null
Page 1 of 8410 123 ... LastLast
Results 1 to 10 of 93

Thread: null

  1. #1
    Just amassing some market figures this Friday night.



    I truly love frequency supply...

    -- Concerning the Journal --
    Trading is very boring once you figure it out...
    Leaves a lot of time to sit, think, and write...
    This journal is where I spend a lot of this moment.

    That is for me to document research/work so I can reference it later.
    Some of this is really for the LOLs and to discuss my comments on matters I find important.

  2. #2
    Quote Originally Posted by ;
    quote Great explanation. That is exactly the issue with the 5-linear subsequence issue from EURUSDD. Although you will see at least 1 of the 4 conditions fulfilled 93.75 percent of the time, the likelihood of P(fulfilled with 5th nmber|not yet fulfilled with the 4th amount) remains 0.52 ...
    This is called conditional likelihood and imho, EURUSDD's managing of likelihood left a lot to be desired. This is a similar dilemma as Gambler's falacy.

  3. #3
    Heres the other side of that.


    Really interesting stuff.


  4. #4
    Quote Originally Posted by ;
    Just amassing some market numbers this Friday night. image I really love frequency distribution...
    can you clarify the aforementioned stuff Please

  5. #5
    The Kelly Criterion

    Can there be one betting amount that is statistically better than another? Yes.

    Have you ever wondered why Casinos have table limits? Is yes.

    If there was a statistically significant sum to bet at any particular time, how do you understand? The Kelly Criterion

    In probability theory and intertemporal portfolio choice , the Kelly criterion, Kelly egy, Kelly formula, or Kelly bet is a formula used to Ascertain the optimal size of a series of bets in order to maximise the logarithm of wealth.



    As an example, if a gamble has a 60 percent probability of winning ( p = 0.60, q = 0.40), and the gambler receives 1-to-1 chances on a winning bet ( b = 1), then the gambler should bet 20% of the bankroll at every opportunity (f* = 0.20), in order to make the most of the long-run growth rate of their bankroll.

    I don't always bet on the market... But when I do, it's a Kelly bet or a fractional Kelly bet.

  6. #6
    A Correlation Problem

    I like the currency market. My favorite thing about it is the liquidity and efficiency of the market. It really is a beautiful thing.

    However, there is one difficulty always facing us when it comes to coping with the currency market. That issue is, there is a major correlation problem.

    Because of the way currencies are traded in pairs, every single major pair (28 total) are all correlated with one another. This might not seem to be an issue, but it is for accounts, and it is also an issue when it comes to assessing performance in connection to the anticipated performance after a substantial quantity of samples.

    When it comes to getting an anticipated result for a method, one thing that has to happen, is you want a huge sample of trials to evaluate. In an ecosystem it is hard to get independent transactionson two distinct pairs using 4 distinct currencies. This causes our results when assessing a batch of transactions to skew a bit.

    Instance:
    Say we're trading EURUSD and AUDJPY. Two Pairs, 4 distinct currencies.
    For simplicity sake, we suppose that we've got a system that's right 50% of the time.
    Now if these two events are really independent and we have a transaction on every pair, the probability of both being wrong or right would be 25% (.5 * .5 = .25).

    However, we know these events are not really independent because even though these two pairs comprise of 4 distinct currencies, they're still pretty heavily correlated.

    EURUSD-AUDJPY Correlation Readings(1 hour, 1 Day, 1 Week, 1 Month, 3 Months, 6 Months, 1 Year):


    In this example, let us say every transaction lasts a day, so we'll use the daily move correlation coefficient of .72 (Shockingly high if you believe that you're avoiding correlations be trading what appears to be two entirely separate pairs).

    Let's plug that significance value into the following formula to see what the probability of both being wrong or right is now:


    The math works out to be 43%.

    Thus if you're trading this system on two distinct pairs with 4 distinct currencies, you might anticipate the probability of the two transactions being incorrect is just 25%, when in fact, as a result of underlying correlations, the probability of being transactions being incorrect is 43% (Important note: that does not mean the probability of both being correct is 1-.43. It is 43%).

    Now imagine how this significance will skew your results in regard to their expected results when you presume your transactions are independent, but they aren't.

    Now imagine how this skew will lead to more losses (or wins) and the way that'll exaggerate the drawdown (or run up) on small accounts.

    Bottom line... If you're trading any two pairs, it is basically impossible to have your transactions put on each pair be independent of one another (if this makes sense).

  7. #7
    Occurrences Expected Outcomes
    DISCLAIMER: In this post I'm talking strictly about win rate when I say expected outcome. I am not talking about PnL.

    In the last entry, I spoke a little about how merchandise correlations can really skew the expected results of a system. Another thing which could really skew your results, is that the number of occurrences that you exchange to test your system. This seems like a no brainer, but that I see people on here all the time trading weekly egies or long term daily egies, that wonder why they aren't receiving the outcome.

    Through this 3-4 weeks they might have taken 12-24 trades. Hell if they required they'd have 20 trades.

    Let us presume that this systems win rate is 50%.

    Our variance within this situation is 5 ((20)(0.5)(0.5) = 5). Our sigma is 2.236 or two

    68 percent of their time, this trial will result in 8 wins and 12 losses or vice versa.
    95 percent of the time that this trial will result in 6 wins and 14 losses or vice versa.
    99 percent of the time that this trial will result in 4 wins and 16 losses.

    The point being, you could try this egy, do 20 trades, have only 6-7 wins and 13-4 losers, and this performance would nevertheless be statically normal based on the number of occurrences. Now factor in the issue with correlations in the post and say these 20 trades were on pairs but nevertheless correlated, these amounts are skewed more. You could easily have the skew from correlated and 17 losses due only 3 wins.

    Consider how many traders would quite this system and continue on the to next, only to get the same lack luster results which are again statistically normal because of the few of occurrences.

    So how do we get nearer to our expected outcome? Independent happenings!

    As a result of the central limit theorem, as we scale up our happenings, the variance of our outcomes will slowly shrink as we get nearer to our expected outcome. And with exactly your expected outcome, you will end up with sufficient occurrences. When we were to perform 80 trades rather than 20, our variance would cut in half. In other words, if you want to slice your variance in half, you have to do 4 times as many occurrences.



    Here is how HFTs make money each and every moment. When they own a system which wins 50 percent of their time and is profitable, they can be green each day trading the number of occurrences. It's not about shadowy pools and front running...

    I think it's crucial for new traders to comprehend how the number of occurrences and merchandise correlation can really skew your results from the expected results. If you do not understand how these things work, you will dump systems which are actually profitable, and heavily invest in systems which are complete garbage, but look do into the skew.

  8. #8
    You are better off computing stats from US as opposed to US dollars and Nikkei and SP500 compared to risk on and risk off.

    Smart money chase returns. Dumb money chase history.

  9. #9
    Non-Linear Trends and Serial Correlation
    This post is essentially only a link to one of my favored answers on this forum. It is from mzvega also it talks about trends in the market. So I and many others can reference it for afterwards, I wanted to create this.


    https://www.forexforum.co.za/broker-...-trade-uk.html

  10. #10

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •  
This website uses cookies
We use cookies to store session information to facilitate remembering your login information, to allow you to save website preferences, to personalise content and ads, to provide social media features and to analyse our traffic. We also share information about your use of our site with our social media, advertising and analytics partners.