Hi all,
I’ve been working on this for a little while now and shared a few draft versions on reddit but now wanted to share the first full version of the report. I’ll add my own interpretation of the results as a comment a bit later. Please let me know if you have any comments or suggestions !
Links:
Franks look op according to the graphs
You might want to consider adding an axis on the right side of the win-probability charts for ‘effective Elo’. Should help contextualize the results better than raw probability, similar to how you added the 25-Elo point for the parameter graphs.
Also I noticed you said you didn’t account for Elo in the naive win rate section, which might be a problem but at the same time if the match-making system is doing it’s job right the mean elo difference over all opponents a civ faces should be 0, right? But it might need to be verified. I’m not sure what the ‘correct’ way to account for this is but intuitively just converting the mean elo difference to a probability and substracting it is reasonable. IDK this is off the top of my head.
From an interpretation standpoint a question arises involving
2) You need to account for players who play the vast majority of their games as a single civilisation
The ‘dual’ in some sense to this is controlling for when players are ‘experimenting’ with a civ. While you’re right that people playing one civ isn’t a problem because of their rise in Elo, people playing a new (new to them that is) civ is because you’re trying to reflect the ‘equilibrium’ win rates. There’s a bunch of ways to handle it if you can get your hands on already aggregated player data. It would be really cool to estimate an Elo handicap that people can expect when trying out a new-to-them civ.
I would expect this would drastically affect Burgundians and Sicilians, maybe the Chinese and Rajas civs as well. Other than those though probably not much would change. On the whole though it looks like a good parametric analysis. I like it a lot.
They are fine now with the forager bonus reduced, they still have popularity in their side because of the Joan of Arc campaign, the extremely easy Knight rush, cheap castles, etc… but they aren’t OP
At 1.7 TG ELO at least they aren’t that picked as they used to be.
He made it again.
I found this analysis in reddit.
https://gowerc.github.io/aoecps/outputs/report.html
In my opinion, having this kind of exhaustive and honest analysis is good for the game
Very interesting.
I think Malay, Italiens and Portuguese are the most difficult Civs to balance. All 3 are overall good / above average civs as Hera stated in his recent Tier list. The only issue they have is their underwhelming performance on Arabia. But since more then 50% of all 1v1 ranked games are played on Arabia this shows in their winrate 
By now Saracens only have their market shenanigans going for them as a real strong suit. But I know many players that play Saracens in this ELO range and don’t use their market at all. This night explain their low performance.
Cumans are overall lackluster and gimmicky. I guess the Devs will change them with the next patch once again. But whether that will be their TC, Steppe Lancers or something completely different I’m really not sure about…
True, those winrates are specially true for arabia and open maps.
I will suggest him to filter stats by arabia vs arena at least.
Then, they are boresomely overused
well that would be because they are one of the most straightforward civs in the game.
they have an early eco bonus that assists with getting to feudal, an eco bonus in feudal that helps their eco be more efficient, with a straightforward and yet strong game plan militarily all backed up with cheap solid defenses.
I am just wondering: Will the devs also use such analysis to determine buffs and nerfs? This would be a great way to see the civs that need a nerf / buff.
Do you also have some stats based on the maps? I didnt find anything about that in the analysis. The current win rates will be mostly based on Arabia, i think. But some civs will perform better or worse on other maps. You could consider to expand your great work with a map analysis.
The Devs stated that one of their balancing approach is to look at statistical winrates for Arabia and try to get all Civs to a winrate between 45% and 55%.
They consider the average winrates at