The first step is calculating this number for your universe of companies. I've told you before, universe of companies on which you can try different constitutional context as well. Get these number from all the companies. The second step is to classify them into deciles. Higher or lower, what kind of score is better? Yes, the higher the better. What does a high score mean? Okay, I think we should take a step back and just spend some time on interpreting this before getting an implementation. A higher score on SUE or a higher SUE actually means that the unexpected earnings per unit of variation or standard deviation is higher. SUE is basically unexpected earnings per unit of risk if you will. Standard deviation is risk, so a company which is scoring higher on SUE when compared to another company. Basically earns higher unexpected earnings per unit of risk when compared to the other company. So the rank ordering should be based on the score. The higher the score, better it is. So you can first take the entire universe of stocks, classify them into the size. And then as we did in the Petrovsky case, we can make, we can try various combinations. The safest to do is to go long on, go long on, think about it, go long on the top decile. Yes, the ones with highest score by which are ranked the highest. And go short on, the bottom decile in terms of SUE. And of course you can choose top two deciles and no bottom two deciles and so on and so forth, and experiment them. This is one way of calculating unexpected earnings. There is another way. Now you'd have guessed that this whole business o taking an average of last four quarters is some kind of an approximation, right? This is not the real unexpected number. So if you have access to data, if you have access to data sources, the best way to get unexpected number is get hold of analyst expectation. Before publication of actual results, analyst expectations are made publically available. So you take these numbers and take the average of analyst expectation. Once you take the average of analyst expectation, use that as an expectation number. And then take the actual number, subtract them, and divide this by variability in analyst expectations. That is an another way of calculating this SUE, which ever way you will do. So the classifications scheme will remain the same. So either you take last four years average earnings get a expected number divided by variability in the last four years earnings, or you take the analyst estimate number. Take the average of it, take the actual number, subtract this analyst estimate number from the actual. And divided by standard deviation of [INAUDIBLE] whichever way you will do, you will arrive at a service score for each company. And the next step is as I've said classifying them into deciles, and deciding on your training strategy. Some more conceptual issues about the strategy. If you remember I told you right at the introduction stage itself that moment you see your strategy working, you should question yourself as to, why does this strategy work? Now if it a fact that if a company announces positive earnings, or an unexpectedly positive earning and the stock is going to go up. Why don't all investors just jump into it and drive the stock price up immediately? Similarly, if on a bad news the stock keeps going down, why don't the investors just jump into the bandwagon and short the stock immediately? They do, a lot of times but then there are some restrictions. And there are some behavioral issues why this cannot be done. First, on the short side. Shorting is not easy in many markets. In many markets, shorting is explicitly forbidden. In many markets, shorting is costly. And in many cases, investors who invest into these institutional funds prevent fund managers from shorting the stocks. In many countries, mutual funds are not allowed to short. Insurance companies are not allowed to short. What about on the long side? Behavioral economists have shown that people suffer from something known as disposition effect. Now what is this disposition effect? Suppose you buy a stock, I'll ask you a simple question. You bought 2 stocks, 1 at 100, both at 100, let's say. One is now gone to 120, and the other one is 95. You have no information about fundamentals or you have equal information about both the fundamentals. Fundamentals are both the companies are the same or assume we all knew information. Now if you're forced to sell one of them, which one is going to be your instinctive choice? Instantaneous choice, immediately, what I want to choose? Most people sell off the 121 and hold on to the 95 one, why? Because I've already made profits. But then, if you think a bit further nothing prevents a stock which is moved from 100 to 120 from going further, from 120 to 200. In most of the case, it actually moves further. Using the same logic, nothing prevents a company which has fallen from 100 to 95 from falling further. But still people who have this position bias tend to sell the winners and hold on to their losers for long. Now if there are a lot of investors with this position bias. And they're already holding this company which has announced a positive unexpected earning, then what is likely to happen? The moment the stock price goes up, these guys come and dump the stocks. Because why? Just because they are making money, that's all. They don't evaluate the fundamentals. So if large search investors are present, then this process of adjusting to new equilibrium price based on the revealed information is likely to take time. And that is one of the reasons why you make money. That is what I am telling you, there are a lot of other explanations, transaction costs. We do not have time to get into that. I strongly encourage you to read papers on this. And also people have shown that stocks which are widely followed by analysts. This time taken for incorporation of the price is likely to be lower than stocks which are not so widely followed. This is the case in Petrovsky as well. So there is reason why this strategy works. And by the same logic, the strategy's likely to work even better in emerging economies, because the frictions are a little bit more. So it is not that investors are stupid, and that's why these strategies work. It's because there are frictions, institutional, behavioral, or it could be something that drives this. The frictions would be created by investors having short term horizon. And hence pressurizing fund managers to act with a short term horizon that can lead to a lot of complication. So it could be any reason. Therefore, you should spend some time to pause and think moment to find result. Jumping straight into trading can be hazardous. Only when you find some convincing reason or when you read a paper, if you think the reason that is given there is convincing then only you should proceed not other wise. Let's come back, now coming to SUE farther. We talked about these two ways of calculating SUE, the analyst way is difficult. The calculation is not difficult. It's actually easy, easier compared to the other one. But then you need to have that information. In markets where it is publicly available, I strongly encourage you do the second way. Where you pick up this analyst expected number, get the variation. And you calculate SUE by actual number minus the analyst expectation number divided by the variability which is given by the standard deviation. That is what I encourage you to do. Otherwise, we can take this proxy of last four years average and standard deviation. Once you have the number, another question is when to trade, how to trade and how long to trade? Now I already told you how to trade, right? So what you have to do is you have to do this decile business. Classify them into deciles, and buy the ones which are at the top and sell the ones which are at the bottom. That is how to trade. The question is when to trade? If we remember while discussing Petrovsky, I do want attention to the fact that in real life you will know this information only after this information is public. So you have to be very, very careful when you do your backtesting. Suppose the year end is December, I told her the same thing before. If the year end is December you can't assume you're training will start on January 1. There is going to be some lag between the ending of the year or ending of the quarter and actual announcement of results. So your at a confident gap and your back testing should start from say next day or next week from the day information is publicly revealed and your actual trading can only start after that day. You will not know what the EPS is unless the EPS is announced. It's as simple as that, right? Now next thing is what should be the holding period? Now in Petrovsky, it could be last as long as one year. But in this case, since information is public, see in Petrovsky there is some guess work going on, right? These are not precise signals. These are some kind of proxies for unobserved fundamentals, right? But here it's not the case, you know the exact earnings. So you don't have an infinite amount of time for the market to realize this and incorporate the information. Time of level is short. People have shown that most of these earnings, most of the returns of this strategy accrue in the first three, four weeks. 60 days is the maximum time for which this keeps working. So this is not as long as Petrovsky, this is likely shorter. But the returns that they have shown is phenomenal. People have shown anywhere between 7 to 15% in different markets. But the key is to start after the results are announced and hold on for say 60 days or so. So that's about prize earnings announcement drift. So this is one more thing you should understand. Petrovsky's some kind of a mean diversion kind of a formula where you look at stocks which are high book to market stocks. This is some kind of a momentum formula where you're looking at stocks which have done well. And go long on them and stocks which have not done well, you go short on them. So that's about price earnings announcement drift. So we'll talk about further strategies after this.
