onGamers News

Over valued or Under valued? An analysis of player's performance in NA LCS

So far player stats analyzed have been just raw KDA, GPM, and things like kill participation. This article takes another approach trying to reduce the winning-bias inherent in these stats.

The struggle with League of Legend analytics has been that LoL at its heart is a team game, and it is different than typical sports because of the snowball factor inherent in these types of games. When looking at sports analytics, Lebron or Kobe could score 50 points any given night, win or lose. This is completely different in LoL, as winning teams, by clearing out base towers, securing objectives, and getting more powerful as the game goes on, naturally get more gold and kills. In fact, the winning side averages over 5 times more KDA, and somewhere between 25% and 35% more GPM when winning vs losing. The full breakdown is in this chart:

So, this winning-bias can often invalidate common analysis of raw KDA and GPM. Instead of a player getting a better kda because they are "performing better", instead they are just getting it because they are winning more. That in combination with the fact that we know this game is a team game, so a player can play "well" in a game, but still lose, made me want to create a winning-bias free model to compare players and see where players underperformed or overperformed over the season.

The Methodology:

To properly see which players are overperforming vs underperforming , I took into account a player's win-loss ratio and weighted their kda/gpm/minion kills against a "dummy" player with equal wins and losses. So, for example, if we look at Shiphtur, one of the people seen as out-performing his team. Right now, his KDA ratio is on average 5.56, with a win-loss record of 10-18. So this means that a "standard" mid player with 10 wins and 18 losses would have 10 games with 8.169 KDA (as seen above), and 18 games with 1.787 KDA, leading to a weighted average of: (10*8.169 + 18*1.787)/28= 4.067. When comparing to Shiphtur's actual kda of 5.56, he has an improvement of 30% over his expected KDA of 4.067, which means he is supremely exceeding expectations.

As follows is the breakdown of each stat, KDA, GPM, and Minion Kill per Minute, as following the methodology above. A note, the fluxuation for something like GPM is less because as we saw in the above chart, GPM fluxuates less compared to Minion Kills or KDA, so a 5% above-average rating is very high, compared to KDA where a very-high rating is something like 20-30%.

Thorin will go on here to break down who is overrated vs underrated:

Breakdown by Thorin - Top laners

Nientonsoh's dreadful KDA leaps off the page at you. On a CLG team that is so improved, with all the other four members receiving their share of praise, rightfully, as the season has gone on, Nientonsoh has been highlighted as the weak spot of the line-up. When one considers how drastically below the dummy he is and considers how unlikely names like InnoX and Benny are outperforming him in KDA, it is somewhat telling of his problems at Top lane. It's not as though he's just been off farming like a monster either, as his GPM is fairly underwhelming. Anyone considering him one of the best Top laners in NA is overrating him, as these numbers and our eyes would suggest.

The aforementioned Benny has a very impressive KDA, but his GPM and CSPM are pretty bad. When XDG were at their peak mancloud, Xmithie and BloodWater were all killing it at their roles, I considered Zuna and Benny to be their weak points. Others would mention him as being a good Top laner and had optimism for his future in the team, which this graph would suggest has been shown to be misguided. We see here that despite doing very well in kills, he is struggling in the other areas of the game.

Dyrus' KDA is worth speaking about, as so much of the focus in TSM is Bjergsen and WildTurtle that if this were just a flat KDA graph one could argue his numbers away with the fact he is winning. Since this is weighted and we're looking at how much he is outperforming what would be expected of someone winning so many games, we can see that Dyrus is having a great season all around and is really outperforming what one would expect of him in terms of kill contribution. In a team that wins as much as TSM does, it's hard to be underrated, but I think these numbers suggest that we can rightfully say Dyrus isn't held in high enough regard.

ZionSpartan has been so highly touted thanks to his big games this season, but a look at his overall numbers show that he remains inconsistent, so that he overall is actually doing worse for kills, despite being thought of as a carry in their line-up. This KDA would suggest we should probably reconsider his overall level, not simply look at his great games and ignore the others. Likewise, Quas was so often talked about a potentially one of the best Top laners this season, but his numbers are bad in every respect. This is not a star player, the numbers simply aren't there.

Balls has good numbers everywhere, he is particularly going ham in farm. With that said, he is now frequently listed as the best NA Top laner and often put ahead of Dyrus. In terms of kills, that's simply not the case, according to these graphs. So that aspect of Balls' game is overrated, even if he is having an excellent season all in all. The next best for CSPM is Innox, who has quietly put together a tidy season, from an individual standpoint. The star power of the EU players will always push him out of the spotlight and then the occasional good game from Pobelter sends eyes to the Mid lane of EG, but Innox's numbers tell their own story here and suggests he is likely underrated.

