Learning, noun: knowledge or skill acquired by instruction or study.

If we were to take these two words, machine and learning, and join them together, we would get two key observations. First, a machine that is able acquire skills by instruction (an algorithm) or study (data). Second, a machine that can use these learned skills to perform a task and thus provide value. At the definition level, this seems quite mundane. But at an implementation level it becomes much more profound.

Programing Today

Let’s first step back and understand how software, via programs (instruction), are often built today by programmers (who have skill) today. This process usually starts when someone has a particular task in mind that they wish to automate:

1. Customer will input data into a form on a webpage.

2. This input data is processed by rules (the program) to compute an answer.

3. The answer is given to the customer.

If we make a diagram, the process would look something like this:

This simple process powers much of the digital world that you interact with today. Sending someone money, buying food with a mobile phone, uploading a picture to a social network; it’s all using input data and rules to produce answers.

The biggest problem with this process is, who creates the rules?

In the digital world, people are producing lots of new input data. All those picture uploads, dinner orders, and website clicks are collected in a vast databases just waiting to be used for, well, something. But to do that, we would need to build more rules. If every mobile phone user and website clicker is making input data, who is going to build these all rules to make these answers? Well, that would be programmers.

There is a scalability problem here: are there enough skilled programmers that can make rules fast enough to keep up with all the people who are making all this input data? At this point, not a chance. Skilled programmers take decades to train (k-12 included) while software has gotten (arguably) easier to use. As such, we are seeing the volume of input data grow well beyond the capacity of skilled programmers who are able to create rules that could produce quality answers from all of this data.

Machine Learning

Creating the rules has become the bottleneck. This is where machine learning comes in. What if we could give a machine high-level guidance on how it can find patterns in input data so that it could make new rules for us? That would start looking like this:

• A machine receives lots of input data (training observations) and separately also receives their answers (training labels).

• An algorithm studies these training observations to find patterns (fitting) and tests these patterns by creating an answer (prediction) for each training observation.

• These predictions are compared to their training labels (validation). If the predictions are far off, the algorithm restudies the training observations to find a better pattern; if the predictions are accurate, we have effective rules (model).

This continuous process of fitting, prediction, and validation to produce the best model from observations and labels is known as training. When we have our model, we can then input a new observation into it and compute (infer) a new prediction. In this way, we now have an algorithm that can turn data into rules.

Because we are giving the algorithm both the observations (to fit) and labels (to validate) we call this supervised machine learning. We can also train our models without labels in certain circumstances, which would be unsupervised or reinforcement machine learning.

From here, we further break out the type of machine learning algorithm based on the type of prediction we want our model to infer.

• Categorical: the prediction represents a specific thing, such as a dog or an application approval. When we use supervised machine learning we call these classification algorithms. When we use unsupervised machine learning we call these clustering algorithms.

• Continuous: the prediction represents a measurement, such as a price or an amount of inventory. These mostly use supervised machine learning and we call these regression algorithms.

What’s Next

We’ve discovered something very powerful in all of this. The vast amount of data that businesses are collecting can now be transformed into new programs with far less effort and cost then ever before. Furthermore, many problems can be solved far better with machine learning then conventional programing (such as voice and image recognition), allowing for previously impossible software to now become a reality.

This is Byte Club, a newsletter about real-life digital transformation by people who’ve led the largest and most complex firms through it. Subscribe now to never miss a post.

Subscribe now

The story around BITO has more to do with digital transformation then with Bitcoin. BITO is a transitional product that shows ProShares deeply understood the complex problem space around cryptocurrency, financial markets, government regulation, and customer sentiment. They then launched an incredibly successful product by effectively solving the most important issues in those areas. But what did ProShares actually do here? Why is a Bitcoin fund that will never own Bitcoin interesting?

Let’s deep dive!

The Current State of Trading Crypto

Can’t you just buy Bitcoin from any number of crypto exchanges already out there? Yes, but enough customers thought these exchanges had sufficient problems to justify a different way to gain exposure. Starting off, what issues do crypto exchanges have compared to traditional brokerages? Let’s compare Coinbase with Fidelity to find out.

