[H][P] which 1000:th post?

APM, skill and statistics

Copy paste from other post:
__________________________________________________________
1 game, or one player doesnt make any statistics. There are ofc good players with low apm,a dn the other way around.

But I am sure that if you look at the ICC rank (+win ratio) I am sure that you would find a strong corelation between high apm and high skill. I think all of you think so too. The average apm at B with >60% win ratio is for sure significantly higher than it is at D around 50% win ratio.

Does that imply that skill = apm?
No.

Does it mean that you get better by having higher apm?
No.

There are three possible explanations when you see a correlation between two things A and B:
1) A causes B.
2) B causes A.
3) Somehting else causes both A and B.

So in this case the alternatives are:
1) High rank causes high apm.
This is aparently not the case. You will not magically get +50 apm once you turn C+.

2) High apm causes high rank.
This is the conclusion that many people, incorrectly, makes from the correlation. It can be true, but it is not neccessarily true.

3) Something else causes both high apm and high icc ranking.
This is what I believe is the main factor. And this "something else" is just training. Training obviously makes you better at sc, that is higher rank, and in 95% at least, training will also make you more active, giving you higher apm.

Also, seeing that the posters on TL tend to be more theretical and strategically cunning than the average sc player, I think that TL players tend to be quite good for their apm, which explains the "I have 50 apm but beat Jaedong 4-1!!" poster that pop up every now and then.
_________________________________________________________

ICCC rank: a better way to measure skill

wtf, what's up with all the confusion on ICC ranking?? + Show Spoiler +
Why are there so many people around going "i am B on ICC, but I only played up to C+ this season. But I am actually B!"? + Show Spoiler +
Are everyone just unbearable braggers? + Show Spoiler +

I will first explain why there is so much confusion with ranking in ICC.

I will then introduce a good definition of what a certain rank means, the CICC (pronounces "kick" if you want...) rank, ending all confusion for ever.

Examples of Confusion
What does someone mean when he says "I am C+ on ICC"? The possible interpretations are many, and each player uses his own way to define his ICC ranking. Some examples that I've seen:

1) He reached C+ last season by joining random "1v1 [his current skill level] MOTW" games.
2) He reached C+ last season by avoiding koreans.
3) He reached C+ last season by bashing D players.
4) He barely reached C+ at the end of the season with a lot of games.
5) He was one of the first C+ last season with a very high win/loss ratio.
6) he believes he could have reached C+ if he would have played more.
7) He is playing close games with other players that are C+, according to whatever definition they may use.
8) He reached C+ the second ICC season.
9) He was C+ on PGT. (lol)

The standard definition, and its drawbacks
If you discuss what the best definition to use is, you frequently get an answer close to 6). That is, the rank that you would get stuck on if you would have continued and played lots and lots of games. You are C+ if you can reach C+, given the time. This seems like a good definition at first sight, as it doesn't depend on how many games you have played the last season. It sometimes comes with an extra specification that the games should be random games against same level, not bashing people 2 ranks below you, or dodging people with too good win ratio or the wrong nationality (racists!!!). Also you sometimes see "current season" definitions, since it seems to get harder every time.

This is not a bad definition, but it has drawbacks.

1) It is not always easy to estimate where you would end up if you continued playing. Because most players don't play nearly enough to reach their maximum. You can be fairly certain that you could get out of your current rank, but how does the the next one look? Just because you have found a few builds that allow you to win 70% on D+, advancing you to C- fairly quickly, that doesnt mean that the builds work at all at C-. You may get stomped the next rank. Still extrapolations of rankings far above what the player has actually reached is not uncommon. We would probably all be happier if we could fix this issue.

