Question of probability

Discussion in 'Trading' started by Kastro_316, Nov 17, 2005.

  1. When you enter a position there is a 50/50 chance it will go either way, correct? Up or down...

    Let’s say there is a 50% chance the stock will raise, and a 50% chance the stock will drop. Assume that you take 1 technical indicator and 1 fundamental indicator that predict the stock will rise from historical data (tree stucture). Will this create more probability that the stock will rise? Or does this just create an anomaly?

    Is it possible to trade on probability?
  2. BKuerbs


    No. This assumption is often made, but nevertheless wrong.

    Just because there are two possible results does not mean they have the same probability: You may die the next minute or not. Do you believe that is a 50/50 chance? Certainly not.

    The problem is that you do not know the chances of the position rising or not. Making false assumption is not really helpful.

    The only realistic thing you can do is look at the distribution of returns of your stock/commodity in your time frame and then deduct a probability.

    Unfortunately, this is only part of the whole equation: you will have to account for external shocks. E.g. if it is a stock take into account the CEO hiding some foul debit (remember Refco?). While these events certainly have a low probability there are others: like the stock market tanking, terror attacks etc.


    Bernd Kuerbs
  3. Ohh ok. Just wanted to know :) Sorry for the dumb question
  4. BKuerbs


    Huh? Did I make it sound like a dumb question?

    Sorry, wasn't my intention.

    Have a nice day.

    Bernd Kuerbs
  5. No no, i just meant i could of found this out with a little research. Just wanted to ask people here. :)

  6. One of the best answered :)
  7. nitro


  8. Well... Yes and no. It depends on how you define your events. I will try to expand a bit on what I mean. Across ET you will read about how and what it is that produces an edge. An edge being something that is strongly correlated to what is going to happen next and thus provides your reason to get in now... The probability there is slightly different.

    Were I to trade probabilistically, I would do several things. The 50/50 is true only when you separate things to a discrete finite manner. This means freezing every single tick and evaluating a single tick independently. Remember, 50/50 says the outcome is either 1 or the other. Where is the 50/50 in the spectrum of market events? It is at a finite level. It is the point where either the next tick arrives at the bid or at the ask. Thus the probability of an individual tick at bid or ask is 50/50. If the bid=ask, I am still defining the condition as exclusive since either the tick was a buy or sell. That is 50/50!

    Beyond that, the probability gets complicated, or interesting as the case may be. WHY??? Well for one, we know that sequential ticks are related, well correlated if we want to be technical. However, immediately when you start grouping pairs of sequential ticks, several other variables have to be considered which for the purposes of my point are OT!

    To keep things simple then, it is only a matter of evaluating and/or defining things independently. For example, what I sometimes like to do is just consider each 5 min bar as an INDEPENDENT EVENT. As a second step, I look for relations between 2 specific variables that construct each 5M bar. So it is one step further since it is regards a second variable that your not aware of as implied by your post... VOLUME! Now, there is the classic argument that price leads volume, volume leads price which I have been tired of arguing but for the purposes of speaking logically, I will briefly elaborate. So for the purposes of your Q, I evaluate each 5M bars construction independently. The entire 5M is the independent event around which I build my stats. Because it is my STRONG (ie. statistically significant) belief that volume precedes price, I am always operating around this basis.

    Why is my belief as such??? VERY SIMPLE! Well, if you look closely at every tick, you see that price is really a pair of prices since each buy is at the ask and every sell is at the bid. Thus all entries are really at a Price LEVEL (ie. either the bid side of the current PRICE LEVEL or the ask side of the current PRICE LEVEL). Of course, I am talking about incoming orders, not the ones sitting at bid/ask (ie. the orders being lifted). So a price level is just a bid ask pair. The incoming flow of ticks is just toggling the flag at the bid or ask for each price level. Given this, then what precedes a new price level? AGAIN SIMPLE!!! ALWAYS VOLUME.

    You can think of a price level as a pair of cups... One for bid, one for ask. Each cups volume is PRECISELY their size which need not be equivalent (ie. the bid cup's volume is exactly the bid size and the ask cup's volume is exactly the ask size). The first cup to have it's volume filled yields a new bid/ask price level (ie. a new pair of cups). To be more clear, once volume at either the bid or ask surpasses their respective bid or ask sizes (ie. overflows), the price level has been filled and simultaneously a new price level begins getting filled. So right away, it is evident that without sufficient volume to overfill either side of the price level, it is not possible for the price level to change, and without continuing sequences in price level changes, there is no differential for which to bank profits. Every bar of any asset of every time frame regardless of whether it is overnight or intraday is bounded by the above logic. For most, it is admittedly hyper activity.

    So with the above logic and a 5M bar, I further expand the above logic to a principle which says if the volume and price are related wrt time, then volume will be related to the time segment's volatility. For illustrative purposes, you can see in the following plot, evidence for this relation. It is taken from another post I had posted elsewhere some time back.


    The chart picks a sample day and independently plots each 5M bar's volatility against the bar's volume. The volume is evaluated as a rank in which the sample of actual volume is sorted according to size (ie. if the bar had the greatest volume in the day, then it would be ranked as 100, low volume bars will have lower ranks). This can be done over a weeks worth of 5M bars as opposed to the same day in order to get a stronger distribution.

    So, here you see that volume and volatility are related. This means that there is high probability that if a bar is going to have large volume, than it will also have large volatility. This is important because this implies that at the large end of this distribution (ie. large volume), the direction of the currently developing bars progression will be sustained. It is sustained because simply, the rapidity with which cups are being filled is predominately to one side, either bid side (short) or ask side (long). At this point it's just a matter of monitoring and analyzing the current bar's progression as it begins by doing a 5M projection of the bar's elapsed volume for a peak to find it's volatility, and bid & ask volume to continually sneak peek at which side is predominating... I definitely recommend steering clear of bar's who's projected volatility is less than 80%, the percentile's are inversely proportional to risk... anomalous? Many think so. But these stats work!....

  9. dac8555


    You can calculate your probability based on the given system that you use.

    For example, If you use a triple moving average crossover with a specific instument, that has been tested to work 60% of the time under specific contitions and a specific time frame on a specific csecurity, you can formulate something. but there are also outside influences that you somehow have to take into account that may or may not be fed announcements or CPI index.

    I think odds are a good way to think about it...just like poker. Make sure your margin of error is realistic.
    #10     Nov 17, 2005