 still a bad news between the dinner and between you and the dinner is now the statistics part so if you think we are talking about Austrian economics throughout the day that there won't be any statistic part and I'm here actually to clarify a bit a lot of stuff that came up recently when it comes to statistics when it comes to Austrian economies and statistics in general but especially when you're applying statistics to Bitcoin so in December I came up with this term bit conometrics and I tried to coin their return for all the phenomenon that we have when we use economic thinking mainly Austrian thinking and then we are standing there with hypotheses and we don't have any further tools to validate them or to test them so the most important hypothesis obviously that we have right now is with the stock to flow model that we are saying the hardness of Bitcoin is driving its value and this in this target I want to go a bit back and I also want to present some new results but I think it's very important to start at the very fundamentals again because there's so much confusion so when we are talking about the stock to flow model we have to first ask yourself how is it from from what is it derived and it's derived from the from the question at the very top what is the demand function for Bitcoin what is X in this demand function what are the main features of Bitcoin that are responsible for its market valuation and I like the three criteria from Jaco Musuko for good money because they are not like 14 or 10 or 12 criteria they somehow compress a lot of subcomponents into three categories and hardness is the category that we are using or that we are looking at because we are deriving it from economic thinking and then we are trying to find a variable which is quantifiable because then otherwise we couldn't do any statistic tests and the variable that we are using is the stock to flow ratio why are we using the stock to flow ratio well the stock to flow is arguable a good measure because if you have an asset where only very little additional supply is coming onto the market like a drop in the ocean it's very reasonable assume that this is a good measure for the hardness of an asset because you can't easily inflate it the first confusion that arises is then well but wait supply is directly impacting demand and this is a major problem but as a polite polite mentioned that the differences between certain goods so you have consumption goods capital goods whatever and you have monetary goods and with monetary goods things are a bit different especially that the supply has a direct effect on demand so everybody who is a bit into luxury goods I personally not but I know a person who is and if you ask persons who are into luxury goods they really understand this point that supply is crucial for demand so if a woman is choosing a handbag I think 90% of the decision is determined by the fact whether other women have those handbags and the same goes a bit with monetary goods as well the supply is directly responsible for your demand if you if you have a monetary good which is easily inflatable you don't want to use it as monetary good and the second category of jacquemore is what he's calling darkness which sounds super scary but it's just a term that compromises or summarizes all the topics that we've heard before with sense of persistence privacy fungibility so one Bitcoin should be same as another Bitcoin and I want what I wanted to point out here is that theoretically in a model that is describing the demand for Bitcoin something like darkness could very well be a part of this function so you could imagine for example there's a bit technical term now that you include coin joins as a variable to quantify darkness so or any other variable that is a quantifiable variable that is correlated with what we coin here as darkness and but the stock the floor model is just focusing on this aspect of the hardness and we will see in the end there's probably reasonable at least as of now to focus on this this because the model is already explaining so much that there's only little room left for other variables but this could very well change so if you think about a situation where Bitcoin is or was illegal I mean it is illegal in many parts of the world but if it was illegal widespread I'm pretty sure that the darkness factor would pop up much more and would probably also be detectable by a statistical test and just one comment on statistics because I got this question a lot and what is statistics for actually what why are we doing all these statistical tests later on that we will go through briefly the point is statistics distinguishes between randomness and structure it says okay is something random or is there some structure behind it so for example with dr. flow model we are saying hardness has an impact and we have data and in the end we can we can't say this is just random there's too much structure behind the data and this is a very useful tool but of course this is and it was the other critique this is relating to the past so of course nobody is saying that this model will exactly be the same in five years or in four years but on the other hand the conditional probability that it will be valid tomorrow is much higher given the evidence we have so if you have complete randomness and you say anything can happen and on the other hand you have some substantial evidence there's a huge difference there of course everything can change in this world but the probability that it will change so fast and within seconds or only days is pretty it's pretty slim perhaps also very shortly the definition of stock to flow because this term here is very loosely defined the problem is all this analysis comes from the gold world and in the goal where we don't have a lot of data really data is already great but you don't have daily data you don't have intraday data so it is always defined as really production and you're usually just working with yearly data but with Bitcoin you have multiple options and we will get into this a bit later because it's important to understand certain points and you could hear