Fine-Tuning Our Tool: The Metcalfe Law Discount Premium Model Gets New Glasses

We updated yesterday our MLDP Z-Score model for the normalized version I used in most of my recent analyses. I think this model is one of the most powerful tools we have in our arsenal that can actually be used no matter where we are in a Bitcoin cycle. Here is more info about the 'why' of this update and what the impact of this change is.

On-chain Indicators

Like I wrote in our eBook, when Jennifer and I started to work on the Bitcoin on-chain metrics-based algorithm, we had two objectives in mind. The first one was to find out of all the on-chain metrics, which one was forward correlated with the price action and that could be used to anticipate main events. The other objective was to craft out of these data different indicators that would give a consistent reading over time. This part was particularly important for me as I had previously seen great analysts who missed calling the top due to data that was drifting. As an example, in late 2021, David Puell, one of the best analysts in Cathie Wood’s team, used NUPL (Net Unrealized Profit/Loss) and MVRV (Market Value to Realized Value) to state that we had still some way to go in that cycle, as both indicators were not at levels consistent with previous tops. Here are these metrics across the last few cycles.

I’m sure you are noticing how these data seem to have decreasing amplitude from one cycle to another. Therefore, it is hazardous to use these metrics for predicting the top or bottom. Four instances don't statistically firmly set a trend, but at least the past suggests that we may have to expect that the top for this cycle should be smaller on these metrics than during previous cycles. I am not telling this story to take credit on top of someone else's work. Like I said above, David Puell is a pioneer of on-chain metrics, and you should probably follow him on Twitter (Cool thing: he has a Ph.D. in Physics). We all make mistakes, but this one shows one dilemma when working on on-chain metrics. On one hand, we want the indicator to be simple enough that it represents a well-understandable aspect of the Bitcoin Network. Let’s take the MVRV. It makes sense to compare the market value to the realized value, and this is a metric that can be easily understood. The realized value, being the average value that people actually paid for their Bitcoin (on the blockchain we can track every bitcoin and look at the last price it changed hands, so we can compute this average price), it can be interesting to compare it to the actual price as a representation of the level of profit in the network. On the other hand, we can easily understand why this indicator spikes with decreasing amplitude at each cycle: as the market cap becomes bigger and bigger, Bitcoin loses its volatility and it doesn’t run as far as it used to from its realized cap. A bit like an optician can find the right glasses to correct someone's myopia, we can find the adequate variable to correct that drift, but by doing so we kind of lose some of the initial sense.

In contrast to this, here is our Exchanges Money Flow Bottom indicator and our Kwiatkowski Market Bottom signal that have specifically been designed to be consistent across each cycle.

These are the two responsible for successfully calling the bottom in November 2022. I know you may tell yourself, 'Vincent, the Kwiatkowski does seem to record a decreasing signal at each cycle.' This is true, but the indicator is so nonlinear (only spike around the bottom) that it actually doesn't really matter (at this moment) in this case.

But like I said above, the issue with these signals is that by putting a corrective lens on the data, we kind of lose some of the sense. You should see the equation behind these signals. It’s a computer that found it because we would never have been able to come up with such an equation. All I can say is that the Kwiatkowski is somehow related to NUPL and MVRV, but I cannot say much more since the equation is so complex and involves many other data points that make it hard to actually capture its meaning.

Data scientists, such as David Puell, face a notable dilemma: basic relationships in data can provide valuable insights into various layers of the Bitcoin ecosystem, yet they are prone to drift. On the other hand, while normalized indicators offer greater reliability, they risk losing some of their intrinsic meaning.

MLDP Z-Score

Our Metcalfe Law Discount-Premium model did suffer from that issue. For those that don’t remember, the Metcalfe law states that a network's value is proportional to the square of the number of network participants. So, this indicator compares that estimated value with the current value (market cap). This is a pretty simple and straightforward metric that we then post-process in a way that was found optimal by our optimization framework. This last part, which is our secret sauce, explains why our model has a different output compared to some other Metcalfe law models. When I designed it, I did see that it has a decreasing amplitude, but since we didn’t use it for capturing the cyclical top, I decided to keep it simple

That being said, I think this is one of the most useful tools we have, and I understand more than ever why some major funds make their purchase decisions based only on such a model. Also, with two more years of history, I finally decided to craft its needed lenses. So here it is.

Without going into too much detail, the solution was relatively simple this time, as we mostly had to account for the existing supply.

Impact of this change:

Within the Bitcoin strategy (Active Holding signal), one of the only uses of the MLDP Z-Score is to exit the market if we register a peak above a certain level when we are still in the first half of the green zone (gauge by the value of the Kwiatkowski Top indicator). Therefore, even if with that correction the recent peak would have been above this threshold, we were already too close to switch to our Blue environment for this modality to trigger a sell signal. The other scenario where it could be useful within the strategy would be at a cyclical top if it ever records an outstanding level like in 2017. Therefore, at the moment, this change does not really affect our strategy nor the recent past of our strategy. I should update the strategy soon to account for this normalized version of MLDP, but I currently see no rush for that.

However, where it could have an impact is on decisions we each personally make about increasing the size of our position or selling a position. Indeed, there is our algorithm, and there are also all the decisions we all personally make about how to play Bitcoin. Like I said, there are plenty of reasons to sell or buy Bitcoin when WU is not doing anything, and I honestly believe that this indicator is a powerful tool that could help a Bitcoin investor in their own decisions. By making it more consistent across years, I hope this will help you get a clearer picture of where we are in terms of oversold and overbought conditions. I hear you saying, 'We have RSI for this,' but unlike RSI, which can remain at a very elevated level for a long period, look how this signal only briefly spikes. This is why we used it recently (and last December) to say that we were going to have a pullback.

I kept the old version alive. That is maybe more a feature for myself, but in case you would like to switch to the old version, simply go to the indicator parameters and click on 'Old Version'. Also, as usual, you may have to restart Trading View to access this new version. I know, Sorry, it pisses me off too, to restart Trading View !

I hope this will help you in holding Bitcoin efficiently. Looking forward to seeing that indicator above 4, and I should be a buyer if I ever see it again under -2!