2) So this player is hopythetically playing a lot of games, and we guess where he would end up. But should we allow all other players to also play a lot of games, or should we "freeze" the rest of the ladder while he is playing?

a) Let's first assume that we allow everyone to continue playing. An infinitely long season.
A game between two D players give 130 points to the winner, and -50 to the loser. That is, after the game the ladder will have 80 more points in total. So when the decent players have andvanced up the ranking, and there are mostly crappy players left on D, then also they will start gaining points from their 50% win ratio against the other crappy players. Then when also the crappy players are at D+, then also the REALLY clueless players, normally hanging around at D-, will start advancing up the ranking. The same story will repeat itself for D+, since also there will a game result in a net gain of points for the two players. Like this, the ladder will settle with all players around B (CHECK THIS UP!!!) where each game, will cost as much for the loser as it will benefit the winner. So by allowing everyone to play in infinity, we would end up with a definition where the average player is B, and you would basically have to be both blind and deaf to get a D rank, and I'm not even sure you would make it then, since you would win at least 50% against players that are blind, deaf AND has no arms, advancing you to D+... This is not what we want.

b) It works better if you freeze the ranking. That is, you assume noone else are playing any games (or at least not as fast), and allow the player to advance up the ranking as far as he can. The drawback of this is that if he ends up below B ranking, he will have a losing record against the other players at the rank he ends up. To get C- you need to win only (CHECK THIS UP!!!)% of the games on D+, and you would need to win (CHECK THIS UP!!!)% against C- players to stay on C-. This seems like a strange property of the definition, that a true C- will get beaten badly by the average player you actually find at C- on ICC. On the other hand, to maintain an A+ ratio, you need to win (CHECK THIS UP!!!)% of the games against the average player on A+, which also is a bit weird, specially in combination with the C- property.

A new, Better definition
This above is quite confusing, but we still need some kind of common definition if we want to comunicate to others what skill we are at, so I will now suggest a rank definition that is easy to use, but still give an accurate description of skill:

***************************
Definition of ICCC rank:
************************************************************************************************
You are [rank] if you have a 50% chance of winning against the average [rank] player on ICcup.
************************************************************************************************

I name this definition the Cascade ICC rank, or CICC rank, or CICCR (kicker? ).

Good things
1) It makes sence.
It does not try to fight the dynamic nature of ICC, but accepts and embraces it. A C+ player is a player with the skill of players you will normally find at C+ on ICC.

Note that this is not the case in the old standard definition, as C+ on ICC would typically be filled with (for example) B players that just hasnt played enough games to reach their "true level".

2) It is easy to measure.
At least as long as you are below B+. Because then you will reach a level where you win 50% of your games (a lot) earlier than you reach your maximum rank. This will cut down on the number of extrapolations, making it easier to be honest with your rank.

3) You can often look at someones gamelist and see their CICC rank directly from which rank they win 50% against.
If it is a low level player, you can see after not so many games, while the standard definition would need hundreds of games before the player reaches it max.

Bad things
1) You can easily get to an ICC rank above your CICC rank, at least if you are a low CICC rank. Since a CICC: D player goes 50% at D, he will advance to D+ if he plays enough, and could possibly even get higher. Correspondingly, you will have CICC: B players that just don't play enough to get above C-.

This is not a big problem. It just mirrors the fact that at any ICC rank, you will find players of widely differing range, and that player of same skill will put different amount of time/effort into getting high up on ICC. A fact that we can not change.

2) High ranked players will never hit their CICC rank.
For example a CICC: A+ player. he goes 50% against the average A+, so maybe 65% against the average A. But 65% win ratio is not enough to advance from A to A+ (CHECK THIS UP!!!) so he is stuck at an ICC rank below his CICC rank.

I dont see this as a serious issue either. To start with: if you are at that level, you will probably be known and everyone will know how good you are anyways.
Second, the dynamics up at those levels are a bit messed up, as you must maintain a winning ratio against same ranked players to even stay in place. Each game between A+ players REMOVES points from the ladder, so you could have two A+ players at the same skill level play a lot of game, and BOTH would drop down to A- as a result... wtf, so I don't care a lot that there are flaws here, since ICC itself is also messed up.

