League of Legends: I finally updated the player list in the OP!

[kitty]

how is dota 2's ranking system not as broken as lol's

I actually wanted to express a similar sentiment to this as well but I haven't played DotA 2 in over a year so I couldn't say it with any credibility. The regular matchmaking (I did not play ranked) was ridiculously broken when I did play though, even moreso than League's.

Edited February 21, 2014 by kitty
I'm a dumb

[Seven]

Why is the team's victory the only way in which we measure the ability of an individual in the game of League of Legends?

Your arguments are well thought out, but this is where I differ the most in opinion. You have to win more games in order to rank up, so what makes you think that what league you're in isn't some fair indication of your overall skill level?

Let's say two players of unknown skill A and B are placed into Bronze V. After playing 200 games each, A makes it to Silver3 and B makes it to Plat5. The only thing that decided where they landed over the course of those 200 games was how much they won and loss; can you draw any conclusion other than the fact that B was a superior player to A?

I'm not sure your example of Faker or Piglet is appropriate, these players are only on a team because they made or are able to make it to D1 or Challenger SoloQ to begin with. They already worked their way to the top of the system in place; it's not as if they were Silver or Gold players in SoloQ, got placed into a team environment and became Challenger+ level players. If you watch Faker stream and he goes 0-20-0 one game, no one (other than trolls on Twitch chat) would say he's worse than a Plat player who just won 6 games in a row.

Would you agree that the concept of strategy in LoL is reducible to actions taken within the game world? If so, would you agree that all actions taken within the game world are measurable and recordable (and in fact are)? If so, why would you think that measurable actions taken within the game world (which the concept of strategy supervenes upon) rigidly oppose the concept of strategy?

This part goes over my head a bit xd.

I do agree that more in-depth and comprehensive stats to track would be interesting, and help point out what we can work on. In SC2:HotS it tells you your average unspent resources each game (which is the most important statistic until you reach the equivalent of Diamond1 in the game) compared to your averaged unspent resources over time, so you can see if you managed your resources more or less effectively than your overall pace in a given game, among other stats I believe. Due to the fundamental differences between the games these type of stats may not be translatable directly, but it is something to think about.

Off-topic, what is the source of that gif in your sig, relyanCe?

Edited February 21, 2014 by Seven

[zircon]

At the end of the day, I think Riot does have the data to see what statistics correlate most strongly with winning. Look at it this way: if we're watching a replay, and we skip to 1 minute from the end of the game, we can most likely make a very good & accurate prediction as to who is winning / who will win. We're capable of doing that based on things like team gold, towers down, team kills, etc. Riot's team is smart: they can put this type of thing into an algorithm that can probably predict things even better than humans can.

Once they have a formula or algorithm that predicts victory with some degree of certainty based on the numbers at end-of-game, they can compare that to the actual result. So let's say the system thinks a team has an 80% chance of victory based on previous data, and that team wins. Great! Normal LP gains/losses. But if the system thinks it's 50-50, maybe the LP should reflect that it was a really close game and not a clear stomp. Or what if a team had a very low predicted chance of victory but managed to win anyway? That should be reflected too.

[kitty]

Your solution is based on team statistics, which doesn't really help in measuring individual skill.

[zircon]

Yeah, I can live with that. To me, the most frustration is when you play well and someone throws it. i.e. An otherwise close game should result in less LP gained/loss, compared to a game where you dominate (or lose) due to overall strong play. That's just my take on it though. Like I said the other day ingame, I can live with being outplayed. But I hate losing because one idiot gets caught out of position.

[Nabeel Ansari]

The thing is, you guys talking about a team game and trying to measure individual skill.

In a team game, what wins a game isn't any one's person ability to do really well; what gets victories is team synergy, preparedness, vision, and collective gold. Mastering your laning phase and being extremely good at it doesn't really mean you deserve a better ranking in my book.

It doesn't make you a better player of the strategy game, it just means you're mechanically good on the single aspect of a larger game flow.

Good players have good mechanics, sure, but they also have good game sense. They know where to be and when to be there. They know what to do and when to do it. They know what the enemy team will think when presented with certain situations (and can devise plans appropriately to catch the enemy off guard). They know which of the players on their own and the enemy team make the most mistakes. These are all of the things that define a good player and the only things I think should make you deserve a better rank; however, the fact of the matter is that these things are impossible to measure with algorithms.

