Stop link the stat "winrate"!

Age of Empires IV IV - Discussion
1

people like to link the “winrate” to support their opinion about the balance, i have to say that’s meaningless.

yes, stats doesn’t lie, but it also can not tell you the truth with one or two stats only.

like winrate is 50/50.
its from 2 games? or its 20000 games?
its form low level pools or its high level pools?
its from casual games or its all ranked game?
its from mutiplayer or its all from 1v1?
its from game under 20mins, or its from game over 40mins?
its from games got alot early engagement, or games peaced like 20mins from begining?

there are so many factors, and it really matters.

like the game over 40mins, it pretty much a economic wars. CIVs give a lit bit differences, but mostly players are the key to win the game. better endurance, higher apm, are the main factors to win the game.
the game ends under 20mins, are mostly likely CIV specialty advantages.
this stat is more reliable than the games over 40mins to reflect CIV balance.

you want to use stats to analise the balance, then you need alot detaided stats, and bind it together. not just a brutal “winrate”.

5 Likes
2

So according to your logical understanding if the game ends before the 20 minute mark its most likely due to civ specific advantages but not if the game goes beyond 20 minutes? You know that some civs have their strength in imperial late game while others do not?

People have always posted stats for very specific match ups to find out whether a match up is fair or not so you are really making something up here. Tbh your just making lists with very confusing points.

2 Likes
3

Wasn’t your reasoning for claiming a civ was OP just that you lost to them and your feelings about it?

How is that more compelling than the data we have available?

Don’t get me wrong, when looking at winrates we should also consider games played and the favorable/unfavorable matchups, but we have quite a bit of data on all of that. Data that is much more persuasive than “I feel English is OP”.

4

that under 20 or over 40 , is just a simple example.
the real problem for that scenario is much more complex tbh, just like what i said one stats doesn’t mean anything. we need to consider multiple factors and combine them all.

when we talk about this, we are talking about balance. one stats could explain in its own region. but balance is not something that easy and clear can be explained by one stats.

5

The elo system is always going to place a french main for example at an elo point where they are winning 50% of their games. Up until Conqueror, the stats are worthless. And in Conqueror, there are still the outliers (top 10) that will make civs seem better than they actually are depending on what they feel like playing .

2 Likes
6

Yes. That’s why in the other thread I’ve posted french’s win rate for >=diamond only. It’s the third lowest with 49.3%. Apart from that we will most likely never find out the exact strength and balance between all civs but taking into account stats for very good players should be already a very good measure. And while we do see some numbers indicating that French are at a very good spot right now I am yet waiting to see an argument that kind of shows the opposite. Does anyone know if we have win rates for the current wololo tournament qualifiers?

7

Its pretty well accepted that winrates have about a +/- 3% variance when the game/faction/character/gun is balanced.

Obviously in a perfect world everyone would be at exactly 50%, but that’s not reasonable.

Small buffs/nerfs will then cause the WR to fluctuate (hopefully) within this window, and create variation in the game.

You can agree or disagree with certain takes on civs and their power levels, but currently civs all are within this window in almost all brackets.

8

Win / Loss ratio,
Kill / Death ratio,
Pretty much any “stats” are something that really triggers me.
I love FPS games that hide away the K/D ratio. Especially when the game isn’t about how great your K/D is, but about achieving the objective / victory.

And I agree with the Message of the OP.

Don’t be blinded by Stats. I see far to many people who praise the stats as if it was some sort of holy text.

I am advocating for Ignoring stats, Stats are certainly a usefull analysis tool, but it is just 1 tool in a massive selection of different tools.
There are many, MANY factors that a Simple look on the W/L dosn’t tell you.

Such as W/L relative to the games. Veterancy of the playerbase. And Agency getes very undermined.

You see this very clearly in games such as Warthunder. Where vehicle (unit) performance is solely based on its W/L ratio.
not taking into account, how many games, and what kind of players use them.
One such example is how one such vehicle got nerfed into oblivion, simply because it had an extremely high K/D ratio. But that wasn’t because the vehicle was good, but because the players who played it in a very specific playstyle that prevented it from getting killed at all before the game ended.

