Aethermancer Stats Update

I've been playing a lot of aethermancer the past couple of days with stats enabled. I mean A LOT. Over 150 games with stats enabled, which means I have 50 games with and against each base set card. Here are their stats, in terms of permilles above the fair winrate of 50%.

F -015 -091 +020 +112 -027 +050 +008 -042 -055 +015 +034 +002
W -102 -016 +007 +072 +024 -016 +020 -029 +028 +017 -001 -016
A +032 -025 -034 +022 -064 +003 -010 +020 -060 +042 +041 +003
E +035 -038 +000 -023 -085 -029 -026 +054 +136 +057 -056 -005 

I'm not going to talk about every card, because for example, with F1, we have -1.5% compared to expected wr over a sample size of 100 games - well within the margin of error.

In fact, if you flipped 100 perfectly fair coins, added the heads, and did this for 48 trials, you would get 40 heads on one of the trials, and 60 heads on another.

My first instinct upon seeing these stats is "Wow, 10% above expected winrate. time to make balance changes!" and start writing theories on why certain cards were performing better than others.

But let's take a step back a bit. Assuming the null hypothesis, the expected std. dev. is 5% wr over 100 games. On this dataset, the std dev I got was 4.7% wr.

In other words, I certainly do not have enough data to decide which cards need to be rebalanced and in what direction.

That said, I'm definitely keeping an eye out on the most extreme data points. F4's great performance (compared to F2 and A2, which it forms a trifecta with) could just be noise, but it could also because F4 has a lot of utility in killing early game mana generators.

UPDATE:

I've told AI2.1 to gather some stats for me (with Wild Magic, Brawler, Ornithopter, Magic Rabbit), here's what it got:

F -029 +009 +053 -009 +011 -015 +010 +049 -026 -029 -039 +007 
W +036 -010 -067 -007 -028 +034 +032 +048 -029 +005 -023 +019 
A +025 -000 +081 -064 -027 -043 -021 +037 +010 +009 +014 -021 
E +004 +003 -001 -007 -044 +072 -028 +002 +012 +024 -002 -016 
Games / Basic: 157.33333333333334
Std Dev: 0.03160631364195439
Expected: 0.03986205025895504

Furthermore, the biggest outliers on this data set actually have opposite sign as my own results with these cards. No base set balance changes seem needed at this time.