Cruzer is about where you would expect him to be overall, many have pointed to him as a weak spot on Dignitas and the overall trend here says that is on the money. He has outperformed in KDA, but not enough to make him some flawed gem.

Junglers

Meteos is not overrated as the best Jungler in NA and we didn't need these numbers to know that. What the numbers do tell us though, is just how incredible his performance has been. The guy is a complete stud and rightfully should be considered the star of his team, blowing the expectation for every stat out of the water. Speaking domestically, it's almost impossible to overrate him, he's the best and by a long margin in NA. In fact, his KDA almost skews every other jungler because he is so far above-average. Someone who is overrated though, is TheOddOne. People want to say that every player of TSM is having an amazing season and is one of the best at their role, but OddOne's numbers aren't there overall on this graph and we can see that he is underperforming, in fact, he is the only TSM member who placed significantly under expectations for KDA.

Just as Meteos can't domestically be overrated, NintendudeX can't be underrated in any sense of the word. These are appalling bad numbers, especially when one considers that every other player at the position has at least one stat that they outperform the expected in. Since it's no shock that Nintendude's numbers are so bad, I should mention that is a huge shock how bad Xmithie's KDA is here. After S3 Worlds, a legitimate debate could be had about whether the best Jungler in NA was Meteos or Xmithie, with the decision largely coming down to considerations of which style of play one prefers and hypotheticals of who would fit best into other teams. After this split that debate is long since dead in the water, though I'm not sure people have leveled as much criticism toward's Xmithie as his KDA number here would suggest he should have coming his way.

It's always been the way in team esports titles that the last addition to a team is often overvalued if the team drastically steps up a level in performance. These numbers would suggest that is the case with Dexter, who is actually doing very badly in KDA for a player who everyone would cite as having had such a large impact on his team and creating a winning culture there. Consider the praise for Dexter as we move over to looking at Crumbzz' numbers. Here's a player who hasn't seen a legitimate wealth of praise in a year, yet has put together a very tidy resume of stats here, on a troubled team.

It's difficult to call Snoopeh underrated, since his legion of fans seem to inhabit a bubble in which they all pretend it's still the middle of season two and he's still one of the elite at his position in Europe, but his KDA shows there is an element to his play that has gone undervalued by those who, accurately, consider him a player who is no longer near the top of his position.

Mid Laners

The most overrated Mid in NA is Hai, this is something I would have said before and this graph suggests a solid statistical ground from which to say it again. On a team battling to be the best in their region, quite successfully, his KDA is shocking even to a skeptic like me. To be in the negative for all three stats is really eye-opening too. Considering he was lifted into the conversation for best Mid in NA before the season began, with people telling fans not to get ahead of themselves saying Bjergsen would go unchallenged due to Hai's existence, the public must now acknowledge that they have gassed up the C9 man too much.

Shiphtur has been the revelation of the entire LCS NA split and the numbers confirm as much, playing absolutely out of his mind. With his team doing so badly, compared to how well their best matches suggest they could be doing, it's nice to see that the numbers back-up that the buzz around this Mid player isn't just hype based on their wins, he's legitimately performing at an elite level in his region. As long as people want to compare Bjergsen to Faker, then he will be overrated, but in NA these numbers show that he is rightfully entrenched as the king of the Mid lane.

Link is an interesting case, since he gets so much credit from his own team and whenever CLG win, but his KDA is backing up that this is a new and improved Link this split. mancloud has had a nightmare season in terms of where his team began and have finished, but he has persevered and these stats would actually suggest he is probably a little underrated, in as much as his reputation has fallen heavily from the best Mid in NA, post-Worlds, to people not even talking about him in those conversations. In his Grilled interview with me, he mentioned that playing Froggen at an MLG in S2 showed him the importance of focusing on simply farming up and winning through CS, but we can see here that that mentality still lives on in his mind and he continues vacuum up every creep that crosses his path.

scarra and Voyboy are the two star names at Mid who haven't played like stars at all this split and so we can say that their fans overrate them, even if I think the general community has accurately labeled them as hype-only stars. Considering how mad people were at Curse removing Pobelter, claiming it would cost them, his KDA suggests he gets a little overrated after his good games and is struggling to approach carry status on EG. Compared side-by-side with mancloud, we can see that both guys are on terrible teams, in terms of performance, and yet mancloud grinds out decent numbers, while Pobelter's KDA completely drops off a cliff.

AD Carries

Wow at Cop's KDA. For a player who has so often been heavily scrutinised on his teams, with good reason, his KDA is really worth noting and a mental note being made that he is underrated in that element of his game. Consider instead that Doublelift, who is always spoken of highly, is actually a decent margin below where his KDA would be expected to be, yet has still gotten a lot of credit this season in terms of being a carry in CLG. Wizfujin, yellowpete and Zuna don't need too much said about them, I think most would consider them the three worst LCS NA ADCs and the numbers show as much. I will just take a moment to suggest that Zuna's occasionally abrasive persona in fact masks a beautifully calm pacifist, who wouldn't hurt a fly and just wishes to end the division caused by the typical pro mindset that considers those in the server from the other team as "enemies".

