PDA

View Full Version : Stock picking and probability

x-ray vision
08-14-2004, 10:56 PM
In this (http://www.investopedia.com/articles/trading/04/051904.asp) article, there are a couple of statements that seem to go against mathematical probability. Am I right or is there something I'm not getting?

According to this veteran program-trading expert and author, "[y]ou can expect 6 or 7 out of 10 trend trades to be losses, some small some a little larger."

Since a stock can only go up or down, shouldn't you have a 50% success rate if your system was flipping a coin to decide whether to long a stock or short a stock?

Exapno Mapcase
08-14-2004, 11:17 PM
It's not absolutely clear from the article, but it reads to me as if Kaufman is saying that in the real world the distribution of trades differs from random chance.

ultrafilter
08-15-2004, 12:07 AM
You're assuming that each stock is equally likely to go up and down. Do you have a basis for that assumption?

x-ray vision
08-15-2004, 09:35 AM
I realize that each stock isn't as likely to go one way as another but should that matter?

Suppose the Bad News Bears play against the New York Yankees everyday. Even though the Yankees are more likely to win each day, if I choose what team I'm going to bet on by flipping a coin, I'll still end up being correct about 50% of the time over the long haul.

Chairman Pow
08-15-2004, 12:10 PM
I realize that each stock isn't as likely to go one way as another but should that matter?

Suppose the Bad News Bears play against the New York Yankees everyday. Even though the Yankees are more likely to win each day, if I choose what team I'm going to bet on by flipping a coin, I'll still end up being correct about 50% of the time over the long haul.

If the Yanks beat the BNBs 50% of the time, you would be correct. If you have 50% each way coin toss but the Yanks beat the BNBs 75% of the time, you are less likely to hit the BNBs on the flips that they do win the game on.

Related thoughts that probably are irrelevant in this context:

Random walk theory is endlessly debated, but there are a lot of people who make a lot of money trading trends (namely large funds) that disprove (or at least downplay the importance of) the random walk theory.

As for the 50% up or down, this does not even apply to randomly selected stocks. Take a look at the advancing/declining issues indicator which gauges the amount of stocks that end the day with higher prices vs. those that end with lower. There is a definite bias to the upside.

You also have to realize that a trend in the relevant indexes (S&P/Wilshire/DJIA and sector indexes) will also likely effect a pull on stock prices regardless of whether or not a particular issue is worthy (refer to tech bubble).

BTW, Kaufman is a highly respected author and his works are interesting. I'd recommend reading them if you're interested.

x-ray vision
08-15-2004, 12:16 PM
Originally posted by Chairman Pow

If the Yanks beat the BNBs 50% of the time, you would be correct.

Aren't I correct no matter what percentage of the time the Yankees beat the BNBs?
If the Yankees beat the BNBs 100% of the time, I will still bet on the winning team (the Yankees) about 50% of the time as determined by a coin toss.

Chairman Pow
08-15-2004, 12:47 PM
Aren't I correct no matter what percentage of the time the Yankees beat the BNBs?
If the Yankees beat the BNBs 100% of the time, I will still bet on the winning team (the Yankees) about 50% of the time as determined by a coin toss.

I turned the computer off, walked away and realized that I had no idea what I was talking about. I think I was conflating two thoughts and ended up choosing the incorrect part of both. Boy, do I look like a doofus. :smack:

You are correct. What I should have said was that you are correct, on any given day, you have a 50% chance of being right about who will win the game. Over a long enough sampling period, your success rate at choosing the winner will be 50%.

These are not betting odds, especially for the stock market. I think it is an incorrect assumption to assume that on any given day, the price of the security will have a 50% of being higher or lower. First of all, this discounts the possibility that the price will be equal to the previous day's price (although that is pretty rare). Second, you are not taking into account the magnitude of the changes in price. For some reason, every argument I've seen along these lines assumes that the advance/decline will be equal to each other. Assuming this is true, if you spend enough time in the market, you will eventually lose due to commissions paid that are not being recouped in the gains.

What Kaufman is essentially saying is that there is a trend that will take hold in the given stock and, while in the trend, each day is likely to follow in the direction of the previous day. There will be corrections, of course and the trend will eventually end, but in the mean time, one stands to make more money by following the trend.* If you have evidence that the price of tomorrow's market will be higher, why short?

What you are not considering is the idea that when the price of a stock rises, someone will take notice and say, "aha! According to my calculations, this will continue. I should buy more shares." This extra participant will create extra demand. Someone else will see this and say, "me too" etc., etc. With all this buying going on, what makes one think that the price will drop suddenly? Certianly these people will decide to take profits at some point, but remember that for every seller there is a buyer and depending on when the seller sells, there's a good chance that the price will increase.