for example say I take the last year I take days and I'm counting the last 365 days and I say this is the yearly production like 600 something thousand Bitcoin for example right now and once that half thing happens this is cut by half or I could say I just take one day of the new supply so once I think it's second May or whatever when the half thing will happen and then we take only one day with the new halfing figures and and I have 364 days of the higher supply and this is importantly we'll get into it at the very end when we look at the new results whether the market is forward-looking or not so this is a short recap if you say okay I'm partially convinced what this guy is saying here that we should perhaps test whether the hardness of Bitcoin is driving its value and I first present this chart here you would say okay I don't see anything here this is completely random I don't see any relationship but if you take the logarithm of both variables or the logarithm of market cap and a logarithm of the stock to flow ratio you see a clearer picture emerging there but if you were a trained statistician and a lot of people in a Bitcoin community apparently are trained statisticians they came up and said they came out after plan B's publication and said well you are you are using times here stator here which is non-stationary which means it's trending up over time there's some some trend here and when you pay attention in statistic classes you know that this is a very dangerous business that you're getting into the problem is that you might think you have a relationship you found a relationship when in fact you didn't find any relationship between two variables they just randomly move over time in a pattern which seems that they are have a relationship because they have some similarities there well in the end and the story could end there if we weren't able to detect whether there is a true relationship and this is the whole point about this conjugation relationship that you might have heard about and it's actually an advanced topic in a chronometrics but the basic intuition is pretty simple you can imagine that you have two cases and perhaps some people already know this example with the drunk and a dog but I think a lot of people haven't heard about it so we just repeat it here the first case is you have a drunk and a dog and they are not related so the dog is not the dog of the drunk the drunk is leaving a bar and the dog is also running around and they are just doing their thing there the guys trying to get home perhaps he ends up somewhere completely different not at home whatever the dog is chasing another dog or whatever and that's that's it the second example is that the dog is actually the dog of the drunk so perhaps for example the dog is on a leash and the drunk is still drunk and he still has some random patterns when walking through the city the dog is still a wild dog and is running around but the difference now is that they are connected through the leash so the dog can run off and the dog can't run off and then if you look at the distance between them in meters for example you will see that this number is not diverging you can be at maximum 10 meters or something like this sure it's worked but something like this and the point is now that the second example is the good case the second example is the conjugation case so the leash is something like the conjugation relationship there and the first example is the bad case the first example is a so-called spears regression case so they are not really related they are just running around and by chance they are just walking into the same direction and you can do conjugation tests in statistics but I will spare them or I will spare you with them but what the essence is what you're doing there is exactly the same with the drunk and the dog in the distance between them that you look at the actual price and the price implied by the model and if you see that the distance between them is not diverging but is like this pretty stationary goes up and down then you have a conjugation relationship the interesting thing thing here is that this relationship is actually getting more stable over time because you see the amplitude is getting less and less and others have also done Nick for example on Twitter and have done a lot of work to show that the coefficients in the model are converging so they are more and more stable which are all so if you're not a statistician and you have never done empirical work you just say okay they are converging whatever but I have never seen any model that stable that robust in my life so and many others haven't seen that either it's very difficult to find a conjugation relationship that is actually getting stronger usually it's the case that you have structure breaks so for example from my profession as an FX analyst I can tell you that there's a period in time where you have a stable relationship between interest rate differentials and exchange rates but suddenly right now I think since many years actually euro dollar is doing its thing it's a thing is actually going in the other direction as opposed to the model of interest rates differentials but with the stock to flow model it's the opposite it's getting more and more stable which is pretty impressive so the point is then that you might say okay the stock to flow variable if you're for more familiar also with the Bitcoin ecosystem and the technical foundations you would say well the stock to flow is not really a dog because it's not that random it's pretty predetermined and again Nick came up with this nice analogy to say that you still have to drunk this is the Bitcoin price we've heard many times today that it's pretty random and it is part at parts very random but there is the structure of the stock to flow road so you don't have a dog but you have a road where the path is pretty predetermined and the price can move freely and randomly within certain boundaries so for example can't jump over the bushes here but within the boundaries it can be pretty wild sometimes and with Bitcoin one has one really has to admit as well that those