WildTurtle quickly gained reputation as an ADC thanks to the success of his tenure in TSM, but his KDA performance here really isn't anything of note. He may still be the best ADC in NA, though the debate with Doublelift is now back and raging at full force, but he's not really performing at any exceptional level that one could look at as underrated in any sense. A man who has gone under the radar is Imaqtpie. He may come off as a goofball who isn't as invested in winning, but after a troubling Dig season he has solid numbers in every single regard here and is firmly placed as Dig's carry player. I think many realise that, but he is likely underrated in terms of never being put near the conversation of who the best ADC in NA is.

Due to playing what is traditionally a carry role, Sneaky has often been thought of as one of the weaker spots on C9, but his KDA tells a different story here. In NA, the man is more than doing his job at ADC, meaning that "WildTurtle could have been the ADC for C9 in LCS" hypothetical gets brought up less often that it otherwise might.

Supports

Obviously I'm going to ignore the CSPM for Supports, as the context is likely too wacky and subjective in each case. XDG Support Sheep made a strong case for being the worst at his position in LCS NA. In a meta where Supports are living the dream, running around and killing opponents in team-fights, he's missing so many kills one has to wonder if Zuna has had a discussion with him about the needlessness of killing in the the modern world. I don't think anyone overrates him, it was mainly just people who wanted to support a new player in the LCS and boost his confidence.

Xpecial is still thought of by some as the best Support in NA, but I don't think the entire community gives him enough credit for his impact on TSM and their wins. The man who engages in seemingly all their fights is posting ludicrous numbers in terms of kills from his position. This the meta in which we truly get to see the power of Xpecial and his mechanics in action. Despite being on a team winning all the time, I think Xpecial has managed to be underrated a little. Speaking of mechanics, a move to NA and a period of refocusing seemed to have put Krepo back on the path to showing himself to be a top Support, but he is much further below the KDA line than I think most would have expected, putting him into the overrated category.

Aphromoo has been one of the reasons Xpecial's name is not mentioned as much, as everyone has been looking at the new name dominating at the position. It's true that Aphromoo has performed above expectations from those who saw his last spell at Support in CLG, one year ago, but I think he is now accurately considered one of the top two at his position in NA now, which the numbers confirm. Lemonnation just does his job in C9 and they win, the graphs tells the same upon inspection. He's right where he should be, and despite all C9's wins I don't think anyone considers him an elite Support, except in terms of his strategical thinking and the more brain-focused side of the game.

With NintendudeX spending most of his LCS games looking at a grey screen, and ZionSpartan forgetting his spinach so often, I don't think many bring up Daydreamin's role in Coast's losses. His KDA really is pretty bad here, obviously not Sheep levels of bad, but bad nonetheless. Nobody particularly every speaks of the Support though, so I don't think anything need be said about him being underrated or overrated.

In conversations I've had with MonteCristo, he always tells me that the Support position is very shallow in NA, in terms of the talent pool. Looking at this graph it's easy to see how accurate that analysis is, as it's basically Xpecial and Aphromoo far out in front, Lemonnation performing a workman-like role and then everyone else being different shades of underwhelming.

Conclusion

To pick out a few names to put as the most overrated and underrated seems a worthwhile task, since we can bring them out of merely the context of their role and teams. I think the most overrated have to be Hai, Nientonsoh and TheOddOne. When you consider that they play for the three best teams in NA, their numbers really are lacking in key areas. The most underrated are Cop, Imaqtpie and Crumbzz. All are on teams a tier below the top three, but all have performed well outside of the context of wins or losses, in terms of their statistics. And finally there have been some confirmed opinions, that Zuna is really underperforming, Meteos is still a stud, and Shiphtur needs a new team.

Data collected with help by the onGamers Stats team: Derek 'Kathix' Adams, Steven 'whedgehead' Falgout, Kent 'Traepoint' Frasure, Jake Morales, and James 'PelkaSupaFresh' Pelkey. Design by Ben 'Sarcasmappreciated' Li.

33 Comments
Posted by Haethos

Great article -- the graphs were a great help to visualize the players against one another.

I would say that it's wrong to call Nien "overrated" though, since I think everyone (including CLG fans) agrees that he is comfortably the worst player on CLG and that he's still really raw after 2 splits as the CLG top laner.

Posted by Spellsy

@haethos yeah I agree that Nien has been historically known as the meh top laner, but it is worth noting how low his stats are. I think one possibly misleading one is dexter's low KDA. I think he is really one of the MVPs of this split, CLG has really turned around since his return.