A big problem with many crypto exchanges is their unreliability during trading hours. Some will say crypto exchanges are more available due to allowing 24/7 trading, while they are online of course. That is fair, but a key consideration is if the times you expected to trade are times you are actually able to trade. During wild price swings, arguably when you would want to trade the most, I’ve often seen downtime screens.

Another problem with crypto exchanges are fees. Traditional brokerages have all but eliminated trade fees for quite some time. As such, getting into and out of positions is cheap, letting you take full advantage of an asset’s volatility to make financial gains.

Taxes and financial reporting are another big feature where traditional brokerages win out. For taxable accounts, I can manage my cost basis for each trade with a granular SpecID4 method, simple average cost method, or a purchase time based method. For Coinbase, I have to manage cost basis myself on a spread sheet; making frequent trades a tax-time challenge. Furthermore, Fidelity offers multiple types of tax advantaged accounts, where frequent buys and sells have no tax impact at all. Until Coinbase offers an IRA or 401(k), traditional brokerages are going to win out here.

Finally, being an SEC approved ETF means BITO also has a deep options market. Options allow you to go long with additional leverage or short without the risk of unlimited losses. You can also transact complicated options strategies with traditional brokerages where you can bet on complex price characteristics such as volatility.

Coinbase does win over traditional brokers where fees would be lower for buying and holding long term. That is because BITO charges fees via an expense ratio instead of charging fees upfront during trades. Buying and holding $30K worth of BITO for a year would result in the equivalent of $285 in fees, $135 more than Coinbase. Unless of course, you also sold $30K of Bitcoin on Coinbase during the year too, resulting in another $150 in fees on the way out, totaling $300 in fees. This is where fees can get tricky. In general though, if you plan on trading Bitcoin, BITO wins out. If you are buying to accumulate Bitcoin long term (not trading), one time fees are a better deal.

There is another interesting reason why someone may want to have a lot of money in BITO versus Bitcoin in Coinbase. Crypto exchanges can get hacked5 and any Bitcoin you have in an exchange such as Coinbase could get stolen. Some say that any Bitcoin stored in an exchange like Coinbase is naturally at risk, and that you should instead store your Bitcoin in a wallet you secure yourself. However, self storing Bitcoin has two major issues. First, you can’t trade Bitcoin that is not in an exchange. Second, you could end up losing Bitcoin you secure yourself anyway6. As BITO never has and never will own any Bitcoin, there is never a chance that it can ever be hacked and the Bitcoin it doesn’t have stolen. In this way, BITO offers a superior type of investor protection where you can completely ignore the issues around owning and securing Bitcoin. That then leads us to ask how we can get Bitcoin exposure without Bitcoin?

Understanding Futures

If BITO does not hold Bitcoin, what does it own? On the ProShares BITO homepage, we see the fund holdings description shows Bitcoin futures contracts. Futures are a financial contract obligating a party to buy or sell a particular commodity at a predetermined price at a specified time in the future. Futures are a type of derivative trade, as opposed to a cash trade. Cash trades are when trade and settlement happen at the same time, such as spending USD and getting BTC. Derivative trades are when trade and settlement happen at different times according to terms within the contract.

Let’s start with a contrived example. You want to buy a 1 ounce gold bar 1 year from now. Right now, gold is $1,827 per ounce. The problem? You either don’t have $1,827 right now or you don’t want to take ownership until 1 year in the future. However, the price of gold will most certainly change in that time. A futures contract means you agree now to buy 1 ounce of gold at $1,827 per ounce (trade) and will pay for that gold bar and thus receive it 1 year later (settle). If gold goes up to $2,827, then you save $1,000 by only paying $1,827. If gold goes down to $827, you will have to pay an extra $1,000 instead, also at a total of $1,827. There are two distinct steps, the trade (making the contract) and settlement (payment and taking ownership). Price fluctuation may mean you win or lose out. But either way you get to lock-in the future price, today.

Back to BITO. The first futures contract, BTCX1, has a maturity date of Nov 26th 2021. That is when BITO is obligated to settle the trade. As BTCX1 is managed by CME group, we can checkout the contract specs on their website7. We see each contract is an obligation to settle 5 Bitcoin on the maturity date for $61,380 each. We can now calculate the exposure value for BTCX1. With 3,885 contracts x 5 Bitcoin each contract x $61,380 each Bitcoin we get $1,192,306,500, or about $1.2 billion.