Also, the old saying, "cut your losses and let your profits run" applies here as well. If you take a trade and it immediately goes against you, the common sense thing to do is to get out. If you continue to make money day in and day out, why would you exit the position? Your trading program could have built-in profit targets, but you could also establish rules to determine at what point the market has moved against you and close out your position then.

* Of course, determining when a trend has started and when it ends are another matter entirely).

Thudlow Boink
08-15-2004, 12:49 PM
Suppose the Bad News Bears play against the New York Yankees everyday. Even though the Yankees are more likely to win each day, if I choose what team I'm going to bet on by flipping a coin, I'll still end up being correct about 50% of the time over the long haul.Sounds right to me, assuming the coin flip is equally likely to come up heads as tails. Whichever team wins, there's a 50% chance that your coin will "pick" that team.

I don't know enough about stocks to answer the question of the OP.

x-ray vision
08-15-2004, 01:07 PM
Originally posted by Chairman Pow

Assuming this is true, if you spend enough time in the market, you will eventually lose due to commissions paid that are not being recouped in the gains.

Yeah, that makes sense, but I don't understand why Kaufman thinks that "trend-following systems are some of the best trading systems around" and that you can expect 6 or 7 out of 10 trend trades to be losses.

Chairman Pow
08-15-2004, 04:58 PM
Yeah, that makes sense, but I don't understand why Kaufman thinks that "trend-following systems are some of the best trading systems around" and that you can expect 6 or 7 out of 10 trend trades to be losses.

Therein lies the problem of market timing. If you select a good spot to get in (i.e. the trend is underway and there isn't one final retracement before the move), you'll make, say \$1000 on your trade (hypothetical number). If you select an incorrect spot (either there's a large retracement before the move begins or you select the beginning of the opposite trend), you will lose. If you have designed your trading system properly, you will have established an amount that tells you when the market is definitely not moving in your favor and you get out, say \$100.

If you are averaging 40% winners (according to his example), you have \$4,000 profits and \$600 losses. You still come out \$3600 ahead. Now, this will still work if you have \$500 and \$100, \$400 and \$100, even \$300 and \$100.

The idea is you are limiting your losses and choosing trades where your risk:reward ratio is favorable so your winnings far eclipse your earnings. You basically have to decide what ratio is acceptable for your needs. 10:1 is unrealistic, and 3:1 is fine for most people. 2:1 and you take you really have to have confidence in your ability to not choose 60% losers.

That said, there are people who make sucessful trades 10% of the time and still end up millionaires year in and year out.

When they discuss how the number of trades adversly affects your earnings what they're basically saying is that if you are trading XYZ a hundred times a day, you're limiting yourself to scalping (say you make 25 trades an hour, how far can the market move in 150 seconds?), or do you trade once per week? Month? Year? The assumption is that the longer you allow the trade to go on, there is a greater possibility that the price will move a greater amount.

If you'd like to see the concept in action, check out the concept of Volatility in options and how smaller time periods can be extrapolated.

js_africanus
08-16-2004, 10:49 AM
Random walk theory is endlessly debated, but there are a lot of people who make a lot of money trading trends (namely large funds) that disprove (or at least downplay the importance of) the random walk theory.
:dubious: People keep debating whether there was evolution, too, but I'm not so sure that means that much. You may enjoy A Random Walk Down Wall Street for a discussion of the debate about market efficiency. Also Fooled By Randomness is another noteworthy book on the beating the market idea.

According to this veteran program-trading expert and author, "[y]ou can expect 6 or 7 out of 10 trend trades to be losses, some small some a little larger."
Okay, what am I missing? If 65% of one's trend trades are losses, then one is going to lose money.

Suppose it were otherwise. How could one make money when 65% of bets are losses?
One could know which bets were better and bet larger on them. If that were the case, then one would stop making predictably bad bets. This isn't a case of having to keep betting to stay at the table until the card count improves.
The bets are such that if they win, they win big, and if they lose, they lose small. How would one accomplish this by merely betting on whether a stock price rises or falls? If stocks price changes have a log-normal distribution, then for a given stock, you can expect a second rise in price to be larger than a second drop in price. For example: rising by 10% twice in a row on a stock purchased for \$100 will yield \$10, then \$11; however a twice drop of the same amount will lose -\$10, then -\$9. But that's not how they're allegedly making money.

