Ok yer, I’m pretty sure my implementation is just wrong, apparently my own manually derived Elo is 1800, as a 1100 player I can only dream 11
2 Likes
I’m going to shoot for the stupid option first, because I don’t understand enough of this to contribute on the smart end. Completely in the first step, when taking that average over 5 games, did the program somehow for some reason divide by 3 rather than 5? Because 1100*5/3~1800.
(Edited because I noticed the original reply had been removed.) ((If that sentence or the quote above confused you, that’s because the original reply is now back.))
1 Like
I think its more stupid than that, I think I’m averaging across both team and solo games for the initial Elo … I should stop drinking whilst coding -.-
2 Likes
wtf… 1.4 %o blood alcohol…
I think that’s the peak where you THINK you can do everything, right before you start to feel something is wrong with you… and release it from your body.
1 Like
Hmm, fixing the initial Elo bug helped a bit but I am still only getting model accuracies of:
- 61% Using the existing Elos
- 56% Using my re-derived Elos across all matches
- 57% Using my re-derived Elos split by map class
I guess will try and have another go in the morning
Rly not surprised to see hindustan at the top. Atelast I was predicting them to be top 3 civ while green said I am delusional 
1 Like
I actually expected exactly this. Hindustanis are totally broken, probably big nerfs incoming.
I hope they preserve the Ghulam though, i like that unit, really cool.
1 Like
ghulam are too fast. Need moves peed nerf.
Then remove the +1/+1 gunpowder bonus completely and move to Portuguese (gunpowder civ with no gunpowder bonus, just UT and UU, also second to last in above civ list as always )
Then maybe we start getting balanced
Btw we were complaining about how poorly designed and bad Indians were in 1v1, now that they are a fun civ to play they need a nerf? cmon
Just another proof how biased people is, keep Mayans, Chinese and Franks untouched but lets nerf Hindustani XD.
Also wtf good way to call Geojak lol a very small data size showing Teutons at 60%WR? is stupid.
4 Likes
First I think they are now too strong in 1v1s BUT the game isn’t only about 1v1s.
Maybe you should read other peoples post history before you make these claims.
Teutons actually have never been a bad civ in 1v1 arabia. Maybe not top tier, but they holded their own, at least in DE. If you look at the error bars it is actually quite ok. Note that only 67 % of the civs will have actual win percentage in the margin of the error bar. So 33 % of the civs will be outside of the bars. That’s how you read this kind of data.
Edit: just read it’s a 95% confidence interval so it’s only 5% of the civs outside the bars. Still, teutons are quite common to have like 51-52 % winrate which is very much inside that error bar.
Portuguese have the best bombard cannons in the game as well as Organ Guns. They dont need buffs for that
They have cheaper gunpowder too
btw just saw the vid of mikeempires with the hindustani HCs…
Doesn’t look really strong to me, especially against most of the ranged units. Sure, it counters infantry, but that’s about it.
The most interesting is the fight against the karambit. I think that’s a nice showcase what the problem of HCs is even against the units they are supposed to counter.
2 Likes
I wont call that countering. Sure, there is no micro, but HC won few times against infantry in that video, and always by little margin.
Would you please ignore all the games before 27th April? I think they are completely irrelevant now.
O yer I will do eventually, in just using them atm because it’s a nice big sample size to test to see if my changes are actually working
1 Like
That’s debatable, Ethiopians, Bohemains, and turks are surely better bbc
1 Like
An this is exactly why those videos are mostly useless when it comes to ranged vs melee fights and sometimes also to the ranged vs ranged fights
2 Likes
Fun fact: Indians occasionally showed up as first (with just as wide of a margin of error) on this graph but people just didn’t gaf cuz that would have gotten in the way of screeching about Franks. So it might mean here that Hindustanis will take back the old Indian’s place once there are more than 2 days of data and people will play them way less often, as there is likely a hype effect going on.
“Oh no civ bonus + UT is too much and UU+UT isn’t enough, surely it will be fairer if it’s UT vs civ bonus + UU + UT”
3 Likes
Normally I agree with this assessment but in the case of HC it is somewhat helpful as the units it counters practically are actually exactly those it also counters in the test. Slow, low HP targets are the ideal opponent for ranged units. And the cavalry does so well in this matchups for the most time, it’s clear it will counter HCs practically in a game, but ofc not by that high margin as seen.
I always try to keep micro potential in mind though it’s hard to approx. But the video showcases quite well that these HC perform quite good vs infantry but still bad against cav and ranged units, despite higher range and armor.