1- While this train of thought is right, it's only true when your whole community adapts at lightning speed. If it's possible to abuse, then 1 person discovers it, then the next day everyone does it, then the token value instantly crashes to $0.

In reality, adaptation will take time, and the early malicious users would profit at the start of the crash, while the people who acted good would just lose everything because they acted good. It's a bit similar to the problem of bid bots on steem. If there was no bid bot today, someone would make one because an opportunity exists, and as long as not everyone sells his vote, the token keeps on existing and being worth a little something, and the thieves keep on profitting, even if it's little, they don't care if it's automated.

Overall I think reputation is a complex problem and I think for a website like yours, trying something new wouldn't be a bad thing :D

2- The way you calculate the ReviewScore seems a bit prone to sybil attacks but considering theres only a *2 factor for the VotePoints part (which is the part sensible to multi-accounts spamming), it should be fine.

I'm not sure I understand, taking the average seems a bit silly. Shouldnt the overall score of a project be the sum of the review scores instead? I mean, if a little minnow creates a new review and upvotes himself for 0.001$, it will lower the average and bring the project down in the rankings?

3- I think it depends on what you want. I don't think curation is viable on steem at 25%, so at 20% it won't be. So in this case you should go for 100% authors. You'd get tons of reviews but then there would be no incentive for people to vote on reviews which kinda defeats the purpose. Anyone can go on your site and create a review and earn from it, without even putting a cent, but in order to make money with curation you need to invest, so I think they need to have a better incentivization than 25%. I'm not even sure 50% is enough, I run an experiment on DTube where we redistribute our beneficiary rewards to increase curation rewards, and while it has a bit of effect, I don't think it's enough.

4- In my view, doing an ICO is just a way for you to give a discount on your tokens. If you believe in your token and it's future utility and use, then it's probably better to do like Steem did. Keep your tokens, then sell a % of it every year to fund the development and so on.

I'm glad to see a real website working on SMT implementation, instead of new steem projects popping, it's a breath of fresh air :D