Thursday, February 6, 2014

What Makes a Good Cryptocurrency, Part 1: Slow and Steady

I’ve discussed some of these thoughts before, but I wanted to get into this a bit more. Vertcoin is a great example of how to make a useful new cryptocurrency, and the reasons for its current success are due in a large part to the design. Fundamentally, I think any good cryptocurrency needs to have the following:
  1. It needs to do something new relative to what is already out there.
  2. It needs to launch in a “fair” manner.
  3. Difficulty adjustments should happen sooner rather than later.
  4. It needs to be designed with the future in mind.
I’m sure there are other things that I could put on the list, but I'm going to focus on point number four right now. There's an old saying: "slow and steady wins the race", which we're all familiar with from the story of the tortoise and the hare. While the latest meme cryptocurrencies may catch fire for a short time, I don’t think they’re a strong basis for a lasting currency. And DOGE, I hate to call you out like that, but long-term I don’t think your prospects are all that bright (similar to your silly Doge face – such wow and all that). A primary reason is that the coins get paid out in a short amount of time and we reach the point where the block rewards largely disappear.

Of course, "short" is all relative, but to me a coin should be designed such that it will still have a reason for mining (i.e. securing the network) in five, ten, twenty, etc. years. If as an example you have a new currency with one million total coins and they'll all be mined in six months, what will keep miners going with securing the network past that point? If transaction fees of 0.1 coins per transaction were mandatory, and if there are on average 10 transactions every block, that would mean even if you're doing a block every 30 seconds, you're still only giving out 2880 coins per day. So the people that mined the initial 1 million coins of this hypothetical currency did so at let's just say a steady rate of 5555 coins per day, and if you had a minimum guaranteed transaction fee of one coin, it perhaps wouldn't be so bad...but that's not really what we have on most coins.

The reality is that transaction fees are far, far lower than 1 coin on most of the major cryptocurrencies. Take Bitcoin as an example -- it's one of the most heavily used coins, and yet looking at the past 20 or so blocks, the biggest block I could find was this one. That has 826 transactions and only 0.21444776 BTC in total fees. If we were depending wholly on transaction fees, all of the power going into the Bitcoin network would only amount to on average something like 0.05 BTC every ten minutes, paid for by those conducting transactions.

Right now, the 25 BTC block reward means there's a bounty of around $20,000 that will go to some lucky miner (or pool) on average every ten minutes. The total network hash rate of Bitcoin has now reached a pretty staggering 19,720,113 GHash/sec. Let's assume for a second that every system participating in Bitcoin hashing is as efficient as the latest and greatest 28nm ASICs. That would mean world-wide, Bitcoin is sucking down around 11,503,400 Watts of power. At a relatively inexpensive $0.10 per kWh, that means in a day Bitcoin consumes $27608.16 worth of power -- not too bad, as the current block reward will pay that in just over 10 minutes. (Realistically, most ASICs are far less efficient so the power cost is probably twice that -- so 30 minutes to pay for all the power use of BTC.)

But what happens in the future, like say in 2030 when the block reward of Bitcoin will probably be at the 0.78125 BTC mark? Most likely we'll be seeing a lot more transactions on the Bitcoin network, so instead of 0.05 in average fees per block, maybe we get to the point where the average transaction fees per block amount to 0.5 BTC (which is probably a bit of a stretch). At that point, we're looking at perhaps 1.25 BTC every ten minutes, and the power use of the BTC network may not actually drop much (and more likely it will increase). What happens then?

1.25 BTC per block right now is still more than enough to cover the cost of power -- in fact, 11.5 MW of power costs something like $200 per block, so at current prices we would only need 0.25 BTC per block for those securing the network to break even. If BTC is worth ten times as much in twenty years (which is either optimistic or horribly pessimistic), an average block reward of 1 BTC with transaction fees will be enough to power a while lot of hashing, so the network stays secure and BTC can continue to succeed. It was designed with this sort of scenario in mind, which is why things should continue to function well. But that's for Bitcoin; what about other cryptocurrencies? Time to pick on DOGE for a minute.

The total number of blocks before the block reward drops to 10K + transaction fees is around 756,250 blocks -- or in just 525 days from the time DOGE first started. Looking at the past day of blocks, here's one of the largest; with a total of 617 transactions, there were just over 575 DOGE paid in transaction fees. Right now the total network hash rate of DOGE is around 86 GHash, but it's happening almost entirely with GPUs. Assuming everyone is using the most efficient GPU possible, so an R9 290X hashing at 900KHash and drawing 350W, that means the DOGE network is drawing about 34,000,000W (and in reality it might be 50-100% more than that due to less efficient GPUs). With an average of 500,000 DOGE produced every minute, that's 720 million DOGE per day, with a value of roughly $1,000,000. Meanwhile, the power cost for the DOGE network is around $81,600 per day, so clearly DOGE is more than paying for the power use. But what happens when the block reward drops to 10K + transaction fees?

With the largest block of the past few hours generating 575 DOGE, it's probably a safe bet that best-case we're looking at 1000 DOGE or less per block in transaction fees. That means 15,840,000 DOGE per day, so to break even on power costs of $81,600 per day DOGE will need to be worth at least $0.0052 per DOGE, or in BTC terms it would need to trade at around 0.0000064 BTC per DOGE. That's only about four times as much as the current value of DOGE, so we can certainly hit that level, but again that's just to break even. If other coins are generating a substantial profit, why would people stick with DOGE just to break even on their power costs? I'd say bare minimum it would need to consistently generate twice as much revenue for those mining (securing the network) as it costs in power, and perhaps 2-3 times the return would be better. Will we see DOGE trading at 0.0000192 BTC/DOGE? Possibly, but more likely a new meme will supplant DOGE before then.

