Important information: It is not about which civ is better. If one civ has approx. 55 % winrate as pocket but 50 % as flank it is more a pocket civ and should not be voted over a civ that has eg 47 % winrate as pocket and 45 % as flank. The poll is about difference, not absolute values.
Most “flexible” civ for TGs
I removed all Civs that I think have a clear role already and are usually never played as the other.
Please correct me right now if you think a civ is missing that can be played flexible.
with Mongols you just spam Mangudai who can serve as both a Cavalier and a Cav Archer basically and you have rly good Siege and can add Hussar if needed. Crucially also good eco bonus and Camel if needed.
Sorry what is the purpose of the poll? To see what people think is the civ that is the most average? Because you specifically eliminated civs that might do better in one field, while still being very good all rounders
For example Mongols are maybe a better pocket than flank, but still possibly better than Incas in both roles
Or cumans being good at both but being much better as a pocket.
What is the purpose of the poll if it’s going to give skew results?
How are you gonna “find” that with a poll? I mean most people voted for magyars who are obviously not equally strong as pocket and flank. If you wanna find such a civ then let people argue for it and pick whatever argument is most convincing.
If it helps the conversation at all I am seeing the following in the data (though note the caveat that there are likely major issues from self selection bias and limited sample sizes which aren’t really captured here). This is based on Elo >1800 prior to the recent Elo adjustment
EDIT : Deleted the graphs as I found a mistake in the derivation code, will repost later once fixed
I voted for Saracens and was so sure that they will be closest. Turns out my assumption is almost right in closed maps while they are 3% better at pocket in open maps (and yet they are picked as flank in every big tournament).
I have no idea how Turks and Italians are equal flank and pocket in open maps. For Mayans I think it is “El Dorado” that is helping them in pocket.
I think you’ve found your result. Bulgarians have perfectly equal w/r in closed map and very very close in open maps. No other civ comes close to them considering both types of map.
I think that a lot of this comes just due to the fact that there are some completely dominating civs that are picked a lot and civs like italians that are rarely seen in general but even more rare as pocket cause… yeah who picks italians pocket? … have a hard time competing with the top dogs in their position and a very low amount of games as the other position => a lot of error.
Edit: Just saw @coolios9876 made extra stats for this… awesome!
I feel pretty confident on the classification method, if you want to take a look my code is here. Will double check tommorrow that the observed slot values in the dataset correctly align to the position values in replays.
I think what I suspect is that due to people self selecting positions that flank civs get rarely picked for pocket and thus have huge confidence interals and are just appearing there by chance. What I might do tomorrow is produce plots of the confidence intervals for pockets / flanks and also maybe increase the point size of the dots in the above plots by the inverse of the SE to give some indication of which points are reliable and which aren’t.
But yer agreed something seems off or at least not aligned with common wisdom. I was originally wondering if its more indicative of early power spikes vs late power spikes where late power spikes perform better in pockets as they have move time to get going, but yer looking at the Civs in question I’m not sure that theory holds much weight either
arrrr crap player colour is a separate variable in the dataset, it is quite likely then that slot is just the position within the lobby not in game. Apologies that is a school boy mistake, will take another look and correct if it is indeed wrong tomorrow. Apologies for the confusion / error.
EDIT: Will pull the graphs for now so I dont give people the wrong impression