It is impossible to measure "game flow" (not to mention how much a player is positively affecting the game flow) without the game's code itself have a strict set of meta constraints and also be able to do on the fly theorycraft to see who is actually in a better position. "We have more kills/gold as a team" is not a reliable indicator of who's actually winning. If you're "winning" by all normal methods of thought, and then your team throws, it'll drive an algorithm insane. You theoretically should have won, but you didn't. Should you get an increase of rank if you lost a game? Could you then theoretically rise through the ranks without actually winning a game?

Every constraint of judgment (KD, gold, etc.) for the proposed algorithm that people are proposing in this thread all have downsides and contradictions. That's why W/L is the only thing judging people, because by normalization, you will rise through the ranks. A good player rotating through a metric ton of different team players will normalize out a win ratio that will reflect his skill level. It's just simple statistics, and that's why that's how it is. Trying to add more things makes the algorithm very flawed and inaccurate.

Edited February 21, 2014 by Neblix

[zircon]

Should you get an increase of rank if you lost a game? Could you then theoretically rise through the ranks without actually winning a game?

No, but it should soften the loss / gain of LP.

"We have more kills/gold as a team" is not a reliable indicator of who's actually winning.

Yes it is. Or more accurately, it's a reliable indicator (when combined with other stats) of how close the game is. Teams that put up a good fight should be rewarded more than teams that don't, and crushing the enemy team should be worth more than winning because one person threw. This stuff is absolutely measurable. In all my time of playing league, I think I can count on one hand the # of times that we had a "close game" with a huge gold difference, or a stomp when the gold was very close. Throw that in with the other available stats and you could tweak the LP algorithm nicely.

[ArmadonRK]

Let's say two players of unknown skill A and B are placed into Bronze V. After playing 200 games each, A makes it to Silver3 and B makes it to Plat5. The only thing that decided where they landed over the course of those 200 games was how much they won and loss; can you draw any conclusion other than the fact that B was a superior player to A?

You proposed:

1. B was superior to A

I propose:

2. B was matched with superior teammates to A

2a. B was matched with fewer trolls/feeders/AFKs/etc.

3. B was matched with inferior opponents to A

3b. B was matched against more trolls/feeders/AFKs/etc.

So, to answer your question: Yes, I can.

Yes it is. Or more accurately, it's a reliable indicator (when combined with other stats) of how close the game is.

Typically, yes, a team that leads in kills and/or gold leads the game overall. But there are so many effective strategies contrary to this. I have seen a good percentage of games where the team behind in kills and gold was ahead in turrets, and won the game off of superior pushing. I have been on that team many times. I have been against that team many times.

Not to mention, late game aces can lead to tremendous comebacks. If you just stall a team until the 50 minute mark, even if it's 6 to 30, 30k to 60k, those 60+ second death timers mean one unfortunate teamfight can turn a game around.

My point being, I don't see a way to reliably measure individual skill in this game. I don't see a way to reliably correlate in-game statistics with the outcome. You can single out such stats for "most of the time", but these are nowhere near "all the time". I'd guess (based primarily on anecdotal experience) that around 10% of games lie outside the normal conditions of how a team won a game, and that's too large a margin to use as a baseline, in my opinion.

Edited February 21, 2014 by ArmadonRK

[Nabeel Ansari]

Yes it is. Or more accurately, it's a reliable indicator (when combined with other stats) of how close the game is. Teams that put up a good fight should be rewarded more than teams that don't, and crushing the enemy team should be worth more than winning because one person threw. This stuff is absolutely measurable. In all my time of playing league, I think I can count on one hand the # of times that we had a "close game" with a huge gold difference, or a stomp when the gold was very close. Throw that in with the other available stats and you could tweak the LP algorithm nicely.

If you're talking about softening LP loss, then that's perfectly fine and I agree.

I was responding to other suggestions in the thread.

I don't see a way to reliably correlate in-game statistics with the outcome. You can single out such stats for "most of the time", but these are nowhere near "all the time". I'd guess (based primarily on anecdotal experience) that around 10% of games lie outside the normal conditions of how a team won a game, and that's too large a margin to use as a baseline, in my opinion.

