Is Bitcoin worth the gamble?

We've been working through the four most commonly used methods for valuing bitcoin.

The total addressable market method and the network effects model tried to quantify the level of demand for the asset and how that might change over time. Meanwhile, the stock-to-flow model and the cost of production framework measure changes in bitcoin’s supply and how those could affect its price.

The price signals they produced are still not particularly useful.

So why did we go through with it?

I’ll answer that question with a little anecdote. As a child, I lived very close to a horse racing track, and during the summers my dad and I would often go to the track together, poring over the stats sheets and inhaling the language of horse racing.

There was one day early on when my dad was still showing me the ropes. I was about six. He asked me about the number 5 horse, which had some promising races earlier in the season and had a reasonable opening line. The opening line is established by sophisticated bettors who look at the race before the general public can bet on it and assess the probabilities for a given horse.

I looked up at the screen to see where the odds stood currently, and I saw that a flurry of betting activity had pushed the odds down significantly from what we saw in our program. Even if we guessed correctly, the horse wouldn’t make us that much money if it won. Too many other people had already jumped on the bandwagon.

According to legend, I looked up at my dad and said, “I like the horse, but not at those odds.”

With bitcoin we have access to an ever-changing odds board, but we don’t yet have an opening line. If you like, think of each valuation method as a different Las Vegas shark, synthesising his data and spitting out his best guess for bitcoin that day. We may not find any one of them particularly insightful, but by putting them all together we can craft a story even a six-year-old can understand.

More on that story in a second. But first, let me boil down our findings into three key takeaways.

1. The inputs for Bitcoin’s valuation models are blunt instruments in need of refinement

Behind the wall of formulas, most of these models rely on very basic math. That’s no knock against simplicity—for valuation models, ’tis a gift to be simple. But for simple models to bear fruit, it’s important that all the inputs are bulletproof.

The inputs for cryptocurrency were about as bulletproof as Swiss cheese. That’s perhaps to be expected because in many cases this type of analysis had never been done before.

For some models, the issue is that we are using a model borrowed from another discipline. To get the numbers to work, the authors had to be scrappy and use variables that are imperfect substitutes.

Take the network effects model. That one is borrowed from stock analysis, which means that we have to shoehorn bitcoin’s tangle of on-chain data into a formula designed for data that’s fastidiously reported in 10-Ks. We had to use the number of wallets on the blockchain as a substitute for users when a significant chunk of crypto investors actually outsource their holdings to the exchange they use.

For other models, modest tweaks change the nature of the results entirely. The total addressable market approach promises sky-high valuations for bitcoin, even if it nabs just a sliver of gold’s market share—or so it seems. But in practice, half of the world’s supply of gold is used for jewelry, and I for one won’t be wearing bitcoin on my wrist anytime soon.

If you account for jewelry, bitcoin’s market share as a percentage of gold’s has already exceeded the 10% threshold broadcast by analysts.

Fine-tuning the measurement of the underlying variables could improve these valuation models’ accuracy without altering the overall methodology. The other two issues we uncovered won’t be so easily rectified.

2. Currently, many of these valuation models rely on constants when they shouldn’t

Take the cost of production model that we did as an example. In that illustration, we attempted to value bitcoin based on the amount of energy it takes to mine it.

In order to get a rough approximation, we took an average cost per kilowatt-hour. The problem is that cost not only varies for each individual miner but will also change throughout time based on global energy markets.

For other models, the issue isn’t so much that the absolute value of the variable itself changes over time. In our stock-to-flow model, for instance, the stock-to-flow ratio that underpins the model appears to be somewhat stable.

The issue is more to do with the fact that the relationship to bitcoin’s price changes. Stock-to-flow’s explanatory power appears to derive from its influence during halving cycles, where the number of bitcoins minted during each block gets cut in half. After that, the stock-to-flow ratio adjusts to the change, and the relationship splinters.

From that point on, anything goes. Bitcoin’s price and the stock-to-flow ratio move entirely independently of each other. Stock-to-flow continues to rise gradually, while bitcoin’s price bounces all over the place.

A more sophisticated analyst than I could easily swap in a time-varied price per kilowatt-hour and resolve at least some of our quibbles with the cost of production model, but someone looking to resolve the issues with stock-to-flow has their work cut out for them. Halving cycles, where most of the explanatory power of the model comes from, are like the Olympics—they occur just once every four years.

Supposing there is a relationship, how can we explain what goes on during the intervening three years? The initial authors of the study didn’t have a good answer to that question, but in fairness neither do I. It’s crucial to address, though, because of the last issue that we identified.

3. All of these models for Bitcoin depend at least in part on a relationship that increases with time

In their current state, none of the models can explain the acute drawdowns that have haunted bitcoin since its invention. This leaves every model susceptible to flights of fancy, moderated only by an analyst’s judgment.

The total addressable market method argues that bitcoin will take over part of gold’s market share, implying growth.

The network effects approach is built on the premise that bitcoin can generate exponential price growth even if the number of wallets grows at a linear rate.

No matter what happens, stock-to-flow will rise at a constant and predictable rate.
If the cost of production outpaces bitcoin’s price, the model assumes that equilibrium will be restored either because higher-cost producers exit the industry or prices rise.

Wrap-up

The issues surrounding these bitcoin valuation techniques aren’t insurmountable, but it is telling that few have attempted to tackle them. It certainly seems like there isn’t much appetite for improving on the current catalog.

Now, that’s not the models’ fault. Each one represents a first stab at valuation. At this stage, it’s perhaps to be expected that a columnist with a week or two on her hands could poke holes in them. But that’s why peer review is so important: Because that’s where refinement gets introduced. This makes the absence of it in the years since these models were published all that more glaring.

Given the lack of peer review, perhaps the reason these models gained traction is not necessarily because they were good at producing a price signal. Perhaps they became popular because opportunistic cryptocurrency maximalists could use the outputs of these forecasts to tell potential investors what they want to hear.

Maybe that’s too skeptical a read. It could also be that people don’t think forecasts for bitcoin are particularly useful, and that’s why they don’t make better models. Maybe we’ve all just accepted that bitcoin is a commodity that is more easily priced than it is valued, and these models are just shouts into the void—albeit explanatory ones.

I find this argument more convincing, but I don’t think we should accept it as the status quo. To me, it seems like bitcoin has a lot in common with the number 5 horse from my days on the track. Yes, there’s a big element of chance to both gambling and crypto. A horse is only as good as its appetite to perform that day, and bitcoin’s price climbs only to the extent that investors wake up and will it to keep climbing. But at least in my eyes, a well-crafted opening line is still valuable. After all, a good horse isn’t worth much if it’s overrated.