EPT[HoN/DotA] Let's Play~!! - Page 1574
| Forum Index > General Games |
|
semantics
10040 Posts
| ||
|
Sm3agol
United States2055 Posts
On October 15 2011 03:00 Durak wrote: The thing is, you can have a positive KDR and play support properly. The common mentality that supports must have negative KDR is due to poor play and false justification. The foremost reason for a negative KDR is that bad support players often wrack up deaths unnecessarily. They'll ward poorly by going to dangerous positions on their own rather than with team support (or just at the right time). They'll also "sacrifice themselves for the carry" when they actually didn't need to die. They'll have bad positioning and get picked off, either in a team fight or while moving around the map (as they're squishy). Alventenie is good and knows what he's talking about. That's evidence that there is a way to play support without killing stealing and yet have a positive KDR. What you should think is, "This good support player has a positive KDR so there must be something else to playing support properly that I'm missing." Edit: Oh, he posted while I took a long time to type.  Alv gives some good examples.You are most certainly correct with the last statement, I do stand somewhat corrected. It's just that there are only a limited amount of kills to go around ,and simple math means the kills/deaths for each team have to come from somewhere. Assuming equal teams on both sides it is mathematically impossible for the supports to have a positive KDR, regardless of skill, and not gimp the carries in the process. If the carries are getting kills and allowed to farm without dying(aka, you're doing your job properly), and you are also maintaining a positive KDR as a pure support, not to mention your gankers, that isn't amazing support play per se, you're just raping their entire team. I'm basically saying if you're consistently having a positive KDR as a support, you probably should be carrying or at least a dedicated ganker type.... Now keep in mind my experience here, and where I'm coming from. I am fully willing to be proved wrong. My HoN experience consist of my fairly terrible(I'm new to playing...watched Honcasts, some Dota and testie for a while) 1550 MMR play, watching testie, and watching Honcasts of the best of the best at HoN. I have basically no experience with the middle ground, aka "decent HoN play" where you are at right now. TLDR: Shouldn't you be playing heroes that benefit more from having a positive KDR? | ||
|
Alventenie
United States2147 Posts
People underestimate how strong a support with farm is. Once you show people what its like they generally don't mind you taking cs/kills from them. Testie and I when we play know it doesn't matter who gets a kill, as long as you kill them. Very rarely do either of us concede kills to each other because in higher level play anything can happen where people survive with under 50 hp. | ||
|
Sm3agol
United States2055 Posts
On October 15 2011 03:21 Alventenie wrote: I enjoy playing supports. I don't like to play carries as I find them generally boring (barring a few). Most of my top played heroes are either babysitting supports or ganking supports. Few of my heroes are gankers that can transition to semi-carry. People underestimate how strong a support with farm is. Once you show people what its like they generally don't mind you taking cs/kills from them. Testie and I when we play know it doesn't matter who gets a kill, as long as you kill them. Very rarely do either of us concede kills to each other because in higher level play anything can happen where people survive with under 50 hp. I know how strong they can be. I've seen testie carry with all sorts of hero ridiculousness. DS carry is hilarious. But if you play so strongly with a pure support, why not go the semi-carry route, and actually be able to take over a game if you get a few more kills than usual, as opposed to playing a hero that is physically impossible to carry with past level 15, regardless of how many ganking kills/farm you can get? Genuine question here, you're obviously far better than me, it's just not making much sense to me from my point of view. Edit: For example, Magmus seems like a decent support hero that can snowball into a decent semi-carry if you get him some unexpected money. While a hero like DS can be ridiculously fed and still be pretty useless as far as carrying is concerned. | ||
|
Alventenie
United States2147 Posts
| ||
|
rabidch
United States20289 Posts
On October 15 2011 03:15 Sm3agol wrote: You are most certainly correct with the last statement, I do stand somewhat corrected. It's just that there are only a limited amount of kills to go around ,and simple math means the kills/deaths for each team have to come from somewhere. Assuming equal teams on both sides it is mathematically impossible for the supports to have a positive KDR, regardless of skill, and not gimp the carries in the process. If the carries are getting kills and allowed to farm without dying(aka, you're doing your job properly), and you are also maintaining a positive KDR as a pure support, not to mention your gankers, that isn't amazing support play per se, you're just raping their entire team. I'm basically saying if you're consistently having a positive KDR as a support, you probably should be carrying or at least a dedicated ganker type.... Now keep in mind my experience here, and where I'm coming from. I am fully willing to be proved wrong. My HoN experience consist of my fairly terrible(I'm new to playing...watched Honcasts, some Dota and testie for a while) 1550 MMR play, watching testie, and watching Honcasts of the best of the best at HoN. I have basically no experience with the middle ground, aka "decent HoN play" where you are at right now. TLDR: Shouldn't you be playing heroes that benefit more from having a positive KDR? you're talking too much about things you havent done and probably dont do very well against somebody who has much more experience and skill than you. thats called being naive. here theres a huge difference between your gut feeling theory and actual in game experience. i realize that not everybody is aware of this when they discuss the game, but please, spare us your theorycrafting now and actually get better first. you also lack a real mathematical proof, let alone a real mathematical argument, that states that supports should have negative kdr | ||
