First, the Ual,WCOm thing was a joke. Second, you are right. We are talking about expectation. I think though that you do not quite understand what expectation means. Expectation is not what YOU guess is going to happen or what you want to happen. If I buy a stock, I have no expectation, only hope. This is because there is no guarantee that a stock will go higher from the point that I purchased it at. A coin, or other random item, dice, etc. Will eventually land on the number or side you hope it will land on. When I speak of expectation I do not mean for example, if I flip a coin I want it to land on heads. That is what I hope. You my friend, are confusing hope with expectation. I HOPE this clears it up for you. Chris
RE: Data on N100 Stocks: Oh goody, yummy yummy data! Thanx jperl. Now, lets look at the results of coin-toss, using the binomial distribution. You will find that the outcome of tossing a 50-50 coin 100 times looks nothing like the results of looking at N100 stocks over the last 100 days. The percentages of sequences with a particular number of "up-days" or "heads" falling within the same set of bins used by jperl is: Range N100 Fair Coin 61-65 0.00% 1.67% 56-60 0.98% 11.80% 51-55 5.88% 32.46% 46-50 30.30% 35.57% 41-45 41.18% 15.57% 36-40 20.59% 2.67% 31-35 0.98% 0.17% 26-30 0.00% 0.00% As you can see, the N100 results are heavily skewed toward too few "up-days." Whereas a sequence of 100 tosses of a fair coin would yield 56-60 heads in 11.8% of 100-toss sequences, the stocks on the N100 only saw 56-60 up-days in 0.98% of 100-day sequences. Likewise the N100 had too many sequences with very low numbers of heads. This bias in the distribution of reflects the bearish trend of the last 100 days and shows that if the N100 stocks are a coin-toss, they are certainly NOT a 50-50 coin toss. Although it is a bit hard to judge the average number of up-days from this binned data, my guess is that it is around 44. The chance that 102 sequences of 100-tosses each of a fair coin would have an average of 44 heads per 100-toss sequence is extremely extremely low. Although the chance of a single 100-toss sequence have 44 or fewer heads is not low (having a probability of 13.6%), the chance that 102 sequences would have an average of 44 heads/sequence is extremely low (this is more than a 12-sigma event, having a probability on the order of 4xE-34). So, we can reject the hypothesis that the daily direction of movement of N100 stocks follows the same distribution as a 50-50 coin toss. (NOTE: this analysis only shows that the N100 is not a 50-50 fair coin, it could still be random with a more than 50% of a downday) <b>RE: profitseer</b> "Besides, the idea isn't to bet on the coin toss, the idea is to bet on how people will bet on a coin toss." -- profitseer Beautifully said! This is why the markets are NOT random and why technical analysis works and why traders can profit. The price movement in the markets is determined by the actions of the buyers and sellers. And those buyers and sellers are not coin-tossers (OK, maybe some traders are coin-tossers!?), but human beings. As inandlong pointed out, human behavior follows predictable modes of fear, greed, and confusion that generate the bear, bull, and choppy price action respectively (as goldenarm mentioned). The key is that price movements are NOT independent of each other the way coin tosses are. Although a coin is not aware that it flipped "tails" many times in row, the participants in the market are EXTREMELY and often painfully/joyfully aware of the number of downticks in a row and behave accordingly. Prior price action does influence the buyers and sellers, which does influence future price action. With price movements being dependent on each other, the goal of the trader is to learn and exploit these dependencies. The challenge with trading is that the interdependencies between past and future price action are very complex (and partly driven by ostensibly random factors like news). Skilled traders learn to recognize the probable sentiment of the market and act accordingly. So price movements are NOT independent of each other. Of course, once you lose the assumption of independence, you lose the central limit theorem, lose convergence to a normal distribution, lose all the nice statistic methods for predicting drawdown and optimizing bet size, etc. Much of the statistical analysis of the markets and trading is a real house of cards -- too much of it is based on assumptions which are too easily refuted. Wishing everyone good & careful trading -Traden4Alpha
Traden4Alpha- In your last post you showed some data for 100 fair coin toss and compared it with that for my stock data on the number of updays over a 100 day period. Perhaps you would tell us how you arrived at your percentages. They don't look like a binomial distribution to me. Bkuerbs pointed out that 75% of the data should fall within 2 standard deviations(which for the 100 coin toss is +-10). Your data for the bins 41 to 60 add up to 95.4% which looks more like a gaussian(i.e normal) distibution. Is Bkuerbs incorrect about this? As far as the data being skewed to the low side, yes you are correct. It looks as if the mean is less than 50. There could very well be a non-random component on top of the random behavior. The question that I would like to raise then at this point is, is there a way to ferret out this non-random component and act on it in a trading sense. We need some quantitative ideas here from anyone who is thinking about this.
Dude, you are an example of a trader I wish there were more of... Buying/selling and appearing to have no clue what they are doing, just "hoping"... I don't mean to be disrespectful, but you started it... There is NO guarantee that the stock will go up, you are right. But it is EXPECTED to go up, otherwise it's worthless. It's a simple math proof, but you don't seem to be mathematically inclined. Your coin flips examples are not entirely accurate the way they are given. Let's assume you are buying an average-systematic risk level stock. Beta of one. You'd expect it to provide a positive return equal to expected market return. Think of it as having a rigged coin that tends to land on one side a bit more. Thus, my friend, you ARE EXPECTING it to land on that particular side on average. That's if you are rational (which might not be the case). From what you stated, it seems that trading is pretty much like gambling for you. You place your unfounded bets and "hope". If that hope is not based on some sound research, then keep hoping man. As long as there are folks like you, there's hope for others to make money
has anyone ever tried to sell ALL bounces off the top bollinger band on a 3 min ES futures chart for 1 point and buy ALL bounces off the bottom bollinger band. I have never done so but at night AFTER THE FACT it looks profitable. this is probably as close as I could ever bring myself to random trading. I am sure one of you super tradestaion wiz people could probably get an answer to this onw. I am truly interested in what the answer is. I trade 2-5 contracts so even profitabily of as little as $8.00 per trade( before commissions) is workable in my opinion. thanks
Vlad, Just because I hope a stock will go this way or that, does not mean I do not do my DD. The only point that I was trying to make is that no matter how much fundamental research I do (consider please research that comes out of brokerage firms), does not mean that the the stock I bought is going my way. My point is that I try not to be an egomaniac and because I think something it will not necessarily happen. I also don't think you understand, the stocks I buy I expect to go up, and the stocks I sell, I expect to go down. But if that not what the market wants (the MARKET's expectation), I don't get paid is all I'm saying. Chris
@jperl Sorry, it is not that easy. When you have some data and want to know whether they belong to a certain class of distributions you have to use statistical tables or better feed your raw data into a statistics program and try fitting. I took your binned data and entered them into such a program, see the result in the pdf-file below. It would be better to use the raw data and leave the binning to the program or do it manually after some analysis of the data, I doubt that in this case it will change the result very much. Regards Bernd Kuerbs
The attachment to my previous post can not be downloaded (wrong name?), so I uploaded a copy here: http://home.t-online.de/home/Bernd.Kuerbs/Dokumente/Nasd102.pdf Regards Bernd Kuerbs