Staff
Edited by GuichexLoL

Great article, and very useful stats !

But about the support's CSPM sta, is it really useful here ? :P.

Really fun to see the EG stats though, where they are (except Innox) underperforming in some ways but overperforming in others. It seems to reveal how uneasy EG's feeling right now as a team : going well some time then not, even inside of the same game.

Edited by Enzo

Does this not also speak largely to team strategy? Sometimes it's a conscience choice to sacrafice x resources for x players, so that y player can get the lead they need?

It's like how last Season Doublelift was sill consistently hyped, even though he was underperforming from a team-play aspect. Sure his stats were good, but his team was feeding him that farm/kills/etc. I still largely agree with the results, but I think there's more to it than this.

Posted by Nash

Love this articles! :D

Posted by Micket

Hey Thorin, thanks for the article. Whilst the premise of the article is good and has a few interesting statistics, some of the numbers unfortunately fall flat for a number of reasons.

1. It is IMPOSSIBLE for supports and junglers to 'do well' when they are on a losing team. This is the drawback of playing a supportish role. If you are getting stomped as a support or a jungler, it is impossible to have a good KDA unless you deliberately forgo winning the game for your KDA. Imagine being Leona or Xin Zhao - there is NO way to get an assist without committing all in to a fight. If you're behind, you WILL die. There is no equivalent of Ziggs or Gragas in the support or jungler role where you can farm up, pick up stray kills at range and be useful. If you are an initiator, you guarantee your own death, and if you don't go in, you might as well be a minion. Thus, the statistics are heavily biased against the supports and junglers on the struggling teams. Why? Because struggling teams get 'stomped' more often, leading to the situation I mentioned above. If you're slightly behind, your KDA will be ok. The statistics don't account for bottom teams losing games harder than the others. I agree, however, that stats like TheOddOne's KDA probably has some significance.

2. KDA is a terrible statistic for AD carries. There are only 2 jobs that a carry has to do - push towers and deal damage. They don't have much utility, don't have to land good ults (mostly), don't have to initiate, just do damage. ADCs have the most liberal positioning out of all the positions, they are positioned at the back, they have protection, they have escape spells, 2 defensive summoners - basically, if you don't care about dealing damage, you won't ever die. And if your team wins the fight anyway, feel free to pick up that KDA, but unfortunately you didn't actually contribute. The ONLY statistic used that should judge ADCs is the 'Damage dealt to Champions' statistic. As I said before, Carries aren't utility monsters and their performances often go completely unnoticed when they aren't solo pushing or dueling. Remember that game when Shiphtur nearly solo carried as Ziggs vs CLG? The unsung hero of that game was Doublelift - his damage dealt to champions statistic, although far lower that Shiphturs, surpassed everyone else in the game by far. No one watching even noticed it - people very rarely talk about Doublelifts impact on CLG games anymore. If you look at the MVP voting on Reddit, universally Link, Aphro and Dexter get the credit. If possible, I would like to see statistics on 'Damage dealt to champions' for ADCs, with the same adjustments made in the article. That statistic alone, imo, determines who the best AD is. I can guarantee that Dlift, Wturtle and Sneaky will have far more average damage dealt to champions than Cop.

3. There are other things that skew the stats such as champion pools, farm priority, split pushing, team rotations. In top lane, your KDA will differ massively purely on champion pool and playstyle. If you play tanks, and your team loses a lot, your KDA will suffer massively, because you won't kill anoyone and you will still die. But if you play tanks, and your team loses a lot, but you stay in top lane for the entire game without moving, your KDA will barely suffer. If you play assassins mid and your team wins a lot, your KDA will skyrocket. If you play assassins mid and your team loses a lot, your KDA will skyplummet. If you play waveclearers mid or Nidalee, your KDA will always be fine, again depending on how much you feel like you HAVE to go in to not lose.

4. Then there is the case of team playstyle. You criticised Hai for underperforming, due to his poor statistics all round, but fail to realise that this fits C9 perfectly. Hai wins lane a lot, with the 2nd largest amount of solo kills after Bjergsen. Past that, he isn't a fantastic farmer and isn't donated all the blue buffs, which means Meteos takes more of the farm and blues sometimes. In teamfights, Hai is incredibly aggressive. His Lulu and Soraka are very 'in your face' and he is dealing a lot of damage (again, damage dealt to champions statistic would be useful here). This leads to him being focussed and dying. But often, this leads to Sneaky and a superfed Meteos to clean up. This sacrificial playstyle works due to C9's situation of having 4 damage threats.

Anyway, I've said enough. Feel free to reply with your thoughts and opinions.

Posted by Amuny

I really don't like this article, and I think you invest a lot of work for absolutely nothing.

