Step Accuracy is based on melee Hit rate (shown below) and caps at 95%. The theory behind these tests is that I can tell how much "Step Accuracy" I have using the difference between my melee hit rate and my Step hit rate. Steps cannot be resisted, so I assume a binomial distribution works. This is a less ideal way to do it than checking monster defense and calculating my exact theoretical hit rate, but I was under quite a bit of pressure (from ze girlfreend) to go to bed when I was setting these tests up, so ya get what'cha get!
Format is going to be:
(Steps_Landed/Steps_Attempted) Step_Hit_Rate% +/- 95%_Confidence -- Normal_Hit_rate 95%_confidence :: Type of Step
Case 1: No Step Acc gear
(904/1916) 47.2% +/- 2.2% -- 41.4% +/- 0.9% :: Box Step
(2247/3588 ) 62.6% +/- 1.6% -- 58.6% +/- 0.9% :: Box Step
Looks like there's an innate +8-10 Step Acc on Box Step.
Case 2: Choreia Earring
(547/972) 56.3% +/-3.1% -- 48.5% +/- 1.1% :: Feather Step
Hard to draw too much from this both due to the low sample size and step change, but I'd guess Choreia Earring is between 2 and 6 Step Accuracy.
Case 3: AF1 Hands, AF2 feet, Choreia Earring
(1274/1863) 68.4% +/- 2.1% -- 49.5% +/- 1.0% :: Box Step
(1040/1779) 58.5% +/- 2.3% -- 41.8% +/- 1.4% :: Box Step
With the maximum amount of Step Accuracy available to me, it looks like there's apprixmately a 15-20% Hit rate bonus (30-40 Step Accuracy all told). I'd bet it breaks up:
10 Step Acc innate
5 Step Acc from Choreia Earring
10 Step Acc from Dancer's Bangles (AF1 hands)
10 Step Acc from Etoile Toe Shoes (AF2 feet)
Looks like I need to do a control test with Feather Step to confirm it gets the same hit rate bonus as box step, increase sample size, do testing at a higher starting hit rate to increase my statistical power, etc. but I've put the tentative conclusions up there.