DOTA2 LOL comparison project 2013

Participants

Emil Kirkegaard - programming, calculating, theory, writing

Julius Bjerrekær - helpful comments

Introduction

Discussion spawned on Facebook after Julius posted a link to a comparison between DOTA2 and LOL. The full comment is archived here. Some LOL players disagreed with various points, including the difference in spammability between the games. It referred to this part of the comparison:

A majority of spells in LoL are inherently spammable. They cost a low percentage of a character's mana pool, or none at all, and can be used extremely frequently to pressure a lane and test an opponent's ability to "sustain" themselves. Many of these abilities are skillshots which need to be landed as much as possible. The spells are for harass and are predictable, and dying to spells like this as a high level player simply doesn't happen: it's very clear what's coming.

Conversely, a key concept of laning in Dota2 is simply "... go." What I mean by this is that when two allies in a lane decide to make an attempt to kill an enemy, you will know it. Rarely seen spells with high mana costs and large impact come out and decide the fight based on the players' judgment of the situation and execution of their attempt. The availability of such abilities dictates how safely you can lane against opponents. In effect, it comes down to a simple challenge: I think I can stay alive if I stand here. If you don't think I can- prove it. A simple AoE stun on a hero such as Sven or Sand King can cost as much as 75% of the hero's base mana pool, and without the mana to use such a spell these heroes are just big creeps. Other spells can be somewhat spammed, but very few Heroes cast spells in lane half as much as Champions in LoL do: you simply don't have the mana for it. In the early laning phase, this mana is as valuable as health, the latter of which can be quickly restored with consumables. Mana potions are more difficult to use, as they last 30 seconds and are dispelled upon taking damage.

These consumables are noteworthy only because of another aspect unique to Dota2, the courier. Utilization of the courier to bring items to you makes being beaten down in your lane somewhat acceptable- as long as you don't actually die. After healing up, you're still at full mana and because your opponents probably took damage and used mana in the kill attempt, the balance of power in the lane may have swung the other way.

The commenter on Facebook wrote that:

2 - Spamability er jeg 100 % uenig i. Jeg ved ikke hvordan det er i DotA, men selv på mit niveau i LoL taber du hårdt, hvis du ikke bruger dine abilities på det helt rigtige tidspunkt. Det kan godt være at DotA er mindre forgiving for den slags fejl, men at spamme dine spells = low elo.

2 - Spamability I disagree 100%. I don't know how it is in DOTA, men even at my level in LOL you will lose hardly if you don't use your abilities at the precisely right time. It might be the case that DOTA is less forgiving for that kind of error, but to spam your spells = low elo. [my translation from the above Danish]

Instead of engaging in length discussion and having a scientific mind set, I set out to check the actual data instead of basing the judgment on an intuitive comparison, which may or may not be faulty. But even if it is correct, the data the intuition is based on cannot very easily be shared as it is stored in the person's brain in no easily accessible form.

Having recently learned how to program in python, this seemed like a worthy and possible project.

Theory

The notion of spammability deserves some explanation. The idea is that a skill is spammable if it is possible to use it many times in a row for an extended period. There are thus quite a few factors that relate to spammability. There is the cost of the skill. There is the cooldown (CD). There is the resource pool size. There is the resource regeneration (regen) rate. When any of these are missing, it is not possible to use the skill. In other words, if the skill is on cooldown, it cannot be used. If the skill costs more than one's resource pool, it cannot be used. Resource pool size relates also to how many times it can be used without depleting the pool. The regen rate relates to how quickly the resource pool refills.

Full use rate

If one uses the skill immediately after it goes off CD, how fast does one need to regen the resource to continue doing this forever? That is the 100% use resource rate. CD/resource regen. The time one has to regen the necessary resource is precisely the time it is on CD multiplied by the resource regen rate. If this does not equal the cost of the skill, it is not possible to use the skill every time it is not on CD. It is also necessary that resource pool ≥ skill cost, otherwise it cannot be cast at all.

