The reason why the price is going down is because we don't like the BU threat that will give the power over the blocksize to just 5 mining pool operators (or effectively 1 ASIC manufacturer.) I like bitcoin for it's decentralization properties, not because I trust Jihan Wu so much. I will never accept the idiotic idea called "emergent consensus."
Big blocks are great. I definitely want to see safely planned blocksize increase after Segwit. And if it won't be merged in the beginning in Bitcoin Core, I will just support another client, but emergent consensus is still a joke and Segwit is still the obvious way forward.
Others have said this before, and I think it is worth saying again. The support that BU is getting is not because the ideas behind BU are so great, it is because of the strong opposition to Core and SW.
I don't believe there are any SW features that are *
needed* today/now, but are rather features that might be nice to have in the future. Larger blocks on the other hand is something that Bitcoin has needed for years now. I don't see a reason why SW should get implemented prior to larger blocks, especially after what happened with the HK agreement.
Even if you don't like some Core developers, it should not be a reason to support a broken risky concept "emergent consensus" and to reject Segwit without any proper technical arguments.
Segwit will give us effectively 2.1 MB blocks (if used.) If "Roger Ver" and "Jihan Wu" really cared about "more transactions", they could have Segwit activated in just a few weeks. Any blocksize increase by HF will take much longer than that - especially if you want to have a non-contentious HF (eg no blockchain split.) This should be enough reason to activate Segwit ASAP even if someone hates "Core" so much. We can just focus on a good dynamic blocksize increase with safe thresholds etc after that.
In my opinion second layers like the Lightning Network are very cool too and will be much easier with Segwit too. Buying coffee was never really feasible with bitcoin, not because of the fees, but because of the slow blocks (10 min avg, but frequently 1h.) LN will not just make micro-transactions very cheap, but also allows them to be instant. Of course it would still take years before we can really use this, but if we could at least test it on mainnet already - progress will probably go even quicker. Note that there are already 6 different LN implementations:
Amiko-Pay,
Eclair (ACINQ),
lightningd (Blockstream),
lit (MIT Digital Currency Initiative),
lnd(Lightning Co),
Thunder (Blockchain.info) - some of them work already on testnet. While I still have many skeptical questions about LN, it does seem nice for small (low-value) payments.