Conclusions
The best thing with the CICC rank is that it is a lot easier to estimate. Partially because it is well defined, but mainly because it kicks in long before the "I could have reached ..." rank is acchieved. Since it is easier to estimate, it is also more reliable. When someone says that they are CICC: C-, you can trust that you are talking about the same thing a lot better than "I'm C+ on icc". Granted they are not plain out lying ofc, in which case they are douchebags and deserve to die a slow and horrible death. The CICC ranking doesn't do that yet though.

All in all, the CICC rank should be standard when you want to describe your skill, or when you are asking someone else.
"rank?"
"I could've reached B last season if i wanted. ^^"
"lol, sure. :D But srsly, CICCR?"

Outlook
A CICC: D+ player wins 50% of the games against another CICC: D+ player. But what winratio will an CICC: C player have vs an CICC: B- player? This can quite easily be checked by just going through random players game lists.

If I pick a random B- player on ICC, that will (by definition) be an on average CICC: B- player. So I look through the players match list and take statistics on how much he won against for example C ranked players (which will be on average CICC: C players) on the way up the ranking. By doing this for some random B- players I will get an estimate of how CICC: B- does against CICC: C.

Note that the corresponding statistics for the standard definition is a lot harder to do, as you would NOT get B- skilled players (accortding to the standard defintion) on average when you pick random B- players from the ICC player list. Many of them would have a higher "true" ranking that they just havent acchieved.

This will without too much work give an approximate list of expected winning % for any ICCCR vs any other CICCR, and it will be EVEN EASIER to estimate your CICCR by how hard you pwn the n00bs on the way up your ranking.
+ Show Spoiler [mathematical disclaimer] +

btw guys, a few hints on how to do the statistics correctly!
+ Show Spoiler [statistics advice for n00bs] +

An example of how to find your CICC
I am 7-8 on ICcup, and I've played mainly other D players. This makes me CICC: D with fairly high probability. I would have to play a few more games for statistics, but about CICC: D. My standard ICC rank would probably be at least D+ since I think I would be able to maintain a win ratio above 33%, advancing me from D. Would I be able to advance from D+ if I played enough? I don't know. Maybe, maybe not. I am however sure that I am not playing enough to ever know, so it will have to remain one of the mysteries of the universe.
So I am myself an example of someone with a well defined CICC rank, but unknown standard ICC rank.

What is YOUR CICCR?

[POLL on both standard and CICC rank]

THANKS FOR READING!!

hahaha, sorry for that man! Help me find a topic and you'll get rid of it faster!

I agree with everything you say, and that is more or less the point in the draft as well. APM and skill often comes together as you get both from training, not because high APM gives you skill. High skill is (I would guess) also correlated with wrist problems. Does that mean that hurting your wrist makes you a better sc player?

Did you read the draft properly?



If you want a model for how the pins get knocked down by the ball, then that would take some time to implement I think, and you would be better of asking one of hte programming geniuses here on TL. Also it must have been done by someone already...
For just calculating the trajectory of the ball with given speed and rotation (and properties of the floor) is something you probably could do with pen and paper though.

It depends on how the friction is as function of the ball. I would guess that for a sliding ball (which is the case almost always for good players right?) the friction doesnt depend on the speed, or very weakly. I base that guess on how the ball quite suddenly catches the ground when it stops sliding and start rolling. But it would have to be investigated ofc. Linear dependence would probably also be solvable. If it is some other power dependence on the speed, then I probably can't do it analytically without going to mathematica. But I would have to sit down and write down the equations to say for sure, and right now im going to bed. You may very well be correct in saying that it is more complicated than I first thought.


I'm not a strong programer unfortunately. My math/physics background makes it easy for me to understand/invent algorithms, but then actually implementing them, there I suck tbh.


I'm not? I can't?
+ Show Spoiler +