How do we measure if $61,380 is a good price for Bitcoin on Nov 26th? The contract specs call out the CME Bitcoin Reference Rate8, or BRR. CME uses this rate to determine the difference between the price the contract holder is obligated to settle at and the market rate when the contract matures. For example, when BTCX1 matures and the BRR is at $65,000, that would mean each futures contract would have $3,620 x 5, or $18,100 worth of capital gains. Likewise, if the BRR is at $55,000 when BTCX1 matures, that would mean $6,380 x 5, or $31,900 in capital losses. Because futures contracts have a transparent way to measure gains or losses, the contracts themselves can be traded on the exchange in a similar fashion to the underlying asset.

So, when Nov 26th comes by will BITO buy $1.2 billion in Bitcoin? No, because BITO will never own any Bitcoin. Before BTCX1 matures, BITO will sell all of the contracts on the exchange and book capital gains or losses based on the difference between the contract price and the BRR. BITO will then use the proceeds to buy another set of contracts that mature further in the future: probably more BTCZ1, which matures Dec 2021 at $61,900; or perhaps some BTCF2, which matures on Jan 2022 at $62,290. This process is known as futures contract rollover and allows BITO to only expose itself to the price movement of Bitcoin while never actually owning any Bitcoin.

Furthermore, if BITO were to retain these futures contracts at maturity, they still would not own any Bitcoin. That is because these Bitcoin futures contracts are cash settled, which means the two parties would only settle for the difference in USD between the reference price on the contract and the BRR at maturity. The only reason why BITO rolls them before hand is to reduce the legal overhead of actually settling each futures contract. Before maturity, futures contracts are financial products that trade on the exchange; after maturity, they become legal contracts settled by law.

Spot vs Futures

You may wonder why BITO is structured as a futures contract based ETF instead of an ETF that just buys Bitcoin, also known as a spot ETF. First, as we discussed above owning Bitcoin means managing Bitcoin. A Bitcoin spot ETF would have to buy and sell Bitcoin whenever purchases or redemptions are made, and thus be reliant on crypto exchanges. The ETF would need to secure any Bitcoin held while also having enough Bitcoin ready to sell during redemptions, introducing additional technical risk. The ETF would also assume the price fluctuations of Bitcoin transaction fees9.

The second reason relates to government regulation. All US ETFs must be approved by the SEC, which advised that a futures based ETF would be preferred precisely as Bitcoin would not need to be managed nor would reliance on crypto exchanges be necessary10. The SEC trusts cash settled CFTC11 regulated Bitcoin futures managed by the CME group more then crypto exchanges or the process around transacting and securing Bitcoin. Some say that Bitcoin spot ETFs are secure and perform well, using BTCC12 on the TSX13 as an example. The counter point is that the TSX is a fraction of the size of the NYSE thus meaning far less chance for financial contagion14 and that BTCC is still much smaller in assets and volume even though it launched months earlier. These opinions are controversial, but at the end of the day all US ETFs need approval from the SEC. ProShares heard the SEC and followed their rules for BITO.

Are futures based ETFs superior to spot ETFs? The inherent indirection of futures contracts does create its own set of issues. First, a spot ETF would track the price of Bitcoin more precisely. Futures contracts can have interesting pricing mechanics which would make the price of the ETF differ significantly from the price of the underlying Bitcoin it tracks; known as trading at a discount or premium. Some traders like this pricing dynamic because it creates improved buying or selling opportunities. However, investors who want to accumulate BITO might end up buying at a premium and losing out on significant returns if it subsequently trails the Bitcoin price at a significant discount. Finally, the futures market is not without constraints as well; BITO quickly became so popular it actually breached the CMEs own contract limit15.

Conclusion

Shortly after BITO launched, Valkyrie also launched a futures based Bitcoin ETF16. More are sure to follow, and we’ll probably see spot ETFs in the future as well. What I find most fascinating is how firms such as ProShares and Valkyrie saw the benefits of traditional brokerages and launched products that effectively navigated the complex challenges of crypto currency, financial market structure, and government regulation to give customers a new way to trade Bitcoin. The tradeoffs for BITO will change as crypto exchanges mature, traditional brokerages evolve, and regulations update. For now, BITO offers investors a new way to trade Bitcoin via platforms they know best.