Your whole intro was interesting. But you basically did exactly what you told was bad; Analyze standard stats such as KDA, GPM and CSPM.

In a team game where every sacrifice can lead the game, I think it's not how it works.

In other words, I think the article is opposing himself, and not really worth anything more than the usual statistics.

You have star players, and you have key players. Sometimes, a team couldn't win anything without their key player, while they could without their star players.

I think GGU vs the new Coast is a good example, and I really think Jintae was more key to the team than Shiphtur. Especially at the beginning of the split. Because I must give credit to Shiphtur; he stepped up his game a lot since the beginning.

Posted by banbanxaixai

I second Enzo's point. Instead of calling Hai, Nien and TheOddOne out as underperforming players I'd say that they sacrifice the most in their teams compared to the average.

These graphs do show the lanes and areas of each team they tend to focus on though. The fact that there's such a big variation with team dynamics throughout the league is interesting especially because all the top three teams do have different mentalities about who's gonna carry the game for them.

Posted by Amuny

The dude just over me completely nailed it.

Edited by Spellsy

@amuny I agree that it is a team game and stats like KDA GPM, CSPM can be misleading. However, they are the best way to measure performance currently. Since LoL is a game of many factors, and to add onto that it has things such as snowballing, and patches to invalidate statistics, I think our best current tools are these raw data of kills, deaths, assists, gold, and minions. From there we have to transform them in order to get numbers that hold the least biases to see insight into performance. That's what the goal is here, for more creative style stats (aka other than KDA, GPM, CSPM) should check out this in-depth CLG analysis: http://www.ongamers.com/articles/the-transformation-of-counter-logic-gaming-an-in-depth-look-at-clg-s-statistics-before-and-after-starting-dexter/1100-1205/

Staff
Posted by Spellsy

@enzo I agree that this does speak to a larger strategy too. Someone on twitter brought up the inverse relationship between junglers and mid laners which i think is a very interesting observation. As one jungle has high stats (ex: meteos) the corresponding mid laner has low stats (ex: hai).

One of the primary goals of these kinds of stats is to encourage discussion and interpretation. That's why this features Thorin, so we can see his perspective on these stats, but his is only one opinion, you can take a look and create your own take-aways, which is the great part of stats!

Staff
Posted by jhunsber

Good article with a fresh attempt to remove winning bias from the statistics.

One thing I'd be interested in seeing is the player's statistics broken up into wins and losses. Basically, the same thing as you have but before you apply the weighted average. Does Meteos outperform the field in both wins AND losses? Or is it just that Meteos puts up average numbers during wins but absurdly good numbers (relatively) during losses? Could even give some insights into how some players are super important to their team as they may have above average numbers in wins but below average numbers in losses. Or if you could just post an easy to access database with this information, I'd gladly make those graphs myself.

Posted by PatMcGroyne

Although this is an interesting comparison these stats are more of a comparison teams play styles rather than who is performing well. For instance if you look at benny in top lane he is never with his team, how many team fights have we seen XDG lose because they are fighting 4v5? A lot. It is a general trend that benny is up on the island and doesn't roam when his lane opponent does.

As for Dexter's KDA he is the initiator for the team and he sacrifices himself for objectives. With someone with such a low average KDA and low CSPM his GPM is above average. This is because he is in charge of securing objectives for his team even if he has to die for them. I would bet the CLG

Overall the KDA data is meaningless without a weighted team KDA and weighted objective control data. Some teams play for kills and some teams play for objectives, this leads to variances in KDA and GPM that have not been accounted for by only weighing players against other players of their position.

Just for another example, TSM has a higher KDA than C9 despite having a worse record. This would lead to overrating TSM players relative to C9 players in your comparison which would likely bring Xpecial who you talked so highly of a little bit closer to LemonNation who you said is merely just doing his job.

Posted by alexm1309

Setting aside whether or not KDA is a good statistic to use, some things could be done to make this analysis stronger:

1. Incorporate EU data. Right now C9/TSM make up 41% of all wins. That's going to slant the numbers.

2. Don't use average KDA. You list Shiptur's average KDA as 5.56 while his overall KDA is 4.48. Those are two very different numbers. Overall KDA is a much better approximation of a player's performance. Using average KDA means a 0/1/0 game is the exact same as a 0/10/0 game.

Posted by Spellsy

@alexm1309 I originally used the "overall KDA" measure and it just doesnt work for this idea. When i did it every support was below-expected KDA. The math just skews it.

Staff
Posted by alexm1309

@spellsy Is there a rationale for using average KDA instead of overall KDA other than overall KDA produced unexpected results? If not then you're fitting data to an expected result which isn't a good practice. The problem isn't with overall KDA but with the model.