Spammability ratio and average use-reuse time

We can calculate a spammability ratio. In the above case the spammability ratio is ≥1 if the conditions are met. However, this is rarely the case, so it is necessary to have a gradient concept. The spammability ratio is CD of the skill/time necessary to regen the cost of the skill. The longer the time it takes to regen the necessary regen, the less spammable the skill is.

Example. Lina has a mana pool of 351 at level 1. DOTA mana regen = 0.01+INT*0.04. Lina has 27 INT. So, Lina's mana regen is 0.01+27*0.04=1.09 mana/sec. Lina's first skill, Dragon Slave, costs 90 mana, and has a CD of 8.5 sec. The full use rate is 90/8.5=10.6 mana/sec. That is, if we ignore cast times, which I will do in this article. Cast times are just added to the CD if one wants to get the numbers completely right. Lina's mana regen is 1.09. So it is clearly below 10.6. The time she takes to regen the mana is 82.6 sec. So the spammability ratio is 8.5/82.6=0.10. The skill is not very spammable. The data is from here.

Some caution with the interpretation of this ratio. If a given skill has a very low CD or none at all, then unless it is also very cheap in mana cost, the resulting spammability ratio will be low. Thus, it is possible to have a skill that is used very often measured in absolute time that has a low spammability ratio. Spammability ratio is not the same as how often a spell can be cast given full efficiency. That time is the denominator of the spammability ratio. We may call that time the average use-reuse time. It is the time it takes to regen the resource needed to use the skill again.

Average spammability rate

The goal is now clear. We want to calculate the average spammability rate of the two games. If the Decency analysis is right, the spammability ratio is larger in LOL than in DOTA2.

Depletion factor

This is another comparison factor. The depletion factor is the number of times one can use a skill before running out of mana without regenning mana. One could also calculate a depletion factor with mana regen, but it would complicate the calculation.

Relative resource regen time

This is the amount of time it takes to completely regenerate the mana pool. It is calculated like this: resource pool size/resource regen rate. This measures how careful one must manage one's resource if one is to avoid being in a situation without being able to use a skill.

Data

There are various ways to do this. For a quick and dirty analysis, one can pick some heroes at random, and check their spammability ratios. For a better, complete analysis, one can copy data from various sites about the heroes. I chose the latter path. Alternatively, one could watch some games and count the number of times the heroes use skills. Skill use/time would be a reasonable statistic, which would take into account items and other things as well. However, it is very time consuming to do it like that with a representative number of games from each game.

I used three sources for data:

Results and discussion

For simplicity of analysis and programming, we used skills that use mana only. It turned out to be too time-consuming or difficult to include alternative resource skills.

Aspect/Game LOL DOTA2 How much larger is the DOTA2 value
Avg. CD 33.8932714617 35.8267477204 5.70%
avg. mana pool 231.0957446809 248.4742268041 7.52%
avg. mana regen 1.34 0.7745360825 -42.20%
avg. avg. mana cost 69.3191489362 99.0757575758 42.93%
depletion factor 3.3337937385 2.5079215429 -24.77%
spammability factor 0.6551866902 0.2800797037 -57.25%
Cast-recast time 51.7307081613 127.9162582836 147.27%

The difference in CD's is small. However, we note that there are many more ways to reduce cooldowns in LOL than there are in DOTA2. In fact, one can reduce the cooldowns without items at all in LOL, but this cannot be done in DOTA2. See this for LOL and this for DOTA2.

Some care must be made with the above results. The spammability factor above can be misleading. This is because the hero has more than one active skills. The above is the spammability of one fictive, completely average skill on a fictive, completely average hero/champion. But even with one skill it is not possible to use it every time it comes off CD.

However, the data does show that it is very likely that Decency is correct in his analysis regarding spammability. The only way to analysis can be wrong is if there is a huge difference in uncontrolled factors that affect mana regen. This could be either items or increased mana regen based on leveling.

Further analyses could redo the analyses for heroes/champions at maxmimum lvl using the skills at their maximum lvl, but still without items.

Raw data and code

The raw data can be found here.

The code and more raw data can be downloaded here.