This is Byte Club, a newsletter about real-life digital transformation by people who’ve led the largest and most complex firms through it. Subscribe now to never miss a post.

Subscribe now

Transformation, noun: a fundamental change in character, usually for the better.

Digital is one of the most spuriously used words of the 21st century, with perhaps transformation being right behind it. However, there is something powerful about the digital medium that is transforming the world around us. From digital services to digital currency, what does digital really mean?

For most of human history, we’ve represented information via physical mediums: chisel marks in stone, ink marks on paper, or sound waves from our mouths. We would settle transactions with paper bills or metal coins stamped with busts from a bygone era. We moved folders full of files across desktops and mailed envelopes with letters across land and sea. Civilizations that mastered the physical world transformed humanity in ways that last to this day. After all, coins were invented around 1100 BCE (over 3,100 years ago) and are still used to pay for stuff today.

Physical medium has limitations though. We want to do four things with information:

• Store, we want to keep it around;

• Process, we want make changes to it;

• Move, we want to share it with others; and

• Display, we want it to be comprehensible.

Let’s take our stone tablet with chisel marks: storage is expensive, changes are difficult, movement is slow, and the person reading it must understand the markings.

Digital Changes This

The world digital comes from numerical digits; and for now, the digits that matter most are 0 and 1. Information represented in digital form immediately gains immense benefits across all four things we want to do with it:

• Store, we can cheaply record information as data;

• Process, data can be modified efficiently via compute;

• Move, data can be sent quickly anywhere via a network; and

• Display, data can be presented via interactive interfaces.

Let’s look at the English Wikipedia as of today:

• At 6,295,404 articles a physical representation would clock in at 2,657 volumes; a digital version weighs in at less than 20 GB, the size of a thumb drive.

• A physical Wikipedia would be difficult to search, requiring a massive card catalog; the digital version takes only seconds to pinpoint an exact article.

• Shipping a physical Wikipedia would take a semi-truck; the digital version sends the most up-to-date article right to your phone, instantly.

• A physical Wikipedia article would only be static text; the digital version has clickable links, video media, translations, and the ability to suggest edits.

A digital encyclopedia is much more than more efficient, it has features that could not exist otherwise. Because data is so cheap to store, it can be replicated across the world to ensure it is never lost. Because data can be processed efficiently, searching is no longer a time consuming activity. Because data can be moved efficiently, you now have the most recent version in the palm of your hand. Because data can be interactive, you can click links to new articles, sort tables to find information, or update articles in real-time when events change. Wikipedia did more than just digitize the encyclopedia, they digitally transformed it into something much better.

More Than Change, Transformation

Change just means doing something differently. Wikipedia could have digitally changed the encyclopedia by scanning the pages of an existing set of books and letting people download them from the internet. However, that would not have taken full benefit of the digital medium. Scans take up more space, can’t be searched, take more time to send, and are not interactive. Change is not good enough.

Transformation is fundamental, it’s rethinking the most basic concepts of something and reinventing it to take full advantage of a new capability. The digital medium offers so many incredible benefits, but only if you are willing and able to imbue them into every aspect of what you’re working on. Holding out only means holding back.

For the past thirty years, firms that fully embraced digital across every aspect of their business have become the most valuable in human history. Firms that have held back are now gone. Many however, have only adopted digital capabilities at the edges or in isolated pockets. Digital transformation means more than just using new technology; it means new ways of organizing employees, understanding customers, analyzing information, building products, and funding projects. It means more than just changing how you do business; it means reconsidering what your business is.

What’s Next

In the past ten years, the digital world has accelerated its pace of innovation and shows no time of stopping. Cloud computing makes expensive on-premises data centers obsolete; machine learning models are leaping ahead of conventional software applications, more employees are working remote across the world, and customers are demanding integration with new services every day. It is not enough to just adapt to these needs; you also have to adopt a new culture that is able to embrace this accelerating tempo and forecast what the next thing will be.

That is where Byte Club comes in. It’s time to for digital transformation leaders to open up, compare notes, and show what works (and what doesn’t). Byte Club is about stories showing the practical work of reimagining firms as digitally native enterprises.

This is Byte Club, a newsletter about real-life digital transformation by people who’ve led the largest and most complex firms through it. Subscribe now to never miss a post.

Subscribe now