boundaries are pretty huge so this is in in log so it's a bit difficult to calculate now but this can mean that you really have swings of 30% and you're still within the co-integration relationship which is obviously a bit too much for many people to digest but that's how it is so and the last part will be even a bit more difficult so I tried really to explain as simple as possible these are our new results so there wasn't the highly debated question okay if the stock to flow model is correct you have the problem or you need to explain why is the market not front-running why it's not saying okay this guy is talking about stock to flow above 120 24 then I would just like sell everything and buy Bitcoin the point for me was first to say let's first look whether the or get a feeling and and do some testing whether the Bitcoin market is at all forward looking so we have another hypothesis now the Bitcoin market is forward looking so and if you don't use statistics you just say this and you don't know it but we can or we came up in this case with a test and we said okay the Bitcoin market is forward-looking this takes some time now this is difficult stuff the Bitcoin market is forward-looking so can we perhaps the easiest thing is always in statistics to disprove something this is the easiest approach so to find evidence against a hypothesis and what we did here is that we use the regular stock to flow model market cap and the stock to flow ratio and we use a definition that once the new flow variable should be applied so a new half incomes we are using the new annual flow data this is this important remark and then we are using what is called leads and lags of the stock to flow ratio so for example a lack of two means we take the stock to flow ratio but not of today but of two months ago and lead means for example 12 means that we don't take the current stock to flow ratio but a stock to flow ratio one year ahead so actually you would suppose if if the Bitcoin market is forward-looking that the leads so this this area here is where you get stronger results so what we did first was to test is the strong is the connexation relationship getting stronger when we use stock to flow ratios of the future and what we came up here is actually it's the opposite so if we take the lags like one year almost from the past we get the strongest co-integration relationship so you could imagine the rubber band gets very tight here because you have to read this that the more negative the test statistic here is the stronger the connexation relationship so exactly the opposite actually and the second criterion that we checked so we are running regressions with all of those leads and lags so one regression with 24 lead and then it goes 23 22 until minus 24 we go to the to the lags and what we get here is what we are doing here is that we use a second criteria how good is the model fit so how well is the model fitting to the data that we have at hand and surprisingly it's almost the same result here better fit again here lower values if you think okay I don't get what this guy is talking about just look at this chart here when we look at the Bitcoin half thing in 2016 where we had a jump in the model price because the half thing was happening or has happened actually and the actual price in this case even it took more than 12 months more than a year while then it overshot probably to compensate for being late and it took more in any case took more than one year to catch up with the model price so for me the point is we don't even have to talk about all these efficient market hypotheses and all this stuff if the Bitcoin market is not even forward-looking if it's backward-looking this is end of discussion for me at least and the interesting point we can discuss at dinner and whatever about it and is now why is this the case my personal view is that Bitcoin is mainly a retail phenomenon still we don't have big institutional investors who are hiring wants to do this stuff and then say okay let's front run a bit that invest substantially it's mainly a retail event so what is happening is that the market is communicating efficiently at the end but with some time like the new information the market is transmitting the the information hey Bitcoin is harder now than it has been before but it takes some time and this is also my summary here the Bitcoin market is big backward-looking as of now but this could change if we have more institutional investors coming in and especially in Germany a lot of regular changes has been made so that is now completely fine for a bank or for an asset manager to buy Bitcoin if it changes internal regulations at least because now it's a completely legal asset for for them so this is something which could probably change but what won't change at least is that we have so much complexity so much news around Bitcoin some many discussions around Bitcoin but as of now at least and probably for some times to come the main driver of the Bitcoin value is its hardness and the hardness is captured by stock-to-flow ratio and I don't think this will change anytime soon thank you very much and I think questions yes I think we're gonna do informal Q&A during the dinner it's always dangerous just before dinner a lengthy Q&A anyone falling asleep on your statistical content because it's really exciting to see and you get a lot of pushback for this or the statistical people in Bitcoin get a lot of pushback on this and the most critical people are the Bitcoiners to say like this is this is too too off here but that's exactly I mean other Bitcoiners they're tough on you validated now other Bitcoiners checked it everybody's checking it that's the same principle with the code it's the same principle with anything else that you will always have people if you're if you're if you're dealing with a hive like it is with Bitcoin you have a swarm you will always have some people who know special things and there are enough statisticians to test this and the Bitcoin network is doing this on its own you don't need any like third-party evaluation service to do this this is all done and very good