Bingo. You can say teams with more gold/kills win games, but too many times this is just not true and the amount of games where this is not true has too much significance to design an algorithm around.

Also, team gold is a flawed concept of itself. For an accurate algorithm, you need to check if your gold is in the right places. Who has the gold? Is it skewed?

Same for kills. Is it all on one guy, or is it evenly distributed. According to a necessarily enforced meta (it would be impossible to make these decisions without one), are the gold/kills distributed where they should be?

If it's 14-15, doesn't mean much if the support has like 8 of these kills.

If gold is 13k-11k, doesn't mean much if your top laner was just sitting top for 15 minutes free farming while your team got pummeled.

There are too many cases of games that disagree with "kills/gold wins games". Game flow is a much more reliable predictor of winning a game, and thus positively affecting the game flow is what makes you a good player (because you are most definitely increasing the chance of you winning the game). As I explained before, game flow and player effect on game flow is impossible to measure.

Edited February 21, 2014 by Neblix

[Seven]

You proposed:

1. B was superior to A

I propose:

2. B was matched with superior teammates to A

2a. B was matched with fewer trolls/feeders/AFKs/etc.

3. B was matched with inferior opponents to A

3b. B was matched against more trolls/feeders/AFKs/etc.

So, to answer your question: Yes, I can.

I should have specified that they both placed into BronzeV, starting out with the same MMR and therefore being placed with comparable teammates. And maybe change the number of games to 2000 so that we can eliminate the chance that player A got feeders/trolls on his team 200 times in a row.

As far as trolls/feeders/AFKs go this is only an excuse over a long enough period of time. Granted, you may have a stretch of several games where many games are ruined by other players, this happens to everyone and is not avoidable. One could argue that, assuming you indeed are NOT a troll/feeder/AFK then there is a higher chance of there being a troll/feeder/AFK on the enemy team because there are 4 potential trolls on your team and 5 on the enemy team.

[The Derrit]

I should have specified that they both placed into BronzeV, starting out with the same MMR and therefore being placed with comparable teammates. And maybe change the number of games to 2000 so that we can eliminate the chance that player A got feeders/trolls on his team 200 times in a row.

As far as trolls/feeders/AFKs go this is only an excuse over a long enough period of time. Granted, you may have a stretch of several games where many games are ruined by other players, this happens to everyone and is not avoidable. One could argue that, assuming you indeed are NOT a troll/feeder/AFK then there is a higher chance of there being a troll/feeder/AFK on the enemy team because there are 4 potential trolls on your team and 5 on the enemy team.

'it happens to everyone thus is fair' is a logical fallacy.

thinking anyone will play 2000 games of league and not be incredibly unhealthy, dead or a professional is also kind of stupid. most people play less than 100 ranked games/season, making your mathematical theorycrafting a moot point.

you cannot argue that because you're not a troll your team is more likely to win, because the likelihood that someone in your game is a troll or not is not a team-based statistic, it is an individual statistic. there is no percentage of the time that any individual will troll or give up, unless that percentage is 100% in which case that person will get banned eventually. this is actually one of the few statistics that CANNOT be used as feedback, out of all the statistics of league.

also the idea of 'comparable teammates' in bronze 5 is nonexistent, as any two people with mmr under a certain range (my guess is 1100-1200, but the relevant point being a large percentage of the people playing this game) aren't actually going to be the same skill. there are hundreds of thousands of people floating around 900 MMR. between those hundreds of thousands are people of vastly different skill levels. ask anyone who's been at very low MMR: there are people who lose games for everyone else and shouldn't even be that high, people who belong there, and people who are clearly better than everyone else, but not good enough to consistently overcome the aforementioned teammate who loses games for everyone else.

MMR works - but only into mid-level gold and higher. seeing as that's less than 10% of the total population of users, the MMR system is wildly unreliable for the vast majority of players.

'get better' is not the only solution here.

Neblix: the algorithm is already flawed and inaccurate. what is worth protecting here?

