Sounds like a classic case of overfitting training data. Bitcoin and cryptocurrency as an asset class is so nascent that I wouldn't be making any significant bets on what the algorithm outputs even if the cross validation is good.
As with all investing claims, the pertinent question to ask is "if this works so well, why are you spending time marketing your method instead of making boatloads of money using this system you developed?"
I went to a tech conference this past year, one of the talks was regarding machine learning and investing. I remember the speaker being distinctly negative on the ability of machine learning to predict prices accurately such that his firm could make money. He was much more optimistic about using machine learning to understand the forces that driving prices
Markets adjust to information, so even if the algorithm was somehow accurate, people acting on the information will correct the market. And at some point it may even become self-reinforcing if enough people buy into this "black box" that can predict the price of Bitcoin.
If Bitcoin continued its 2017 growth rate for 5 more years, the market cap of Bitcoin would be about 140 quadrillion dollars, which is about 300x the value of all public and privately owned assets in the entire world.
Currently, bitcoin uses about 1 out of every 5,500 Watts being generated on the planet.[1,2] So, unless bitcoin is secretly a paperclip optimizer, that should offer some upper limit.
I think I had like $60 NZD in mBTC back in 2014. I pulled the backup key for my wallet, waited to sync 40 months, and was surprised to see over $800 USD. Like everyone else, I was I had gotten more bitcoin back then. But it all of us had, it probably wouldn't be worth as much.
Everyone lost their private keys over the years, you are the reason btc is so valuable today.
> A lot of people are investing expecting 1000x returns based on historical data, but that ship has long sailed.
You are right. A more reasonable expectation is a 5x return. Analysis below.
-----------
There have been approximately 187,200 tonnes of gold ever mined (+/- 20%, I know I know), which at current spot prices of $41.56 / gram means that the value of all gold ever mined -- almost all of which is held in circulation, or ownership -- is $7.7 trillion.
Bitcoin is a similar store of value in that its worldwide existence increases slowly. It is expensive and a pain to transfer bitcoin, but so is gold when used for monetary purposes.
Likely a good estimate for the "proper" market cap of bitcoin is $7.7 trillion, matching the market cap of all gold ever mined. While many bitcoins were lost, the bitcoins still in circulation are substantial in quantity and have a higher velocity than gold.
Bitcoin is susceptible to technical problems, but the price of gold is predicated on the difficulty of transmogrifying any other element (aluminum, say) into gold. But I include this paragraph just to say that in no way is there some metaphysical guarantee that gold will be equally scarce forever, just as there are issues that could happen to the bitcoin network.
For the next 3-4 years, a decent target for an "appropriate" price of bitcoin is parity with gold - $7.7 trillion, say, or about 40 times its current price.
In my personal opinion you should consider the fundamental price of bitcoin to be around 40 times its current price, which you should then discount by certain systemic risk.
If you have gold it cannot just disappear (without being stolen), but bitcoin can disappear if its network has some kind of systemic problem.
An extremely safe way to generate 5x return on any amount of money (up to about fifty billion dollars) in the next five years, is to buy bitcoin after finding an credible insurer who will sell a policy against a systemic problem that causes it to go to 0, or against your personal bitcoins being stolen.
The risk of the bitcoin network having some unforeseen problems is vastly undervalued today - nobody seems to consider that possibly it will not be a functional network in five years, at all.
In terms of technology few p2p networks survive 5 years. I would take a 5-year position on bitcoin only if I could adequately protect against this risk.
Bitcoin is certainly not a bubble and there is next to no risk of an adjustment of losing, say, 98% of its value, while containing to remain healthy at 2% of its current value by its legitimate users. It is not a ponzi scheme. It is not subject to sudden hyperinflation.
You do not have to worry about this eventuality if you are considering a long position in bitcoin. You don't have to watch its price day to day. You do have to have a very active policy against its going to zero for technical reasons. If you don't have a signed contract with a traditional, brick-and-mortar insurer, you should not have any position in bitcoin, period.
Nobody is giving technical network problems the correct probability of surfacing. They're idiots.
Let me put it in these terms for you:
-> I would bet better than even money that the price of bitcoin on December 3rd, 2021 will be more than 5x its current price, if it is at least 10% of its current price.
-> I would not bet even money that the price of bitcoin will be >= 10% of its current price.
Do you understand these two constraints? You can get a 5x return, easily - as long as you ensure against a total loss, which is very, very likely.
I get it, I just don't know who would sell such insurance? It'd probably be extremely expensive to purchase, extremely expensive to write the contract, extremely expensive to litigate when they want to claim the losses weren't caused by systemic tech failure, etc.
Those costs would take so much off the top that you'd probably need to make this play with tens of millions just to offset those costs and make the potential returns worthwhile.