|
Sm3agol
United States2055 Posts
On October 15 2011 03:44 rabidch wrote: you're talking too much about things you havent done and probably dont do very well against somebody who has much more experience and skill than you. thats called being naive. here theres a huge difference between your gut feeling theory and actual in game experience. i realize that not everybody is aware of this when they discuss the game, but please, spare us your theorycrafting now and actually get better first. you also lack a real mathematical proof, let alone a real mathematical argument, that states that supports should have negative kdr Hypothetical game. Equal skill on both sides. 50 kills for each side. If one side has their carries at a 3:1 k/d ratio, with their supports at 1:2(bad ratio math lol), while the other team has a 1:1 k/d all the way around, I'm pretty sure the team with the stacked carries has the advantage. | ||
|
Glull
Germany404 Posts
i am pretty sure. | ||
|
Sm3agol
United States2055 Posts
On October 15 2011 04:02 Glull wrote: hypothetical game. equal skill on both sides. 50 kills for each side. no leavers. no flames. 3 doombringers. the perfect game of hon. i am pretty sure. It's good to dream. | ||
|
Glull
Germany404 Posts
| ||
|
Sm3agol
United States2055 Posts
On October 15 2011 04:07 Glull wrote: not in a discussion, no. -_- Because your comment added SOOO much. | ||
|
rabidch
United States20289 Posts
On October 15 2011 03:52 Sm3agol wrote: Hypothetical game. Equal skill on both sides. 50 kills for each side. If one side has their carries at a 3:1 k/d ratio, with their supports at 1:3(bad ratio math lol), while the other team has a 1:1 k/d all the way around, I'm pretty sure the team with the stacked carries has the advantage. only covers one (im)possibility of many. why 50 kills and why should they be equal? why 3:1 ratio? you just picked completely arbitrary numbers without any regard to lineups, levels, time in game, map control. if you want to make a sound argument, you cover all cases, and it has to be at least feasible. 50 kills for each side is an uncommon number for the number of kills in a game, especially it being an equal number. ok assuming pigeonholing roles is a good idea (and it really isnt), lets assume team 1 has 3:1 carry kd ratio and 1:3 support kd ratio. there are 4 supports and 1 carry on that team. now lets assume that team 2 is at 1:1 kd ratio all around. there are 4 carries and 1 support on that team. can you say that team 1 has an advantage over team 2 in most cases? ok, lets say you can. now let n be then number of kills (equal in the game). for the case of 0 < n < 10 can you make the same reasoning? how about 10 < n < 20? how about 20< n < 30? how about 50 < n 100? how about 100 < n < 10000? and lets just say what if team 2 was a push strat and team 1 couldnt defend pushes at all, then how does it change? | ||
|
Manit0u
Poland17872 Posts
Sm3agol, please watch some more of the competetive game and take notice of what exactly are carriers doing in those games and how supports act. Also, please stop pestering Alv about playing carry or semi-carry. Even if he gets positive KDR on supports he still might like playing supports more or he's playing with a team that has their roles already assigned. | ||
|
Boblion
France8043 Posts
Poor Sm3agol, all the 1800+ are on you now  | ||
|
Gummy
United States2180 Posts
On October 15 2011 04:17 rabidch wrote: only covers one (im)possibility of many. why 50 kills and why should they be equal? why 3:1 ratio? you just picked completely arbitrary numbers without any regard to lineups, levels, time in game, map control. if you want to make a sound argument, you cover all cases, and it has to be at least feasible. 50 kills for each side is an uncommon number for the number of kills in a game, especially it being an equal number. ok assuming pigeonholing roles is a good idea (and it really isnt), lets assume team 1 has 3:1 carry kd ratio and 1:3 support kd ratio. there are 4 supports and 1 carry on that team. now lets assume that team 2 is at 1:1 kd ratio all around. there are 4 carries and 1 support on that team. can you say that team 1 has an advantage over team 2 in most cases? ok, lets say you can. now let n be then number of kills (equal in the game). for the case of 0 < n < 10 can you make the same reasoning? how about 10 < n < 20? how about 20< n < 30? how about 50 < n 100? how about 100 < n < 10000? and lets just say what if team 2 was a push strat and team 1 couldnt defend pushes at all, then how does it change? His argument is sufficiently specified and totally valid. I don't know why you have a stick up your ass but his argument works for all N. But let's rigorize this in the form of a well-defined allocation problem: if both teams have the same number of kills N (for the sake of keeping consistent domain lets only do N s.t. N mod 5=0), then the average per team is N/5 kills and N/5 deaths. Also we consider only "well-composed" teams of 5 with >= 1 support and >= 1 carry. We define the following partial ordering on player-tagged KDR vectors e: e = <player 1(i): (Ki:kills,Di:deaths), player 2: (kill,death), player3: (kill,death), etc....