Edited February 21, 2014 by The Derrit

[ArmadonRK]

I should have specified that they both placed into BronzeV, starting out with the same MMR and therefore being placed with comparable teammates. And maybe change the number of games to 2000 so that we can eliminate the chance that player A got feeders/trolls on his team 200 times in a row.

There are currently approximately 62,000 players in Bronze V on NA. But you can be matched with players from as high as Gold in a typical match. Let's keep it simple, though, and restrict ourselves to the 315,000 players in Bronze, and possibly the 295,000 in Silver.

At maximum diversity, you can be matched with/against up to 800 of those players.

So, the chances that both players are matched with "comparable" allies and opponents is actually pretty slim. 300 to 600 thousand players offers a lot of diversity.

As far as trolls/feeders/AFKs go this is only an excuse over a long enough period of time. Granted, you may have a stretch of several games where many games are ruined by other players, this happens to everyone and is not avoidable.

Actually, you have this backwards. The fewer games they play, the more their rankings are skewed by luck-of-the-draw.

After 500, 1000, 2000 matches at the same MMR the differences will normalize and become statistically insignificant. In 200 games, their matchups are more likely to be wildly different.

One could argue that, assuming you indeed are NOT a troll/feeder/AFK then there is a higher chance of there being a troll/feeder/AFK on the enemy team because there are 4 potential trolls on your team and 5 on the enemy team.

Since I'm not a stats person, I can't explain why this is wrong. But I believe this is wrong. Going to look it up now.

Edited February 21, 2014 by ArmadonRK

[zircon]

Thinking along another line, I really like the idea (suggested elsewhere) of having an "MVP" for each team in each game. The gist is that after each game, you pick someone on your team (but NOT yourself) who you think was best or contributed the most. If you were required to do it, or got a reward for doing it optionally, then it could tie into the honor system or have cosmetic rewards. This could also be interesting to promote players who are consistently doing well.

[ArmadonRK]

Thinking along another line, I really like the idea (suggested elsewhere) of having an "MVP" for each team in each game. The gist is that after each game, you pick someone on your team (but NOT yourself) who you think was best or contributed the most. If you were required to do it, or got a reward for doing it optionally, then it could tie into the honor system or have cosmetic rewards. This could also be interesting to promote players who are consistently doing well.

This seems like an interesting idea. Dominion does have a stats-based MVP marker. It tracks points for kills, assists and point capture (and, I believe, kills and assists in turret proximity as point defense, though don't quote me on that) and designates an MVP at the end of the round based on those points. It's a rewarding TF2-esque system I wouldn't mind seeing outside of Dominion, as a purely cosmetic feature.

On the other hand, a vote based system for MVP would suffer from some of the same failings as the current honor system, which was so heavily "nerfed". Honor ribbons are so rare nowadays I sometimes wonder if anyone other than myself is giving out honor anymore.

[relyanCe]

The thing is, you guys talking about a team game and trying to measure individual skill.

that is what SoloQ does. It uses a team game to measure individual skill and enforces the reward system of a team's performance upon an individual player, which a source of major frustration among the community of ranked solo players.

Do you see the dissonance now?

In a team game, what wins a game isn't any one's person ability to do really well; what gets victories is team synergy, preparedness, vision, and collective gold. Mastering your laning phase and being extremely good at it doesn't really mean you deserve a better ranking in my book.

This is what the current SoloQ system rewards. Why are we in disagreement?

Do you seriously not see the dissonance?

Good players have good mechanics, sure, but they also have good game sense. They know where to be and when to be there. They know what to do and when to do it. They know what the enemy team will think when presented with certain situations (and can devise plans appropriately to catch the enemy off guard). They know which of the players on their own and the enemy team make the most mistakes. These are all of the things that define a good player and the only things I think should make you deserve a better rank; however, the fact of the matter is that these things are impossible to measure with algorithms.

You saying it is impossible doesn't make it so, unless of course you have 15-20 years of programming and statistics experience to back it up. Otherwise, if you're going to constantly repeat "It would be hard so we shouldn't do it", then you really have nothing more to add to the discussion except variations of that theme. If you'd care to explain in an at least semi-rigorous way why such would be impossible, then please, do.