Also, by your argument, if bitcoin is going to be treated as a stable deflationary value store like gold, then that should put downward pressure on the price of gold (because they would be competing asset classes). It should also mean that bitcoin will eventually stabilize and merely reflect inflation.
But the appeal of gold is that it's a hard asset and it's stable. Bitcoin is the opposite of that. Why would its price be based on it being treated like gold when it isn't and will never be anything like gold?
No real comment on anything you've written (which seems a good reply to me). Right, I think it could be quite expensive to insure. My point is that if you are investing $10 million into bitcoin today you will likely make a great return by Dec' 2021 if you actually spend literally $6 million of that (60%) on insurance that pays out only if it goes to less than 10% of its current price or gets stolen from you. So out of $10 million invested, only $4 million makes it into your position * 5x = $20 million, or 2x return over 4 years for an 18.9% year over year return. (1.189^4 = 2, i.e. the left side is the compound interest formula.)
I also have perhaps these two quite minor points,
- Why do you say gold is stable? (You write "the appeal of gold is that it's a hard asset and it's stable"). Here is its real price over just the past 10 years: https://www.kitco.com/charts/popup/au3650nyb.html - doesn't seem particularly stable to me. It seems more likely that the appeal of gold is 1) its historical use, and general acceptance as value, and 2) its inherent scarcity, i.e. its value reflects its scarcity.
- Also you say bitcoin would "reflect inflation" - but where do you think inflation comes from? Wikipedia says it's just a direct effect of an increase in the money supply, but bitcoin doesn't have such a mechanism.
What makes the market cap for gold a good estimate for bitcoin? I would like to know the price of bitcoin where the dollar cost of a transaction was roughly inline with a credit card transaction.
Another systemic risk: government regulation and/or shutting down exchanges to combat money laundering and tax evasion. Good luck insuring against that.
I just invested figuring that it won't crash until I buy into it, and I sorta want it to crash. Unfortunately, it has gone up another 20% since then so my strategy is failing.
Sigh. It's a bad sign because I never wanted to invest in bitcoin, even if it would mean enormous profit (not a fan of speculative profit). But, I just bought some. So what if other people equally skeptical and disinterested in profit should suddenly decide to buy in? Then all of us will unfortunately reap the profit of our own folly. :(
There's some solace knowing that I could just leave the BTC sitting in my "wallet" indefinitely, forever refusing to convert it into any other currency and refusing to spend it. Could serve well as a personal test to know how much I'm authentically morally opposed to speculative profit.
It not only predicts a "bubble burst," but also the price of Bitcoin in the near future. If such a system were actually accurate, what would compel anyone to share it publicly?
If a system is good at predicting prices and makes you money, why sell it to others. It would be worth millions.
The quant hedgefunds spend millions developing "black boxes" to carry out trades and guard the secrets within like it was their own life. And yet here we have these people giving it away or selling it for 2.99 on the Play store.
Hard to believe that any one who could credibly predict this would give it away, instead of making money with shorting Bitcoin.
Seems more plausible that this is itself something that's supposed to crash Bitcoin, so that the owner can make money selling it short. If it isn't, someone will hack it and ...
There is a difference between the naive sample size (just your raw n value), and the information content of your sample size.
How many crashes has bitcoin had? Depending on how you measure it, somewhere between 3-10? And even then, these crashes varied in magnitude, and some occurred at times when bitcoin was fundamentally different than it is now.
The point is, you can have the richest data sources you want, at any time granularity, but you can't solve a hard information problem, which is that you only have a small small handful of true points to train on.
The rest of the training data is going to be extrapolating or learning from small movements, which should have different dynamics from crashes.
To all those who claim that it can't predict the bubble burst accurately, you are probably right, but it may not matter at all to the author.
If this software gets seeded to many people - and presumably the target market is Bitcoin investors - the author may have a simple way to trigger a selloff. All you have to do is send a push to your app users saying "Burst About to Happen!" and watch the selloff happen from your panicky users.
If you can trigger a selloff at will, you can start playing the market (short).
The nice thing about it is that it's a self fulfilling prophecy at that point. You can trigger the sell off manually, and those receiving the notification will say "man, this app is really good, it really predicted it!", and will be conditioned to sell off the next time they get another notification.
Given the volatility of the market, this isn't as wild a theory as it sounds, especially when you account for not only the app's direct impact but also the following chain reaction of news stories.
> Using Artificial Intelligence to predict Crypto currency value.
The AI thing had me rolling on the floor with laughter. Some shitty TA stuff used to "predict" cryptocurrency value. The lengths people are going to incorporate the words - AI/ML into their products.
It crashed $1200 yesterday ($11868 -> $10600) when it approached $11868 which seems arbitrary until you realize that 10000€ = $11868, which triggered a round of sell orders.
reply