Here is where the Stat blinded people failed to see:

The vehicle was so immensly slow, had very good frontal armor, and a OK gun. By the time it finally reached the frontline, you usually manage to fire 1-2 shots before the game ended.
This resulted in the vehicle being rarely killed as it rarely saw action. And whenever it did see action, it usually had the innitiative of a surprice attack and before the enemie could react, they either got killed, and if they didn’t, well the game ends before they get a chanse of counter attacking.

Just looking on K/D didn’t take account of the actual situation on the ground.

Same thing can be said about any stats.

Maybe I’m to old to understand why people have such a Performance obsession.
I gotta have the best W/L ratio. I gotta best K/D score, I gotta make smurf accounts because i got 1 to many losses on it!
I don’t get the why people feel the need to be “better” than everyone else.

in the end, none of that really matters.
My wife dosn’t really care that I manage to achieve perfect rounds without K.O. in the PvP Battlegrounds in Wow with my max lvl rogue lol.

2 Likes
9

We are using those win rates because we do not have any better measure for finding out whether a match up is fair or not. Threads like “civ/strategy is OP” literally pop up every week and in most cases they are just wrong and are based on overestimating someones abilities. I’m quite sure that everyone including me at some point thought that a certain match up was OP after losing to it. The first thing to do though would be to watch the replay or/and youtube videos of pros with a similar match up and find out what one could have done better. Stats also are a good additional indicator. What we see here is people immediately stepping into the forums blaming the game for something they should blame themself for. If people do not bother finding out why exactly they lost but just blame the game they certainly deserve a “get gud”.

3 Likes
10

I do think its worth it to note how the difference between “best” and “worse” civ in terms of win rate is of 3% in all ranks and 4% in diamond+.
I honestly think devs deserve some praise for that considering we have 8 asymetric civs.

4 Likes
11

While there is certainly a Get Gud factor for majority of people who jump on a forum to complain whatever is OP.

There are certain elements that shouldn’t be overlooked, but are overlooked simply because a Civ’s W/L rate covers up from a underlying issue.

Take for example the Balance of Heavy armor units in early age.
There is a underlying problem with them that they in reality, Do not have counters. As their counter unit is locked behind Age III. And majority of Civ’s with an exception of few, Can’t make any Heavy units before Age III.

Another example is the early-hulk problem in navy.
While early Hulks are certainly really strong for the French civ, the Galleass is an absolute useless unit.
It does less damage than a regular Hulk, and it leaves French increadible weak in the Castle-game navy.
Another problem with the early hulk, is that they can actually be countered economically, by outproducing with Arrow-ships, which shouldn’t be to hard to pull off if you play it correctly.
So this leaves, ironically, the french with the weakest navy of all civs, believe it or not. Because their hulks are weaker than a regular hulk. Their Galleass servers no purpose.
Now none of this is adressed, simply because its all hidden behind thier otherwise Good W/L ratio.
And because very few players have figured out how to counter the early hulk pressiure on navy maps.
And Because Navy maps RARELY gets played at all.

The French hulk really only shines in hybrid maps, when they can combine it with knight-raids. Because in order to counter French hulks one needs to fully dedicate oneself to the feudal navy fight and economy to support it. Leaving you incredible vulnerable to land-attacks.

now This isn’t a issue in the long term as Navy will get a complete rework.

But it’s just an example on where things are clearly broken but nobody adresses. Because the Stats dosn’t tell you the situation of individual units performance.

12

As was mentioned above, skill based matchmaking means that players have an approximately 50% win rate, so if a player always plays the same civ, their win rate with that civ will inevitably be around 50%, it would be around 50% no matter which one civ they always play, they’d just be at a different rating for each civ. The win rate would only be useful if it only included wins or losses of players who chose random civ.

