[MUSIC] We're now following up on the previous poll. Are markets more or less predictable when they are efficient? Well, it may sound a little bit counter-intuitive, but if markets are efficient, they are actually going to be less predictable and fluctuate more in a random fashion. So let's dive in to more detailed explanation of these concepts and how they relate to trading strategies. Olivier could you remind us of the first definition of market deficiency? >> Yes, so if you remember I talked about three forms of market deficiencies. So weak form, semi strong, and strong, so let me just remind you what is a weak from. So the weak form is related to past prices. So if I just Look at past prices, this information should be fully reflected in the prices. >> Let's look at an example of past prices that we typically depict in a graph. As you can see on the slide, I'm displaying here in blue the trajectory of a stock price. In red I'm displaying the average computed over the 20 previous days. So you can see that the red line is much smoother because it reflects only long term variation of the price. And the green line is a moving average, again computed on a longer time interval, 40 days. As you can see, all these lines crosses each other and we might be interested in looking at the structure of these graphs. So, Olivier do you see any pattern in this? [CROSSTALK] >> Yeah, so it looks interesting. So if I remember well how we can use it if you want to do trading is basically when the short-term moving average is getting downwards the long-term average I should sell. And vice versa in the short term is crossing upwards I should buy. But this looks really strange to me because I have a paper showing that in fact if you look at 8,000 different selling strategies, it's impossible to make money. Because as soon as you introduce some transaction costs, everything disappears. So I'm a little bit surprised by what I see here. >> Well actually Olivier, these prices are not real prices. I generated randomly this trajectory so any form of signal trading pattern that we might want to infer by looking at this graph is really the result of our imagination. There is absolutely no predictability here. So you're perfectly right to doubt the validity of a strategy related to previous patterns in a price trajectory. So the weak form of market deficiency actually implies that it is not possible to identify structure that allow us by just looking at previous prices to predict future prices. But more generally even if we do find structure in prices it does not necessarily imply that we will find predictability. What we mean by predictability is can we find a link between yesterday's price and today's price? And predictability and structure are two very different things. Let's illustrate this by looking at successive returns points. So we are going to take a stock, compute it's return let's say today and relate this to the return tomorrow. And we're going to do this on a very large number of points in time. We're going to display all of this in a graph so we will have the return today on the x axis and the return yesterday on the y axis. This is how it looks. >> Wow, this is very surprising. So this really nice, so I didn't know that you were really a fan of modern art. >> I didn't even draw this. This is actual data, this is what a plot of these particular time series would look like. It looks very regular. Actually this shape is called a compass rose. It indicates the direction of north, south, east, west with thicker lines And it seems that if prices actually follow that particular behavior they must be predictable but actually they're not. They fall into this particular pattern simply because of a market micro structure effect. The explanation is a little bit complicated, but its actually only depends on the tick size. The tick size is the minimum price movement that you can observe in a quoted security. So for example, a stock can only move by a cent, it's minimum movement can only be a cent. There is no smaller possible variation. The tick size depend on the particular market it's orchestrated on. So here, this structure is due to this particular tick size, and is related to the fact that the price level and the price change are small multiples of that tick size. This particular structure seems to indicate predictability but in reality it does not. You can do absolutely nothing with that except contemplate it it's a nice picture. More generally when you want to relate past return to current return there are very specific ways of testing for that. And as a statistician, Olivier, how would you address such a question? >> So thank you very much for the question. In fact, there is really a whole zoo of tests out there. And in fact there is no real consensus among statistician on whether we have a In the predictability or no predictability when you look at prices. So there are very simple tests and I will just mention two of them. One which is called the Unit root test and what the test is what is called the handle mode hypothesis. So, what is the handle mode? It's very simple to understand. So, if I want to predict the price for tomorrow, my best predictor is in fact the price today. So it's like a drunk man. So if you want to find a drunk man, the best way to do it is just to look at where he was just one second ago, because he can just go either right or either left. Okay so this is the handle mode hypothesis. There is another test which is out there which is called Variance ratio test, and it explore the fact that for Handle mode in fact the volumes of the returns scale linearly. So what does it mean? It mean that if I look at the variance of a, Two week returns. It should be twice the values of a one week return. And so you can indeed develop tests for that. So, there are other more sophisticated tests. But what I would insist on is, in fact it's still a very, very vivid debate among economists, among statisticians and definitely we don't have a definite answer. >> So before we move to the second form of the efficiency let's just summarize this first point. If markets are weak form efficient, you cannot create profitable trading strategies only based on the behavior of past prices. But there could be more information that you could introduce to construct your trading strategy so let's go back to the second definition. [MUSIC]