Byteball  Because compared to any Blockchain protocol, it is vastly more scalable ON-CHAIN and has easy to use smart contracts.  The inventor and main developer Tony is also brilliant and pragmatic. If you have not read the Byteball Whitepaper then definitely start there; at least the first few paragraphs and last few. Next download the wallet and get an invite to the SLACK channel or join the Telegram group; someone will likely give you a few bytes to experiment with. Then install some bots onto your wallet and see the daily activity happening there. Insurance, betting, gaming, trading via easy to use smart contracts. Have a look at the merchant cash-back programme for distributing Byteball Bytes free (e.g. buy something like dinner in Milan, a hair cut or AirBnB rental in Japan, etc. with fiat and get 10% back as Bytball Bytes). Or buy with Byteball and get 20% back. Blockchain is a huge technical achievement, but it has become evident that it just can't scale to global proportions. Which may be fine if we break it down say to one Blockchain per institution or even industry sector. Hence proposals to simply take high volume transactions off this great invention and onto say a Lightning type extension. But if you think on-chain scalability is still very important then you should investigate systems using a DAG architecture such as Byteball, IOTA and Hashgraph. They have unbounded scalability. Each implements DAG in a slightly different but important way. And each has a different level of brand recognition, due to how each has been marketed at launch. Of the three I would argue that Byteball is technically the most mature in terms of utility. And utility has always been the approach used to create more value in the Byteball network. Actually using it in real life. IOTA is also interesting and I noticed that they are now beginning to pay more attention to showing that it is actually useful too; but that system does not use a deterministic method for consensus (it is probabilistic), I assume this is why they don't yet have an Oracle for transactions to be time-stamped, and no smart contracts yet. Transactions are described as 'free' but remember they still require significant proof of work to be done on other transactions before completing(which is a cost). Hashgraph does a good job of explaining DAG but it is not free to use and has patented various parts.
By the way in my opinion, DASH, BCH, IOTA, ETH are still worth holding but for other reasons mainly around brand recognition and how well they are organised as a project.