Magic Damage: How I learned to stop griefing and love mage bot

Preamble

The team magic/physical damage ratio. You probably have an opinion on it. Maybe you think you need a 50/50 split. Maybe you think you only need that if the enemy has tanks. Maybe you think you can have 2 AD solo lanes with an AP jungler and AD carry. But tell me, do you KNOW? Do you like really know?

I’m here to answer your questions and tell you you’re probably wrong. About something. A safe prediction really, being right all the time is hard even when you have the data. And oh, I have the data. 25 million diamond plus games since patch 14.6. More every patch. So let’s go see what the DATA has to say about all this.

Perfectly Balanced

First, let’s take on the question: “How bad is having too much or too little magic damage?” and “Do I only need a balance if the enemy team has tanks?”
To answer these, I manually ranked all champions from 0 to 3 on a “tankiness scale’. Here is my listing:

Essentially 0 are champions with nearly no defensive stats, 1 could build some, 2 are fighters who could easily build a few, and 3 are champions who want to build resists naturally.

Next, I defined the phyical/magic ratio for each champion. I ignored both true damage and damage amount for this analysis. True damage is too champ specific and caused anomalies with the win rate output being skewed by Vayne basically. Damage output skewed the analysis because there is in fact an optimal amount of damage to have on YOUR team, but that’s not the question I want to answer right now. So instead, we stick to the physical/magic ratio for each champion. I scraped the physical/magic damage values for each champion from leagueofgraphs. By position, they looks like this:

Note that not all champion names are displayed here due to overlap, but the violin plots do show the distribution of all of the champions, even the ones not shown. Ergo, there are more magic-damage heavy supports so there is a big bulge near .15 Physical/Magic ratio.

Ok! So we defined tankiness and physical/magic output. Remeber, level 3 is what I call tanks, level 2 is what I call fighters.

Let’s unpack.

1. You can be versus 0 tanks and you are STILL griefing if you go all AD, a bit less so AP. Is it steelcaps? Is it GA? Is it Zhonyas? I don’t know, but apparently a little resist goes a long way even on the squishies.
2. If the opposing team has 2+ fighters/tanks, you damn well better be between 30% and 60% physical/magical damage. That drop off is steep. Over 2.5% from 30-40 phys/magic to 20-30 phys/magic. The magic cliff is similar, dropping 2% from 60-70 to 70-80.
3. A slight AD skew is preferable. 40-50 is the clear peak over 50-60.

The Golden Rule of Topside

Back when I was playing with mid-jungle synergies (unpublished as of yet, jungle is soooo complicated), I noticed this very clear trend. AD mids like AP junglers. AP junglers like AD mids. Same with tops to a lesser extent (top lane cares less about synergy everyone else than anyone else). But i was left with a question:
“Should the jungler always have a different damage type than his solo laners?Both laners? Just one?”

So I went to the data. I first divided the top side laners into classes:

I kept mid/adc/sup as 2 classes since most fall into one of the two. Top and Jungle are more flexible so I introduced a mixed class.

Next, I determined the individual win-rate corrected win-rates of different top-side combos: Phys top, Phys jun, Phys mid for example. I also did this for either ADC or APC in the bot lane. Then we plot them as heatmaps:

This is a small plot but boy is it powerful. We have our answers.

1. Do not have the same damage type on all 3 topside roles. It’s the golden rule. If you do you’re staring down a 2%+ dip in win rate. The ONLY exception is if you have a mage bot. Mage bot means triple AD top side is just fine.
2. Wait all these mage bot numbers are like… higher. Interesting observation. Could it be the title of the article is about to become relevant?

Learning to Love Mage Bot

I’m here to tell you that the entire league of legends meta is wrong. Mages should be bot. Yes, I say that partially because bold statements breed reddit karma, but also partially because of this:

Notice the top. Yeah that’s mage bot lanes absolutely rocking your world. And it ain’t new. The split 1, split 2, and split 3 plots all look like this. And it isn’t because APC win rates are high, we subtract all that away. This is just synergy.

So I guess the golden rule of bot side is…

1. Play Mage Carries Bot.

Potential Caveat:

• Mage carries are MUCH rarer bot which means those that play it are likely high mastery. I don’t account for mastery here which could play a role in their higher win rates. Maybe someday Phreak will share his team’s mastery calculation methodology with me :p

Next Steps

Well, I owe you jungle match-ups, I know. Long promised. The thing is jungle is so darn connected to EVERYTHING that it makes finding coherent clusters difficult. It also means I need to consider match-ups vs every other role. There is no “lane”. The rift is their lane.

I’ve got 3 splits of data and I do want to do comparisons. What’s been striking is how little changes in terms of match-ups and synergies. Win rates changes, relationships stay the same. But I won’t be totally confident claiming that until I publish the figures.

After we have classes set it will be model making time. Until then… play mage bots I guess?