grateful to the guy for putting in the effort! finally someone put in the math to show what we’ve been saying all along.
obviously doesnt prove any civ is broken, just supports the arguments that civs can be statistically better (wrt the ladder) than they seem and still have a ~54% win rate(eg franks), and civs can be actually worse than they statistically seem and have a ~48% win rate (eg burma)…
im not using it as ammunition to start an argument, i just wanted to share this for anyone thats interested
I dont see anything that is really shocking. You dont just look at win rates, but you look at the full picture, which also include stuff like pick rates, maps, civ match ups, … It is what i always already say about winrate. They are still a good starting point for balance, but you really need to understand the win rates before making assumptions based on them. It is not like just looking at the win rates and you are done. It doenst work like that.
I think this isn’t as Important. There will always be civs played more or less and important is the overall winrate, not a hypothetical if all civs would be played equally.
Interesting would be an Analysis how “Civ Picking” influences the winrates. Most Players have a “favourite” civ they just like to play the most. Some players also just play one single civ all the time. With the Elo system and Matchmaking the vast majority of players will approach a 50 % winning record. So if they only play a single civ, this civ will be pushed towards an overall winning record of 50 %. But if they play let’s say 50 % one civ and 50 % random they will most likely win more games than usual with their favorite civ, as they are used to it but less games with all the other civs as they don’t know them as much. Also the overall elo of a player who plays one civ only will deviate from players with the same skill that play 50/50 or random only, so they are paired with players from different skills. If a Franks only Player of 1600 is matched with a 50/50 Player of the same elo, the 50/50 Player is very likely to win as his real skill is actually higher than the shown elo. On the other hand civs that are played more commonly are also better known by the playerbase and will therefore have less chance to “surprise” the opponent.
It is also important to note, that civs are differently hard to master. Like Saracens and Franks. Franks are a very smooth civ with a clear gameplan, easy to execute. Saracens on the other hand are very special as to use their full potential you need to learn the market abuse. It’s not even worth trying this if you don’t play them on a regular basis. Ofc there are Saracen Players who know how to do it. But the vast majority of players doesn’t even bother using it.
So which winrate is correct? The one of the players who play Saracens “standard” or the one of people who mastered them? Or do we just accept that there is a middleground of these two and thats a valid representative of the “power” of the civ? What happens if for some Reason Saracens get a buff cause they are down the rankings - and this buff in conjection with the market abuse makes them OP? Possibly thousands of players will immediately learn the market abuse and play a lot of games with saracens so the ratio of random and main players would largely shift towards the main players. The winrates of the civ would raise way more than the buff really increased their power, just because more players learned how to play them to their full potential.
This kind of effect can also be already seen with the chinese winrates on different elo. In low elo basically nobody picks chinese and if a player that never learned how to play chinese gets this civ, he has porblems with managing their vill production. Therefore their winrate is very low. But with higher elo there are more and more players willing to master chinese as it actually is one of the strongest overall civs out there if played correctly. Because more player picking the civ which can actually play it to it’s potential the winrates are way higher than at lower elo, besides it is still pulled down by the players who get chinese just by “chance” and never bothered to learn the civ as it is only one of many they occasionally get.
All of these are factors we actually should take into consideration when talking about win stats at least at higher levels of play. But I think it is also quite hard to approach these influences, as it would mean to collect individual player data and try to figure out how much impact all these factors have in winrates. And then we would need to correct the winrate data accordingly, remove all the one civ only players with 50 % winrates. Correct the elo of players which have main civs but play random occasionally also, try to estimate the “experience” factors of the different civs to see which potential civs have, so the risk of overbuffing them is reduced. By showing that eg Sarancens may already have a 52 % (just a random number) winrate potential we could then probably focus our interest how to make saracens more appealing for casual players instead of just buffing them which the raw winrates could suggest.
The problem is, that this kind of influence factor analysis is really hard to achieve - and even if we did it, it would still be questionable if we actually chose the right influence factors to analyze.
This is a beautiful point. As much as I appreciate the pure statistical analysis linked at the beginning of this post, it does not account for sentient growth and adjustments.
Great, another “math” thread about winrates!
So I guess Chinese need buffs?
TL;DR we should be playing more random civ
Great analysis of win rates and how they can be influenced by different playstyles. Very insightful