Edited by The_Sgt_Pepper

These Stats are horrendously Miss leading. Specifically for the Jungle Roll, Since when does Gold Per Minute and KDA have anything to do with snowballing your team as a Jungler. The Impact that Dexter had on the CLG roster is absurdly large and exceeds the stats by a mile. When Meteos farms for 20 minutes and does nothing, He gets high KDA and GPM and does nothing to help his team, These stats are not an accurate representation of how well a Jungler does in game.

Edited by Spellsy

@alexm1309 Yea the thing is this method breaks it down game by game (because you are like "9 wins and 4 losses"), with the nature of KDA (specifically, how much deaths can change your overall KDA when losing). Putting each game at the same weight (which is what averaging does) makes it easier to compare to a weighted average.

Staff
Posted by rob_ee55

Some of these numbers are very misleading and by some, I mean a lot. KDA is something a solo queue player trying to get hyped should focus on. If your a tank or an initiator for your team and you die 4 times but ace the enemy team in the process, your KDA will suffer to an extreme extent. However doing so has a much more positive impact on the game than tanking 2 spells and running so you don't die and causing your team to get flattened. Very situational at best. You can have 1 bad game and you're now the worst player in the world......

For example praising Dyrus over Balls. In every head to head match-up Balls wrecks the lane and gets more 1v1 solo kills than anyone else by far that I can think of in the top lane. No disrespect to him, Dyrus is really good, but he has admitted himself that Balls wins almost every match-up.

Or are we gonna say Doublelift is the 6th best adc in NA? Fun to look at but doesn't paint more than a little bit of the picture.

Posted by Douglas_F

Before anything else I wanted to say that I really appreciate the work you are doing, trying to get more relevant statistics into LoL. I find most the attempt to use things like unweighted KDA to approximate player worth to be pointless.

That said I want to note some issues I have with the conclusion you drew from this, because I don't think from the data you have you can call the players in question over or under rated.

First I think in many cases the data you have is not conclusive. Without the Std Dev (which you probably should have shown) it is next to impossible to tell what deviations from the norm are statistically relevant. So when you say ZionSparaten for example is overrated due to having bellow average KDA, I don't know if that small of a difference is statically relevant, and suspect it is not considering you have a limited sample size so far.

Second as some other people have noted I don't know how much KDA reflects player skill versus team priority. I player who is in the same position, but fulfilling a different roll on the team, could easily have a differing KDA from the norm, even if they are playing well. So say Balls for example could have a low KDA do to him playing more of a supportive role on the team, or due to him being a massive feeder, it is hard to say. Another example would be a player who primarily split pushes, versus one who mainly team fights in the late game, having a huge effect on kill participation.

Third I don't think it is accurate to approximate all wins as beings equal, and same for losses. The problem is I suspect that game length changes the relative KDA for the winning and losing teams. If a game goes longer, the winning team can easily end up padding it's stats, since the fights will be one sided. Alternatively if a game is longer it could be a very even game that only swings at the end due to a large team fight. Either way I doubt that it is a consistent average across game lengths, and since some teams on average have longer games then others this could skew the stats.

Finally I do not think we know which stats have relevance to what makes teams win. To crib a bit of money ball, we don't know is KDA actually drives team success, it could easily be GPM, or time to first kill, or anything else. So I don't know if we can rely on KDA as a good measure of player skill.

None of this is to say that you should stop doing this sort of analysis, or that this was bad. This is to hopefully point out more factors you could work on isolating/separating out in the future, to make the analysis better.

Posted by Spellsy

@douglas_f Thanks for your comment, I think it was very good.

1) it was hard to get a standard deviation for this data because of the weighting. I basically showed a different method of measuring variance from the mean, through comparing them to a "dummy unit". I do think a lot of these values are significant values away from the mean; however, what that means is really left to interpretation (As you mention in your second point, the value of these statistics is on a case-by-case basis)

2) I do agree with KDA being a questionable-stat. Right now though, given what we have I think that the 3 big "success points" in LoL that are easily quantifiable are 1) kills, 2) gold, 3) objectives. Objectives are clearly a team-stat, as they involve much strategy and are rarely secured by a single person's isolated play, so that did not come into factor here. Now, when looking at measuring gold and kills, as mentioned in the intro there are lots of factors here - and the snowballing nature of LoL doesn't make this any easier, so it comes down to trying to isolate a player's stats from the team's stats. And to do this I treated it on a win-loss basis, which I do think in general was effective.

However, as you mentioned there is still biases that occur, such as champions, game time, team performance etc. These things should be in the person's mind as they browse these statistics. I think we are far from crafting the "perfect stat" that can singlehandedly determine a player's performance, if that is even possible. I do think, as you said, this is a step in the right direction though - and much better than analyzing a player's performance based on just raw kda over a few games (cough, riot, cough)!

