I am trying to develop a system and need some statistics
So I was thinking about developing a stock trading system based on statistics but I am missing some key ingredients.
I am trying to find out, if such statistics exist (maybe as an overall proportion for all stocks historically)... say If a stock goes up one day, what is the probability that it goes up the next day? Similarly for the stock going down, what's the probability that it goes down the next day? Is their statistics like that available in a broad scale?
thanks
Not sure, but you can do it yourself with a few IF statements in excel.
I made one last month; really interesting I must say.
When first getting into stocks, I remember developing this exact system. I was 13 though. Sad Story is that you're presumably far older than 13 and honestly think that a system similar to this has a chance of being profitable.
If you're trying to learn about stocks, I suggest you do this little experiment, but incorporate options with varying strikes/expirations to provide some protection against particularly volatile days. Backtest and attempt to optimize the system. I could see some educational value in that. There is no worth in your original proposal though. It's just a lot of simple, boring data collection for a system that has zero chance of being profitable in the long run
You have and will never get laid. I really do hope you make millions, because that's the only way you will be able to afford the hookers and therapy.
you're in the wrong forum
this place is more about S&T and prop trading.
For developing a system you should go to another forum that's specific to trading. Google "stock trading forum" or anything like that. the only one I've been to is elitetrader, which has an okay forum.
Shut the fuck up please. you have 10 posts and are telling him to not post here. I am happy to see some variety in the posts here besides typical
"OMG I have a 3.8949 from a target "think HWYP" majoring in finance. What do you think are my chances of getting a SA at GS on their prop desk."
"Oh the ladies ever tell you that you look like a fucking optical illusion" - Frank Slaughtery 25th Hour.I did not tell him to not post here. I told him he was in the wrong forum.
Are you saying WSO (consisting mostly of members who likely do not trade securities) would give better advice to what he is asking than what he would receive from a stock trading forum (consisting of actively trading members, and those who have been trading for much longer than many of the users on WSO)?
I hate the typical questions like the example you gave as well. I'm only helping the guy with the best response.
why do you think you know what my proposed system is? This is only a small part of it. Do you seriously think a system would be something like "oh..since the stock went up today, I should go in tomorrow?"
So here's a rough idea.
Assuming I know that if a stock goes up one day, the proportion of times it goes up the next day is 70%.
Say if I find 2 stocks that go in opposite ways historically with a a correlation coefficient close to -1 (obviously unrealistic, but a high negative correlation, think oil companies and airlines or something like that)
Say the regression equation is 2airline = 1oil company
Then I would buy call options of oil company and twice the amount of call options of airline with a strike price at the current price (to eliminate bias) (twice the amount to fit the regression model)
The first day of trading if oil company goes up and airline goes down.. Statistically, you should break even that day minus the time value of the option of course (due to the high negative correlation). I'll dump the airline at the end of the day. Assuming the 70% statistic (completely made up). There would be a 70% chance that my call option for oil company would go up again, and you also limit the chance of losing going in the first day with a directional bet. There, you have a statistical edge. Obviously it's not sure win and I dont even know if it is realistic, but mathematically it works.
obviously the hedge should involve more components to make it safer, stepwise regression would be a good tool.
If you are going to spew bullshit, it can generally be done in a couple of words
you just did exactly that, maybe a little more than a couple words but close enough
at the end of the day, your concept is still based on the fact that if a stock goes up today, whether or not it will go up or down tomorrow.
you didn't add any real material with your elaboration apart from some terms from stats 101. if there was anything to gain by this "strategy," don't you think wealthier and smarter people would have long exploited this little opportunity to the point that it is not longer exploitable?
also, to answer your question about the stats, there probably is, but i have no idea where to find it. a friend of mine just wrote a program to do this and mined all the info off of yahoo finance.
yes. but you have to start somewhere. Don't you think someone told David Shaw the same thing when he did his research? (not saying I am the next David Shaw but just an example)
It is not exactly. The thing is, if you jump in with a directional bet, you expose yourself to more risk. What if you initial directional bet is wrong? Then you are taking all the loss. My idea is, you are statistically going to break even the first day (you might gain, you might lose, but the average based on statistics says that you should break even minus time value which would be minimum because it's only one trading day), but the second day you have a statistical edge of gaining (obviously, you might gain and you might lose, but you have the statistical edge).
