Welcome to NexusFi: the best trading community on the planet, with over 150,000 members Sign Up Now for Free
Genuine reviews from real traders, not fake reviews from stealth vendors
Quality education from leading professional traders
We are a friendly, helpful, and positive community
We do not tolerate rude behavior, trolling, or vendors advertising in posts
We are here to help, just let us know what you need
You'll need to register in order to view the content of the threads and start contributing to our community. It's free for basic access, or support us by becoming an Elite Member -- see if you qualify for a discount below.
-- Big Mike, Site Administrator
(If you already have an account, login at the top of the page)
As a part of the edge exploited by most random entry systems is derived from kurtosis, adding stops and targets in the initial model doesnt really give you anything useful to build on as the theoretical expectancy will always be zero minus transaction costs.
Without exploiting the fat tails there's nowhere to go really
Fat Tails would be the result of positive auto-correlation of price. You could also exploit negative auto-correlation of logarithmic returns. What cannot be exploited, is no auto-correlation at all, which means random behaviour.
Not feeling exploited. Threads like this one are a fun thing. Can be used for sharpening one's mind...
Is there any way that I can please see the code so I can tweak it for what I need it for? I have tried writing this on my own, but I cant get the random function to recalculate would really like to see how you did it. Thanks.
I created a random entry system using 2 set contracts on the 6E back adjusted contract from 2001 to present day. Again because market orders are used there is possibly a slight bias, but if you compare each system to the others that bias should be relative as opposed to comparing to an absolute benchmark. It does not follow the wait 30 second rule, it goes 1 takes a trade on the next candle which are 5 minutes. But I don't think that makes a difference as my samples have very large numbers. Everything uses 2 contracts, no slippage or commissions are added.
We can see that the 1:1 and 2:1 win rates converge very closely to the mathematical expectations of 50% and 33%. using the formula 2/sqrt(N) to estimate the error at the 95% level the smallest sample size here 42435 has an error of ~9.7%.
I have no idea why the 2:1 expectancy is so low compared to the rest, i thought it would be closer to 0. But the scale out definitely shows that you are infact cutting your winners short over the long term. Also since you may take a winner at the 1:1 level and the 2nd contract may lose your time in trade is about equal for winners and losers. Basically coin tosses will never have an edge, but you can minimize your negative expectancy based on targeting/stop placement. Which would state that a 1:1 all in all out is the best methodology.
Random theory is interesting wherein financial markets are not random. If they were truly random, then our coin toss trading method should have positive expectancy.
Interestingly, options trading can create a higher probability win ratio where one can make money on options' trading in a market that can move 4 ways out of 5.
Futures trading, on the other hand, can make you money in a market that can ONLY move 1 way out of 2.
I recall years ago an FX trader so exasperated with his losing trades, that he purposely chose to trade opposite his daily trading plan. He began to make money! It continued on for some weeks. My guess was that he was "dis-lex'ic" and looked at his monitor 'upside down.'
After watching the FT71 webinar's I have taken a great interest in this. I've been a futures trader for >10 years but have always suffered from letting winners turn into losers and holder losers too long so that my losses outstrip my gains.
I have decided to test this myself. I've been programming C/C++ for >20 years but since Multicharts.NET is based on C# I've decided to learn that. I will post results once I have them. But I want to make some observations.
These observations do not take slippage or commissions into account.
For the all-in/all-out case with an equal risk reward the expectancy is zero since a move of the same amount in opposite directions is equal probable, i.e. 0.5*(-x)+0.5*x=0. Thus it seems without a positive edge in entry the all-in/all-out is not a winning strategy.
For the scale out case things get more interesting. Let's look at level 1 Risk/Reward of 1:1 and level 2 Risk/Reward of 2:1 and 2 contracts. There is a 50% probability of stop out for 0.5*(-2x) loss. There is a 50% chance of hitting the 1st target. The next assumption normally presumed is that the 2nd level is 66% vs 33% for total retracement once the 1st level has been reached, i.e. 2 times more probable to move x ticks to 2nd level vs retracing 2x ticks to stop. So expectancy is 0.5*(-2x)+0.5*(2/3)*3x+0.5*(1/3)*(-1+1)=0. I'm not sure I believe that this is true, i.e once the market has moved in a certain direction it may have a more favorable tendency to do so. Any "momentum advantage" will lend positive expectancy to the scale out which the all-in/all-out can never achieve.
After implementing a random entry with a scale exit system in Multicharts.NET using the E-Mini(ES) open outcry system and exiting any open orders at the end of the day, the results confirm what has been expressed in this thread before. Basically with a risk/reward ratio of 1:1 for target 1 and 1:2 for target 2 and exiting 50% at target 1 and 50% at target 2 and no movement of the stop if target 1 reached, the result is that the expectancy is approximately 0 using 3 months of tick data.
Some runs have a small positive result and some runs have a small negative result depending on the initial seeding of the the random number generator but the average is very close to 0.
I haven't tried different Risk/reward ratio's yet but without an edge straight scale out at the above targets is a losing system when commissions and slippage are taken into account.
I think the point of the coin toss exercise (let's say 1 contract - N ticks gain target, N ticks stop loss - random long/short) is that (neglecting slippage and fee) the expectancy is 0. An expectancy of 0 means that one can't use this method to generate income but also that one can't lose either (assuming an account large enough to avoid let's say 5 consecutive losses that will empty the account.)
So, this show to beginner traders that it doesn't matter which strategy is used if looser are not cut...
Imagine, a trader trading for few months and loosing let's say $500 every week. Switching to the coin loss exercise will show that he will do much better (P/L essentially fluctuating around $0 - fee/slippage) just by using stop loss and profit target of the coin exercise -)