Staff
Edited by zepidan

Nice stats to look at but I think you should use this article instead to educate those who think KDA/GPM etc is the standard which a player should be measured that they are wrong! One glaring statement here was that Hai is overrated... Hai overrated?? He's UNDERRATED in my opinion. He has intangible qualities that no other NA mids can touch and that's leadership and sacrifice. Most people think Bjergsen, Shiphtur, Link etc are "better" players but what they are not taking into consideration is that Hai is a huge part of the successful C9 machine and it's not because the others are picking up his slack. Hai has a lot to do with their dominance.

Posted by AbuDhur

First of all a very interesting article. I would like to see an overall statistic for the whole team incooperated in the data as well as the discussion. For example it seems like the average CLG member is underperforming in KDA for team with so much wins. Does that mean that CLG is underperforming? No, because a high KDA is not the goal of the game! It rather seems that CLG's focus is not on getting kills, but getting objectives .. Surprisingly, not a single one of them is underperforming in GPM...

Posted by Zerasad

I really don't like this article. The author makes assumptions based on three stats, and tries to say whether someone plays good or bad. You mention how this is a team oriented game, yet you use these stats as measurement. I don't think that junglers and ESPECIALLY supports can be in any way measured by these stats. A support is not good if their KDA is good or their GPM or CSPM is good. You can have a support that has presence in lane and zones out the enemy ADC, making them lose farm, and that doesn't show in these stats. You can have a support that puts wards in the right places, and saves people's lives with clutch plays, but that doesn't show here.

Same thing with a jungler. If someone plays Lee they might get 3-4 kills for their team in the early game, then fall back into an utility role where they go in kick the enemy ADC out and then die, and their KDA will be avarage, eventhough they were the reason their team got rolling in the first place. A supportive jungle that pushes towers with their team and counterganks really well has no way to show here.

Also a team that likes to force 2v1 situations will have a toplaner with lower than avarage farm, gold and KDA, but that doesn't mean that their top is bad. And their bot will have higher than avarage stats, but that doesn't mean that their bot is superior. It's just a different playstyle.

With an example in soccer if a team's strategy revolves around getting the ball to the middle striker, than obviously the middle striker will have more goals then the other strikers, but that doesn't make him better. Same with LoL. Every team has a player that they rely on to carry them and they feed them kills and farm, often to the detriment of other lanes.

Edited by CJPzecond

As a big CLG fan I'm not surprised to see Niens statisitics at all. He prioritizes controlling the lane, pushing waves and proxying waves, so that he will be able to impact early mid game dragon fights more. He has played so much shyvana, CLG has learned how to work with Niens play and really values his playstyle, this essentially reinforced his style. Also, Nien has a penchant for trying to make solo kill plays, like when he gets ganked, or on the mid laner, but has been pretty unsuccesful. So the combination of him never winning lane, giving a few kills really hurts his kda, but he is central in helping CLG win objectives. So in a sense, this statistic does expose his weakness but it hides his strength, which underrates him.

And in response to @banbanxaixai who said that players @spellsy called out, hai, oddone, may just be "sacrificing" gold, cs, or their lives for the team. I agree this is possible, but that is letting them off the case. I think stats are really good for raising eyebrows. But we must now inspect each case to come out with real information. Is Oddone really sacrificing his stats for TSM, or does he underperform at times? We must consider his playstyle and if he is being effective at it. He gets low KDA and low farm in comparison to other winning junglers. This would suggest that he neither succesfully applies alot(above average) of jungle pressure nor farms alot. I personally think, that oddone is very good at making safe decisions, but lacks the playmaking ability to ever have a carrying presence. Hai is another case that is probably more interesting.

Posted by LesVerite

If we look at each of these three statistics, KDA is most representative of team fighting, GPM is a balance between laning and team fighting, while CSPM favors laning.

I think that using KDA can be a problem becomes it can alter depending on which role each team prioritizes. If we look at TSM, Bjergsen definitely deserves the praise for that KDA because he so often secures those solo kills. I would guess that his affinity for out-playing his lane opponent accounts for that better than expected KDA. In the case of WildTurtle, one could argue that the decrease in KDA is more to due with the addition of Bjergsen, and the changing of TSMs priorities, than it is any reduction in WildTurtles skills. His CSPM definitely points to someone with exceptional laning mechanics.

If we apply this same thought process to Sneaky and Dlift, we can guess some different possibilities. Sneaky has a higher than expected KDA, but one could definitely argue that this is because C9 has moved him away from utility ADC to a carry role. His GPM, which would combine CS, objective gold, and Kills is only slightly above expectations. This same logic paints a particularly terrible portrait for Doublelift, considering that his team goes out of its way to feed him. So the fact that his KDA is lower than expected points to a serious deficiency in his team-fight mechanics. Which is actually in line with the general opinion of dlift. Strong lane mechanics, and weak positioning/team fighting. Taking this thought process into account, Imaqtpie is without a doubt the most under-rated of the ADC, as he is above average in all 3 metrics. His statistics are quite impressive, and imo, paint him as better than doublelift.