13

Either I do not understand you or you do not understand stats.

imagine: old mongols with TR (in winter) VS old China. (only 2 civs in Ladder).
Basically: Mongols beat China.
You play vs fiend with your skill level:
You pick mongols:
WR against mongols = 50%
WR against China =100%
WR avg = 75%
What opponent will do: pick mongols for the win, so now you have: 100% Mongols pick + 50% WR cause it’s mirror.
So either WR =75% and shows that civ is good, or Pickrate =0% and shows that civ is trash.

With large numbers it does not matter, what you pick. It’s large dataset, where are some statistic rules and theorems. Which are proven.

The “random” here is players(not civilisations), players all around the world select civilisations.
Even if you do not understand “how it’s random”, you can randomly choose 10000 different games from Gold league.
and it does not matter what “favorite civ was for certain player”, cause you select games, not players.

14

You don’t seem to have understood me. For a player who plays the same civ every time, their win rate will be around 50% no matter which civ they pick, the civ will affect their rating not their win rate. This is just how skill based matchmaking works.

Imagine civ A is really strong, and civ B is really weak. Imagine you have a player who is truly 1000 ELO. They play the strong civ A and they have a 50% win rate but are 1500 ELO. They play the weak civ B and have a 50% win rate but are 500 ELO. The civ determines their rating not their win rate. It doesn’t matter how many players we look at the data for, the same remains true, it’s their rating that civ strength affects, not their win rate as win rate must always be around 50%.

If only players who select random civ are included, then the civ determines the win rate not their rating.

1 Like
15

I usually pick >=diamond level as I thought players would on average play a wide variety of civs there but after doing some checks this seems to be wrong. Only top 100-150 of the ladder really varies when picking civs. Do you know if we have win rates on the current wololo tournament? Sample size is small but would still make a decent indicator imo.

16

Where are you getting this info? How would a civs “strength” dtermine your ELO? If you have a 50% WR? No matter who you play you’d stagnate at a certain ELO range with that WR.

17

I 100% agree with you on that. However devs are going to be reworking water on season 3 though so that issue and many more other issues are probably going to be addressed then

18

While I understand your sentiment, “play games for fun, not to be better than everyone else”, and it is true in ‘core concept’ kind of way, a large part of the fun in playing a multiplayer competitive game is the competition.

And why play in a competition if you are not trying to be better than your opponent? Sure, your wife or other people may or may not care that you reach into the top 100 players, but it is an achievement you made. A goal you strove for and succeeded in.

No one cares if you make $100,000/year, or if you have the most tone body in the world either. These are goals you set for yourself. I can work at a gas station and be the best WoW arena player ever, I can be CEO of Microsoft and be trash at Fortnite. It’s about you, not other people.

If being the best CoD 1v1 player is your goal, you should go for it, and using metrics is how to know your position relative to others on the road to your goals.

If you are not having fun playing a game, you shouldn’t play it. If you aren’t enjoying the work you do, find a new job. This is the concept that I agree with. But if part of the fun is measuring yourself against others, why hate on someone for it?

1 Like
19

how is it possible to have “not their rating”?! Is it my brothers rating? Or my friends rating?
if I mastered the civ and play it on 1500ELO, then it’s my rating.
If he pick another civ → ELO will “drop”
The probability, that playerA is wining with Mongols-main and loosing on purpose with others. Is the same as.
The probability, that playerB is wining with China-main and loosing on purpose with others. Is the same as.
The probability, that playerC is wining with English-main and loosing on purpose with others. Is the same as.
So it’s even probability to have “villain” doing the same but with another civ.

So I understand your idea. But it’s wrong.
Randomly select 1000 players between 1100 elo and 1200 elo.
Randomly select 5000 games from them, and calculate WR.

You will get “randomness” in the games for given ELO. As you wished.

Or just select all data, cause it’s proven(there are proven formulas + proven statements), that you can do it.
Higher number of examples → better. And outliners can be ignored.

20

If a game developer or balance team is using a singular metric to balance a game, I’d ask if the dev really cares, or knows what they are even doing.

In-game stats are one tool to use, so is community feedback, so is developer’s intent. ALL these things and countless more should be considered by the team working on each game.