> with the constraints that the ordinary pairwise sum of player i (kill,death) across all i in [1..5] = (N,N). Thus our partial ordering can be computed based on a linear combination objective function using the indicator variable Ci that takes the value "1" if player i is a carry and 0 otherwise: u*(e) = sum(Ci * (Ki - Di)) across all i in [1..5]. We use Kill minus Deaths as a metric because it is nicer mathematically and you don't have issues of undefined values when D = 0. The following properties relating KMD and KDR are obvious however: KMD > 0 => KDR > 1 or undefined. KMD < 0 => KDR < 1 Finally, we presume that kills and death can be "allocated" across players. This is not an entirely unrealistic assumption, seeing as many support heroes have abilities with which they can sacrifice themselves to save teammates. Nymphora ult-stun combo, engineer ult, rhapsody ult and tank, andro swap, etc.... Obviously left out of this model from the previous assumption is the "ease" with which certain heroes can get kills or escape from being killed at different points in the game. Also skill level of the individual players and communication issues are ignored. For example, if a support gets a kill because he is solo and gets bum-rushed in a tower, that's not a re-allocable kill in a real game, but our model presumes that all kills are reallocable. (Maybe there was a ward there and nymphora tele'd a scout in just in time to get the kill). Now for some observations: Proposition: There exists an optimal allocation of kills and deaths with at least 1 support with KDR <= 1. Proof: + Show Spoiler + By the MVT, there must be at least one character on each team that has >= N/5 deaths. Would you rather that person with N/5 deaths be a carry or a support? Support, obviously, from the computation of u*. On the flipside of the MVT, there must be at least one character on each team that has >= N/5 kills. Would you rather that person be a carry or a support? Carry, obviously. Likewise, at least one character has to have <= N/5 deaths. You would obviously want that to be a carry. One character has to have <= N/5 kills. You want that to be a support. Thus, with locally optimal decisions (since our partial ordering is based on a real-mapped objective function they are also valid with respect to globally optimality), you can easily construct a situation where the same support is the one with >= N/5 deaths and <= N/5 and thus has a KDR of <= 1. Thus we have proven existence. But now that we have this framework, we can prove a stronger statement about optimality: Proposition: If a support in this situation has > 1 KDR (KMD > 0), the kill/death allocation is suboptimal. Proof: + Show Spoiler + This is trivial really. Assume by way of contradiction that the allocation e with a support having > 1 KDR is optimal. Then for all other allocations f, we have u*(e) >= u*(f). But if we take 1 kill from the support and give it to an arbitrary carry, our new allocation f', which is guaranteed to be well-defined since KMD > 0, has value u*(f') = u*(e) + 1 > u*(e), so we have a contradiction. Thus, we have shown that with the given assumptions, no support should have >1 KDR. All this being said, I've played with Sm3agol a few weeks ago, and at least at that time he was a horrible baddie. Afk jungling on keeper while our towers were being 5v4 pushed (instead of counterpushing or teleporting to help) and me having to buy tele stones and sending them to him.... | ||
|
wooozy
3813 Posts
On October 15 2011 02:09 Sm3agol wrote: Then you are probably doing it wrong. Unless you are playing a "support" like plague who has a super powerful damage ulti + amazing damage Q, you SHOULD have a negative kdr. That means you gave the kills to your carry and sacrificed yourself to keep him alive a few times. DAMN SON. 1500 calling alv wrong. kid's got nuts edit: oh wow i didn't even notice this entire page. guess this was late | ||
|
Ack1027
United States7873 Posts
Alv is among the best support players in dota or hon on this website. Please just stop. This is like bronze leaguers rationalizing something fundamentally incorrect as valid to a GSL code S winner. | ||
|
Gummy
United States2180 Posts
On October 15 2011 04:56 wooozy wrote: DAMN SON. 1500 calling alv wrong. kid's got nuts edit: oh wow i didn't even notice this entire page. guess this was late I think Alventenie being a good support and Alventenie having a high KDR aren't causally related. Alventenie is just godly has a good time stomping noob 1800s on a support. | ||
|
semantics
10040 Posts
On October 15 2011 04:58 Ack1027 wrote: Sometimes people need to not talk or come up with their own ideas [ COOL I'M ORIGINAL ] and just listen to players vastly superior than they are. Alv is among the best support players in dota or hon on this website. Please just stop. This is like bronze leaguers rationalizing something fundamentally incorrect as valid to a GSL code S winner. Have you not seen this site, that is exactly what happens all day theorycraftin~ | ||
|
TurpinOS
Canada1223 Posts
Positioning and map awareness go together, and even though they are always important whatever hero you might be playing, it is that more important on supports. (and since youre not farming and youre moving around warding, you have the time to actually pay attention to the minimap). | ||
| ||