Put another way: if you believe DOGE will manage to maintain current hash rates for the next two years, you'd be a fool to sell any of your DOGE at the current prices. All the "DOGE millionaires" (currently around $1300 worth of DOGE) would be looking at the equivalent of $10,000 or more if that happens. A nearly 10-fold return on your investment in under two years is "pie in the sky" sort of thinking in terms of investments, but yet cryptocurrencies are all beating that mark -- often by a large margin.

As usual, this is a bit long, but when you start thinking in terms like this it should help you to start seeing why coins that pay out most/all of their block rewards in a short amount of time are a bad idea. They start out looking pretty interesting and might garner some headlines and make waves, but a couple years from now I expect Litecoin will still be chugging along -- the little engine that could -- while most/all of the meme coins are going to end up fading away. And really, it's better that way in my book, as if I'm talking to friends or investors and trying to get them to understand that cryptocurrencies can succeed, the "success" of a joke coin like DOGE doesn't help at all.

Now if you'll pardon me, I'm going to go create the All Your Base Are Belong To Us (AYBABTU) coin. Does that sound old and stupid to you? Well, that's what today's memes will be in another decade. All memes die, and the meme currencies will die with them.

Getting back to the main topic, what I'm saying is that you need to build a cryptocurrency that will pay out block rewards long enough to reach the point where the transaction fees can actually sustain the network. Or you can be like DOGE and go with a deflationary approach and always have 10K DOGE per block minimums, forever. But that's what got us into the mess we're in with fiat right now, isn't it? As far as a long-term payout, there are plenty of ways to do that -- Bitcoin, Litecoin, and Vertcoin cut the block reward in half every 4 years or so while other coins might drop linearly over time. Coins that pay out too quickly on the other hand (DOGE, QRK, FZ, etc.) are very likely to reach the point where there's no profit in mining/securing the network. If that happens, the coin(s) will die. You've been warned -- don't get caught holding the bag for a poorly designed cryptocurrency.

EDIT: Note that I missed the fact that DOGE has a 10K minimum reward, apparently forever. I don't really like that as a solution either, if you can't tell. I've updated the text to reflect this with new calculations. Thanks to several readers for pointing out my errors!

9 comments:

  1. Did You consider this when writing about doge?
    http://www.businessinsider.com/dogecoin-just-solved-a-problem-bitcoin-will-face-2014-2

    ReplyDelete
    Replies
    1. I updated text to reflect this fact; no, I didn't know that DOGE had a minimum block reward of 10K forever. I thought it was 10K until we hit 100 billion DOGE, but I read that wrong. (Part of the problem of being a joke/meme in the first place is that some technical details were obscured.)

      Delete
  2. 100x returns isn't necessarily pie in the sky... Someone dumped 30k VTC for .00001 back when it first hit coinedup. I bought some of it, but sold it soon after for not much more. If I had kept all the VTC I mined I would be able to pay off my house right now..

    ReplyDelete
    Replies
    1. It's not if you're in early, but at this stage it's probably more than I would expect. Especially with a meme coin like DOGE. Anyway, DOGE could very well prove me wrong, but I'm just saying I like coins with a bit more substance for the long-term.

      Delete
  3. With regards to DOGE, I think you're missing something. The minimum block reward is 10k DOGE, not 0 plus transaction fees - this has been the case since the original announcement. At the 1.5 year mark (of roughly 100 billion doge), it will still be generating 14.4 million DOGE/day or 5.25 billion a year for 5% inflation (assuming little to no lost wallets and in reality there will be some for sure), not counting transaction fees (I do not know if those roll in to block rewards but I assume they do). This is honestly it's most differentiating and interesting feature, counter to the solely deflationary nature of every other crypto currency I'm aware of.

    Whether that's sufficient incentive to keep a strong network hashrate of miners at that point is still a separate matter...

    I quite enjoyed the post! Keep on writing!

    ReplyDelete
    Replies
    1. Edited the post, as noted above -- I misunderstood the block payout and thought it was 10K until the total circulation was 100 billion, but apparently there's no limit to the total number of DOGE. I'm not sure I like that -- it's what governments do with printing money, albeit at a steady rate. Quark does the same thing, and it's not really working out too well for newcomers; only those with large QRK holding continue to mine/secure Quark. Of course, Quarkcoin paid out 250 million in six months before dropping to 1 million per year, which is much (MUCH) worse than DOGE.

      Delete
  4. No what about bringing CPU only or ASIC proof coins in the discussion, do they have a unique place in the long term view of things? If so which coins have that outlook? Does that include Quark/ Frozen? Also in terms of doing something new? Are there devices that are not CPUs, GPUs, ASICs that can be used for unique mining or use?

    ReplyDelete
    Replies
    1. Well, there are FPGAs, which are basically in between an ASIC and a general purpose chip (i.e. CPU/GPU). They tend to be pretty expensive however and often don't substantially outperform GPUs.

      As for CPU "only" coins, it's mostly a question of taking the time to develop a GPU miner for each algorithm. CPUs are general purpose, but GPUs are pretty much in the same category -- there are only a few specific instructions that are really difficult to use on a GPU, so you could always make a miner that does most of the work on the GPU and a little bit on the CPU (if you have the programming skills, which I lack these days). Anyway, quark-based coins end up at the same place as scrypt-based coins in that you need to have enough value in mining to make it worthwhile for people to spend the electricity on mining.

      A quad-core desktop Intel Core i7 CPU will do around 800KHash on quark algorithms, which means with QRK right now you'd mine about 2.17 QRK per day. Your power use would be around 2.4 kWh, so $0.24, but the value of the QRK is only about $0.184 -- so with a fast CPU you're losing $0.06 per day you mine it in power costs. It's why I think QRK is basically dead.

      Delete