There doesn't seem to be any sophisticated financial instrument on the table. How is it that winning wins big and losing loses small in a predictable manner? This is how:
The trick here is to analyze the system for its effective time spent in the market. For example, if a system is in the market only 50 percent of the time, you can put on twice as many contracts each time the system enters into a trade, as compared to what you would have in a buy-and-hold situation over the entire period to achieve the same amount of contract hours spent in the market. Looked at this way, in the 50 percent case the profits made per contract traded should be at least half that of the constant-contract, buy-and-hold strategy. Proper money management also allows you to increase the number of contracts traded as your equity grows, whereas a buy-and-hold strategy doesn't give you the same flexibility.
In other words, you have to know when your bets are going to be good and then get into the market heavily at those times. But that just brings us to bullet #1: Why get into the market when one know's the odds are poor? Not only is this telling us that technical analysis informs the best times to be in the market, but also the best sub-periods of those times. If one bet evenly over those periods, the 65% loss rate would net an overall loss, so the trader needs to know, when he is in the market, what times to bet even more heavily. But if he can tell that, why be in the market when he will be doing predictalbly bad?!

Something asserted in the article bears repeating: "Proper money management also allows you to increase the number of contracts traded as your equity grows...." In other words, they are selling the very same betting schemes that casino patrons have been falling for over the last who-knows-how-many years. Guess what: No betting scheme can improve a punter's expectation (www.amazon.com/exec/obidos/ASIN/012240761X/qid=1092670853/sr=ka-1/ref=pd_ka_1/103-8643344-7252600). The thrust of this article can be put in easier terms: If you're playing black jack, then you should be more concerned with your betting scheme than with your strategy for play.

scm1001
08-16-2004, 11:04 AM
65% of trades could still be losses in any system, as there is a small cost to each trade. E.g if each trade costs you a dollar, and you make profits (exclusive of the brokerage) of \$2, 1\$, £0, -\$1, -\$2, then gross you are even, but net you have lost on 60% of the trades.

x-ray vision
08-16-2004, 03:09 PM
You guys are getting off topic. What I want to know is whether or not the quote by Kaufman in my OP is some how correct. As I stated, it seems that you should make an incorrect choice 50% of the time, not 6 or 7 out of 10 times.

Chairman Pow
08-16-2004, 05:16 PM
He's not making his choices randomly, that's how.

x-ray vision
08-16-2004, 05:30 PM
BTW, Kaufman is a highly respected author and his works are interesting. I'd recommend reading them if you're interested.

So, you'd recommend I read his works when he recommends following systems that are less efficient than flipping a coin?

Chairman Pow
08-16-2004, 10:48 PM
You are missing the point entirely. Go back and read my previous post (I suppose it's actually two previous) wherein I state that the idea is that you make several trades and your wins are larger than your losses. If you do that, your percentage ("efficiency" as you call it) will be worse, but you will make more money than you would by randomly selecting stocks.

x-ray vision
08-16-2004, 10:56 PM
I understand that point. I understand that you can pick incorrectly more than you do correctly and still make a profit by limiting your losses and maximizing profits on winning trades.

My point is that you can randomly pick stocks and pick a winner 50% of the time and limit your losses on the 50% that are losers and maximize profits on the 50% that are winners. Wouldn't this system be better than one where you pick incorrectly 6 out of 7 times?

Chairman Pow
08-18-2004, 05:14 PM
Somehow, this didn't show up on my new posts list.

You will have to describe your random approach for picking stocks before we can go any further. Are you getting in every day? Are you always in the market, or are there days in between the trades? As soon as a trade closes, are you going to re-enter the market again the next day? What stocks will you look for? How many shares will you buy? Are you going to add on/subtract from your position during its lifetime? If you don't have profit/loss targets, when will you take profits/losses?

I will argue that you cannot beat a well thought-out trend following system because your random entries will not move quickly enough in the direction you would like to make a profit.

The easiest way to test to see if you are right is to go to www.barchart.com, select a stock and check the chart (I think the default is something like 50 days). Take your coin, say, "heads = long 100 shares, tails = short 100 shares." Use all of the risk:reward ratios and profit targets in my post and see how far you get. To keep it really simple, pretend that you have an unlimited account, no commissions and use only the opening prices for entry and closing prices for exit.

To test this against your trend following system...well, you'll have to figure one out (and make it a decent one with a track record of returning positive returns while limiting losses) and compare them over a trial period. I think that you will kill yourself if you are always entering and exiting positions and that a couple of trials even over 50 days (even less if you feel like it), you'll find that this is, if not inferior to a trend following system (again, unless you make your own, you really won't find anyone whose willing to sell you theirs. Maybe in Kaufman's book though), then it is a money losing proposition.

Chairman Pow
08-18-2004, 05:19 PM
I was hoping that they would allow you to see a chart over some absolute ranges, but they don't. Theoretically you can sign up for their service and download the history.

If you do perform this experiment, I think the best format would be:

Date: Open Close Entry @ Exit @ P/L (this trade) P/L Cumulative