I just spent the better of 4 hours trying to figure out this very question. Here is what I found....
Changing the way host ratings are calculated benefits potential guests even more than it would the hosts. Currently, guests are not being shown an accurate measurement of what they can expect due to a flaw in the system of calculating host ratings/percentages.
One of the inaccuracies centers around reviews "host cancelled" and/or if a guest didn't leave a star rating for one of the review categories. Airbnb seems to still count these "missing" stars into overall calculation -- 0 stars are awarded but 5 stars are still possible.
DISCLAIMER: If I am incorrect in the way the host ratings/score is calculated then I apologize but from experience dissecting my own ratings this is the only system of calculation that even came close to matching mine.
1 review = 7 categories x 5 stars = 35 stars possible
EXAMPLE
Guest #1
5 stars in 4/7 categories = 20 stars
4 stars in 2/7 categories = 8 stars
3 stars in 1/7 categories = 3 stars
Total 31/35 stars or 88%, score 4.42.
Guest #2
5 stars in 1/7 categories = 5 stars
4 stars in 2/7 categories = 8 stars
n/a stars in 4/7 categories
Total 13/35 stars or 37%, score 1.86
Guest #3 (Host canceled)
n/a stars in 7/7 categories
Total 0/35 stars. 0%, score 0.00
Guests #4-#7 (4 separate guests/reviews)
5 stars in 7/7 categories = 35 stars each
Total 35/35 stars x 4 (140 stars total), 100%, score 5.0
After these 7 guests the host averages are:
31 + 13 + 0 + (35x4) = 184/245 stars
(88%+37%+0%+(100% x 4)) / 7 = 75%, score 3.75
To illustrate this flaw remove the “missed” categories in Guest #2 review…
Now Guest# 2 totals are: 13/15 stars, 87%, score 4.33.
Increasing host’s overall 7 guest average to 82%, score 4.1.
Going one step farther….
Removing the “host cancelled” review decreases the number of reviews (6 guests, instead of 7) and reduces the total number of stars by 35.
This increases the host’s score/rating to 96%, score 4.84.
That’s a huge difference 96% score 4.84 vs 75% score 3.75!