It is impossible to measure "game flow" (not to mention how much a player is positively affecting the game flow) without the game's code itself have a strict set of meta constraints and also be able to do on the fly theorycraft to see who is actually in a better position. "We have more kills/gold as a team" is not a reliable indicator of who's actually winning. If you're "winning" by all normal methods of thought, and then your team throws, it'll drive an algorithm insane. You theoretically should have won, but you didn't. Should you get an increase of rank if you lost a game? Could you then theoretically rise through the ranks without actually winning a game?

No, because the MMR system (rightly) enforces a 50/50 win loss ratio.

Further, this "game flow" you speak of is not necessary for a system to take into account to accurately measure the basic actions necessary to take in order positively affect "game flow", and your assertions that "game flow" is impossible to measure is because you are purposely thinking too big.

Every constraint of judgment (KD, gold, etc.) for the proposed algorithm that people are proposing in this thread all have downsides and contradictions.

Please list some of these downsides and contradictions, compare them to the downsides and contradictions that have been proposed against win/loss, then argue your point. Otherwise, you're just stomping your foot.

That's why W/L is the only thing judging people, because by normalization, you will rise through the ranks. A good player rotating through a metric ton of different team players will normalize out a win ratio that will reflect his skill level. It's just simple statistics, and that's why that's how it is. Trying to add more things makes the algorithm very flawed and inaccurate.

W/L is easy to measure and aggregate client-side, and is for the most part accurate. Nobody is arguing this. What is being argued is that the system in place is hugely responsible for how plain unfun and toxic soloQ is. People are penalized for good play and rewarded for bad play at least 50% of the time, and this is extremely dissonant with trying to measure and reward individual play (this is what SoloQ is).

Seven, the conversation moved a bit since I last checked, so I'll reply to you in pms later.

[Brushfire]

Here is a mock up of the ingame overlay. Kinda minimalist, and I think that is fine. Something I thought about doing was taking actual player pictures and putting them over the characters protraits when I do actual highlight reels.

Thoughts? Also does OCR Heroes even have a team logo?

Edit: A real annoying thing about LoL's programming is that the game is a separate instance from the lobby client. So I have to reset the stream every a new game starts.

Edited February 21, 2014 by Brushfire

[SkyRiderX]

@ Brushfire : No we don't atm. I have a few ideas and I was going to get one made for us soon.

ATTENTION ALL OCR HEROERS - NEWS ABOUT TOURNAMENT!!

IT'S BEEN CANCELED!

So I finally got word back from the guy in charge of running the tournament. Apparently there was a major issue with a local gaming community that had a bit of a falling out...and they apparently had 4 teams that were supposed to compete and now aren't....so that brings the number of teams competing too low for the tournament to run...

I'm sorry for getting your hopes up guys..if you guys want I can try to find another tournament down the line for us to join in?

[relyanCe]

I'm personally down with just continuing to play as OCR Heroes.

I think we have a great team going right now. We could probably do pretty well this season.

if another tourney comes up, I'd be down to play it.

EDIT: I can work on a logo for us. If anyone has some ideas as to what they'd like to see in our logo, shoot me a pm

Edited February 21, 2014 by relyanCe

[The Derrit]

does anyone want to play today

[zircon]

Servers just pooped themselves but I'm on throughout the evening.

[kitty]

I nominate OA for making a logo.

[Brushfire]

I nominate OA for making a logo.

I would also like to nominate the Bald Man from Minnesota Land!

[OceansAndrew]

reylance and I can collab on something - dude hit me up, and lets talk

[zircon]

Random late night theorycrafting: wouldn't it be cool if Runaan's Hurricane had an active ability? I was thinking something that would push enemy champions within ~150 units (or so) away, by maybe 200-300 units. The issue with Runaan's, among other things, is that you have to be close to use it. AD carries generally don't want to get in close, because they don't want to get jumped on. An active ability like this could be used to instantly get people off your back and let you get in 1-2 more shots (or run away). But the item would still be sub-optimal for damage, and a situational pick.

[kitty]

I don't like the idea of giving a champ free AoE CC that's not affected by tenacity but I also don't like an alternative of giving the champ a free dash since there are plenty of champs with escapes that would potentially buy Hurricane (e.g. Lucian, Cait). But you said sub-optimal for damage so it could work.