any realistic method of quantifying skill, it completely ignores psychological effects and game specific roles.  Basing skill purely on results over time doesn't work, not everyone plays for the same reasons.  Learn the difference between skill and likelihood of winning a game :/

jiriki
176
old people
Admins
Provider & Webmaster
Posts
490
Location
Oulu, Finland
Joined
1 May 2009
10 October 2009 - 20:40 CEST
#53

I never said you were waste of time, but I said those posts were. The former is an ad hominem, latter isn't. In any case, this reply looks a lot better.

On the subject: if a player does something that benefits the team's likelihood of winning the game, whether its fragging, commanding, teamworking, res biting etc. it will show up in statistics. This is not only about quantifying skill, this is about probabilitiies. If skill doesn't increase the likelihood of winning the game, then what is it good for? :)

Get to the spaceship!
lump
81
Posts
499
Location
United Kingdom of Great Britain and Northern Ireland
Joined
7 May 2005
11 October 2009 - 06:12 CEST
#54

My posts are an extension of me and stating my posts are a waste of time has implications.  Obviously if I say your posts are self righteous, without thought, without any knowledge of human behaviour, niave and oblivious to reality; it says nothing about you right? lol, grow up.

Well jiriki, this system assumes all players play the same class for a start, that's VERY far from the truth in NS where certain roles pull a lot more weight in the outcome than other roles:

1. You could be the best capping gorge in the world with the most detailed knowledge of the game around but if your team sucks at lerk, fade and onos... it wont do shit.  Repeat this 20 times and you'll seem like a shit player because you have a negative score from the jps who keep spawncamping you.

2. Be a very strong lerk and hold the team for the first 6 minutes, but your team never eats res and the rest of the game you're just sporing against sgs, hmgs with mt who have a seemingly endless supply of meds..  Will it looks like you're doing anything?  Probably not.

3. Tank for other players who don't respond, you're just killing yourself basically.

4. Certain teamwork based actions require the rest of your team to have the same understanding, if they don't..  It'll actually make the situation worse.

5. If you're the only person who parasites and SCOUTS on your team but again, no teamwork, it wont do shit while being an almost crucial skill based aspect of the game, 20 parrasited marines is worth more than 7-0 k:d imo.  But then, do the stats rate this as skill?  lol.

6. I've already stated about 3 other situatoins in previous posts that were ignored and GiBBZ stated a situation too.

7. The list goes on.

 

NB. I haven't even started on game psychology.

lump
81
Posts
499
Location
United Kingdom of Great Britain and Northern Ireland
Joined
7 May 2005
11 October 2009 - 06:27 CEST
#55

Heres something that I think makes my point PAINFULLY OBVIOUS

Q: I am always playing together in the same team with my friend JoeDoe. Will the TrueSkill ranking system be able to differentiate between us two in terms of skills? In other words, is the TrueSkill ranking system capable of finding that I am the more skilled player of us two?

A: If both you and your friend only play ranked team games together then the TrueSkill ranking system cannot distinguish between you two; it always compares the team's skills (sums of the player's skills in the teams) and 'distributes' the gain/loss proportional to the individual player's uncertainties (see detailed description).

 

I don't know where to begin with applying this to situations in NS that determine player skill, I could write essays.

jiriki
176
old people
Admins
Provider & Webmaster
Posts
490
Location
Oulu, Finland
Joined
1 May 2009
11 October 2009 - 13:04 CEST
#56

On the subject, it doesn't require you play the same class. But it is true, that if you are the best fade in the world and almost never fade, your score will be lower than it'd be if you faded always. However because of your amazing fading skills, the probability deviation σ will be higher.

1. Negative score from spawncamping is meaningless, because it doesn't count frags.

2. We're not talking about a single game here, but thousands of them. However if sporing increases your probability for winning the game, then it will increase your score.

3. Read 2

4. Yes, and if they are worse than you, the expected score will be lower.

5. Read 2

6. Read 2

7. Read 2

Much of that information was in previous posts, and now you might realize why I'm not really fond of replying to you. It doesn't help much either when you take it as a personal insult.

And you don't play NS with the same guys all the time? You forgot to paste the rest of the answer, let me do it for you:

But note: if your friend also plays team games with anyone other than you then the TrueSkill ranking system will be able to identify the more skilled player of your tw o. Also, if both of you always only play together, you might consider forming a clan.

Get to the spaceship!
GibbZ
Noavatar
9L
Gather Moderators
NS1 Gather Mod
Posts
154
Location
Sierra Leone
Joined
7 May 2005
11 October 2009 - 14:02 CEST
#57

Yea lump. Go form a clan. That always works lol!

ZiGGY
815
Posts
66
Location
Earth
Joined
15 January 2006
11 October 2009 - 20:32 CEST
#58

I hit this right on the nose, booyah! xD

also leave lump alone lol, he has charismatic leader bitch skills that make him good at making clans xD

sublime
5
Posts
488
Location
United Kingdom of Great Britain and Northern Ireland
Joined
6 May 2005
11 October 2009 - 22:46 CEST
#59

Jiriki if you look at most clans when activity was at an acceptable level, you have a group of 6 players - rarely more than 7-9 - who play with each other all the time. This means that you probably won't be able to distinguish between them, especially since the core 6 players are more likely to play than the backup ~3 players. I know you come from megahugeclan Quaxy and may not have experience of this, but in my experience I rarely played with more than 5 other people in a clan.

lump
81
Posts
499
Location
United Kingdom of Great Britain and Northern Ireland
Joined
7 May 2005
12 October 2009 - 05:05 CEST
#60
jiriki wrote:

On the subject, it doesn't require you play the same class. But it is true, that if you are the best fade in the world and almost never fade, your score will be lower than it'd be if you faded always. However because of your amazing fading skills, the probability deviation σ will be higher.

1. Negative score from spawncamping is meaningless, because it doesn't count frags.

2. We're not talking about a single game here, but thousands of them. However if sporing increases your probability for winning the game, then it will increase your score.

3. Read 2

4. Yes, and if they are worse than you, the expected score will be lower.

5. Read 2

6. Read 2

7. Read 2

Much of that information was in previous posts, and now you might realize why I'm not really fond of replying to you. It doesn't help much either when you take it as a personal insult.

And you don't play NS with text/javascript"> the same guys all the time? You forgot to paste the rest of the answer, let me do it for you:

But note: if your friend also plays team games with anyone other than you then the TrueSkill ranking system will be able to identify the more skilled player of your tw o. Also, if both of you always only play together, you might consider forming a clan.

Sigh, you still missed the point.  Please read my postS more carefully.

1. you missed the point, you lose.

2. I know, I'm not disputing that, but spores don't win games.  YOU CAN STILL LOSE EVERY GAME.

3. YOU CAN STILL LOSE EVERY GAME.

4. YOU CAN STILL LOSE EVERY GAME.

5. YOU CAN STILL LOSE EVERY GAME.

6. YOU CAN STILL LOSE EVERY GAME.

7. YOU CAN STILL LOSE EVERY GAME.

 

I already realised why you didn't want to reply, but I also realised you completely missed my point.  You're just assigning this model to a far more complex system than it's intended for and assuming that it will work.  And not only that, you're assuming it will show skill level which is a highly ambiguous personal factor and can't even be accurately defined for this game, never mind quantified!!!!

New Reply