Thursday, October 30, 2014

Predicting Difficulty, ROI, and Price of BTC

If you read that title and your first thought was, "Anyone that can do those three things accurately is guaranteed to make a ton of money!" Well, you're absolutely right... and your next thought was probably something along the lines of, "Too bad most of that stuff boils down to pseudo-science and lame charts that are wrong 90% of the time," again, you're right.

One of my recent posts basically had a couple of people call me out and say something to the effect of, "But Jarred, you're not even factoring in the increasing difficulty of BTC mining for your calculations, so they're all wrong!" To which I responded, effectively, "Yes, and I didn't factor in price either."

Of course course the difficulty of mining BTC -- or any other cryptocurrency -- changes the ROI prospects. The price of the target cryptocurrency likewise changes ROI prospects. Factoring in both of these is basically... well, at best it's just pure guesswork, and so I've stopped bothering. At this point, we should all know that difficulty and pricing for cryptocurrencies change regularly, but there's no telling what they'll actually do.

I've said before that difficulty follows price, but that's only part of the equation. Difficulty follows price as well as hardware efficiency, with other less significant factors also playing a role. BTC has had steadily increasing difficulties every cycle for so long that everyone seems to have forgotten that at one point, the difficulty actually plummeted! It did so right after the first big crash from $30 to $2. At some point it will almost certainly happen again, so if you're really smart you'll sell all of your Bitcoin right before the next big crash and then buy back in at the bottom so you can make tons of money. (Good luck doing that, by the way.)

So again: difficulty basically follows price as well as the efficiency of the hardware, and unless you can predict both accurately, any guess at difficulty is just that: a guess. Let me give you some great examples.

If we look at April to October 2013; the average increase in difficulty for BTC was 22% per cycle (a bit less than every 12 days). Then if we look at the next six months, October 2013 to April 2014, the increase per cycle averaged 25% (with a few instances where it jumped more than 40%). Wow, difficulty never stops increasing, right? But then we go the the past six months (April 2014 to October 2014) and the average increase is now only 12.5%. More importantly perhaps, look at just the last month where difficulty only increased 7% per cycle -- or the last two increases of 1% and 3%!

The reason the difficulty increases have fallen is that the price of Bitcoin has trended down lately (about the lowest it's been since a year ago), coupled with the fact that there's not really any more efficient SHA256 ASIC hardware coming online right now. As I've pointed out previously, the best SHA256 ASICs are doing something like 0.5W per GH, and in the not too distant future (~3 months?) we might see 0.25W per GH... but the next bump after could be much further out and eventually we might only get 0.1W per GH before we basically are stuck following Moore's Curves. That means the efficiency will be very difficult to improve without a process technology shrink, those only come every two years or so now, and in many cases a process shrink only improves efficiency by 30% (or maybe 50%).

We might see some other "tricks" where companies focus on making truly efficient ASIC designs and get a boost without a process shrink (e.g. look at NVIDIA's recent GM204 launch, where they increased performance and reduced power use while staying on the same 28nm process technology), but given SHA256 hashing isn't really all that complex I wouldn't expect major breakthroughs going forward.

Short summary: difficulty increases in Bitcoin are going to slow down, and at some point we'll even see difficulty decrease. That will happen when the price and difficulty make it unprofitable to continue mining for a bunch of ASICs, at which point they get shut down (just like my AntMiner S2 ASICs are now resting peacefully). The only thing that will get those ASICs turned back on is higher BTC prices to compensate for their power use. And if we see a massive jump in Bitcoin prices to $10,000 or whatever, you can bet that difficulty will skyrocket as well.

I of course make no claims of being able to accurately predict the future difficulty or price of Bitcoin, so when I say "best-case, the current ROI for buying a Hashlet Prime is 429 days", that means that if you buy a Hashlet Prime at $34 from the Hash Market and Bitcoin stays at $338 with difficulty also staying static (which means the payouts stay at 0.0003 BTC per MH, plus 50% for Double Dipping), you'd need 429 days to break even. If the payouts drop the time to ROI will increase, and if the price of BTC goes up the time will decrease.

Hopefully that clears things up.

No comments:

Post a Comment