What can't be stressed enough is how much better than the average players Bjergsen and Meteos really are. Specifically Bjergsen and is GPM being 9% higher than expected. That's by far the highest percentage in that metric in any role. KDA, and CSPM are just numbers that contribute to GPM. So while his KDA and CSPM are both much better than expected, it's rather odd that his GPM is only 4% better.

I also believe the same logic applied to lemon nation could be applied to the OddOne. I don't think anyone believes the Odd One has god like mechanics. But I do believe he brings a decent amount to the team in the way of game knowledge.

Posted by LesVerite

One thing I would like to see here is how each overall team compared to their expected performance. This might make some of the data more definitive, and give answers to why Benny's KDA is so much better, but his GPM so much worse and other such things. If the entire team performance is bad, than it would drag down individual GPM as he would be getting less objective gold.

If we look at the Oddone, his GPM is only 2% better than expected, the lowest on his team. Something that further illustrates that in comparison to his team's performance, he's been subpar.

So a chart comparing performance of player compared to expected, against performance of overall team against expected would be interesting to look at.

Also, lol at the CLG fanboys defending Nien. He only appears useful because the current tanky top Meta makes top laners somewhat useful by default. They're still a big tanky dude walking around, most of them with aoe damage that gives them assists just by existing near a fight. Top lane is a pretty forgiving role right now. People need to remember that an assist is equally valued in the KDA formula, so the fact that his KDA is so bad despite playing Shyvana, a champion with significant AoE who should get lots of Assists, means he really is quite bad. I feel like CLG fanboys consider Nien the Darian of NA. But when Darian dies, his team often gains solid objectives off of them, and those deaths are often somewhat planned with those famous chases around the map. Nien just dies to ganks, or makes god-awful engages. It still surprises me that CLG beat TSM that game. I remember that one Nien teleport right into 5 members of TSM when his entire team was 1000+ units away and backing off. Nien's poor decision-making and mechanics so often cause CLG to lose team fights, or objectives.

Edited by Spellsy

@cjpzecond Yea i lokoed more into Nien's stats, and here is his champ breakdown: http://puu.sh/8adFp.png . The thing that stuck out to me (other than him playing a LOT of shyvana), was that he has 100% win rate on renekton, but less kills than deaths. This is really low given the stats we saw at the begining of this article, where winning top laners average 8.8 KDA, his was only 3.38 after 3 wins! That is very low. A lot of people have been saying that he pulls people top lane which is not reflected in kda, but i have not really seen that in practice :O, i will look closer at it when it comes to playoffs.

Staff
Posted by Spellsy

@lesverite I like your take on the data a lot. One thing that i would note is that KDA really is influced a LOT by how many deaths a person gets, kills and assists are good but are mostly stable.. if you die less though, your kda will skyrocket (ex: cop).

I know when looking into teamwide stats, TSM was well above their expected values, as you can see by looking at each stats down a line you can see they are mostly above-average. i looked through my database to see if i had that data still (i did something like a win-expected for teams, but i think it might have been older :( )..

Staff
Posted by bluesawdust

Hi Spellsy, I don't know if there is a way to email you specifically so I am just going to comment on your most recent article to talk at you. I saw that you were one of the onGamers' staff that was shadow-banned on reddit, which I think is really sad because I enjoyed the expanded discussion about your articles. The reason I noticed you were banned is that I wanted to see if you had created power rankings and predictions for the NA Playoffs like you did superweek, because I was really looking forward to reading it again. Is there any status message on that, or are you no longer working on that bit?

Thanks for the analysis that you do, I think that it is really interesting stuff to read over.

Posted by Spellsy

Hey bitesawdust. Thanks for being such a devoted follower! i really appreciate it a lot, as i have much less fans than the other ongamers guys :p..

Anyway, as to your question, I might be able to pull out a NA predictions article, right now ive been kinda bogged down with other content (and the whole reddit drama), should go check out "Spellsy's Stats Supershow"! .. and working on an EU version of this article. But i might be able to squeeze it out tomorrow, we'll see ;O.

Staff
Edited by EscapeArtist

Statistics are pretty flawed because you're not taking into account that if a good player is on a bad team, he gets better stats and because he's the one doing plays and getting rewarded for it. Also if you put a good player into a good team his stats will fall.

It's same like putting a full gold team play againist a bronze team with 1 gold player. Now if you move the gold player from bronze team into gold team, he's overall avg stats will fall because he's team is also able to make plays.

Also you're not taking into account the champions people play. Hai mostly played sustain mids in the second half of the split so obviously you ain't getting any kills with those (soraka, nerfed lulu etc).