Back then, slaying the small dragon used to give us only gold. Unlike lolmaths, the site doesn't have that problem.The dragon pit has been a feature in League of Legends since the game’s early days. On hit effects on Azir are nerfed to 50% effectiveness since his soldiers don't proc them.įor attack speed runes effecting, how quickly he attacks, etc. This is a magic constant between 0 & 1.5 (set to 1.0 for Azir), which represents how long you stretch out your fights for. The other Engine Option is Extended Combat, that controls the value of DoT style items like Liandry's Anguish and Demonic Embrace.
Usually you can change this value with Add Engine Option -> Combat Length Override, however it looks like I locked it at 4 seconds for Azir, I'll fix that in the next hour for you so it can be overridden. Basically, how much damage you can do against an target dummy in X time.įor Azir this is currently hard coded at 4 seconds (rather than the usual 1.5 seconds). Hi, so the engine maximizes damage over a set time period. Use the line: Įdit: Forgot to say, that it does understand how Rammus benefits from Armor. That said, if you are looking for hyper armor Rammus, you can click on Add Engine Option -> Override Stat Ratios and increase the number on Effective HP Auto Attacks to 10.0 and set Effective HP Magic to 0.0. So even if items don't syenergy well, if they are the best item, the engine will still select them. Assume all items are balanced) isn't really true. Therefore select items that syenergy well with the champion and don't waste stats, etc. That can be quite different from how a human player does it. So it's a generative form of machine learning (generic algorithm), it reading the stats (and passives and actives, etc.) off the runes, items and champion and puts them together to meet the desired stats (damage is implicit) at the top. They failure of this model is really dependent on the fact you won't ever be able to obtain accurately weighted secondary values.
League of legends builds full#
Unless you have a way of simulating millions of full games of league on summoners rift accurately you really don't have a way of collecting these values with accurate weights. Assuming the balance of these values with arbitrary values really ruins the whole model. However the way every single champion is balanced in league of legends is a mix of raw damage/ effective health, crowd control, and damage uptime, often times with large multi-target multipliers. This weighting problem is really quite significant, because calculating raw damage dealt/damage taken is extremely easy with any machine learning model. Again, I'm suspicious these values are not anchored to anything which is a big problem. To expand on my concern for this issues, you have sliders in which you can adjust the model's weighting of various attributes. These sort of values would appear to have a large impact on item predictions and the fact they are not anchored to anything is concerning. This number seems very arbitrary, and not anchored to anything. For example in your UML diagram on the site you have gap close on proto-belt with a value of 2. You use gap close, resistances, hp recovery, ability haste ect, however I see a significant problem with weighting these variables. I can tell you've put a ridiculous amount of time and effort into this project, and for that thank you very very much. Recent User Guides (Guides by champion is under Builds) People can then view your guides and even edit the builds, say to add Seeker's Armguard, Quicksilver Sash or Grevious Wounds (listed under force items) to the build and the build will will reoptimize itself with those new changes. You can create and publish guides that contain all of the changes you made to the builds alongs with the standard guide text. You can also change most of the engine settings under Engine Options. So if you want a different keystone, want a different mythic or you want to ban an item, you can now do that. The engine will then run in your web browser and change the build to match your new settings. You can now force/ban runes and items and change all of the engine settings under "Engine Options". It will show you the build order that corresponds to those settings. You can select the champion, the build (set of stats to optimize for), and then vary the ratios between those stats. It takes account of runes, champion actives, champion passives, item actives, item passives, burst damage, sustain damage, cooldown reduction, gap closing and most of the behind the scenes formulas in lol. It can find the theoretical optimal build even if no one has ever played it before. It's based on the raw maths of the champions, items and runes. I've got automatically generated builds for all champions in all roles. I've fed League of Legends into my PhD work.