the thing is, just think of it as a markov chain. Today you are in a "going up" state, what is the probability that it remains in the same state tomorrow? Assuming the stock cant stay exactly the same (which you break even and it doesnt really matter)
There are two states of the markov chain: State 0.9 = going down (you lose 10%) State 1.1 = going up (you gain 10%)
(assuming the probability that stock goes up given that the previous day was up take is 70%)
going from state 1.1 to state 1.1 would be 70% going from state 1.1 to state 0.9 would be 30% The expected return would be 1.1*0.7 + 0.9 * 0.3 = 1.04 > 1
Obviously this is from a perfect finite time markov chain which is not completely representative of the real world but it makes sense mathematically, but I am also just starting. And I know it would be simply statistics based on previous data and have nothing to do with what's happening right now, but think of it as "as time approaches infinity, p-hat = p." if you do it continuously day after day after day.
I am not saying this is even a system at all, I am just starting. but I do want to find the statistics to start.
Start off with something simple.
Make an excel model with a few stocks in your universe. Link historical stock prices from Yahoo so you can get the last 40 or 50 days stock prices and use a 30 or 40 day moving average to indicate a buy/sell trigger. When stock price 30-day avg price, the cell is conditionally formatted to turn red and say SELL.
Pretty simple. Will it make you the next Renaissance Technologies? Probably not. But for learning purposes its a good place to start. See what that strategy would return and try to modify depending on whatever technical analysis you have a good grasp of.
As for you other idiots, like the douche bag who's been coming up with algo strategies since he was 13 years old... whats the correlation coefficient of the distance of your ejaculate shot and the type of star wars figurine you're jerking off to?
The only time i have ever heard of something anywhere near along these lines working was Gil Blake doing mutual fund timing back in the 1980s. He was interviewed in "New Market Wizards" by Jack Schwager. I admit I was extremely skeptical of his overall method however he had great results using it back in the 80s. Highly doubt his method still works.
In general system design and backtesting is a pain in the ass. I would say just about every idea you would tend to think of will fail to display good backtesting results. If it hasnt worked in the past it is highly unlikely to work in the future. Also just because it worked in the past does not mean it will continue to work in the future.
I have been trading a system for the past month or so with good results but the problems are 1) this trade is beginning to get too crowded which can cause lots of whipsaws from the added volatility.
2) I constantly feel like today may be when the system fails to work. 3) A lot of the trades are not easy to put on. They do not "look" like they are going to work out when they are put on. Of course from my experience I know that the trades that often look the worst turn out to be the best.
"Oh the ladies ever tell you that you look like a fucking optical illusion" - Frank Slaughtery 25th Hour.
I'm just brainstorming right now, perhaps you should also consider what the overall market was doing those days also. Isn't it possible that many stocks would rise and fall with the tide of general market conditions. Not sure how or if this should factor in, just a musing.
I think this could be a difficult system to develop with any effectiveness. I understand what you are saying with your Markov chain argument, I am just skeptical it will work that simply. One of the reasons I'm unsure is that this is very possibly a random probability, and your examinations will give you a lagging probability measure that is useless for predictive purposes... I don't know if I'm explaining those thoughts correctly. Who knows, maybe I'm wrong.
"New Trading Systems And Methods" Great summer read.
Ok i understand now what you mean. Didnt mean to sound like a prick I just got annoyed with how you came off. I just would welcome any new discussions here. Ultimately I would like to see the site evolve to more than just a career advice forum. This site is dedicated to finance professionals which includes independent traders. Just want to see the community grow and the only way for it to is for posts like these to be made.
I welcome all discussions about trading systems on this forum.
"Oh the ladies ever tell you that you look like a fucking optical illusion" - Frank Slaughtery 25th Hour.
Sed pariatur enim culpa omnis. Odio ipsum quia dolores nihil qui veritatis. Incidunt ex perferendis nisi tenetur vel nulla voluptatem.
See All Comments - 100% Free
WSO depends on everyone being able to pitch in when they know something. Unlock with your email and get bonus: 6 financial modeling lessons free ($199 value)
or Unlock with your social account...