When to start trading real money

[REDP1800]

Purely subjective.

When do you feel comfortable throwing all of your cash at it? You are going to get a lot of different answers. I will throw mine in as well:

1. What is your maximum exposure?
   
2. What is your risk-adjusted metric? (Sharpe, Calmar, etc)
3. What risk-adjusted metric number makes you feel comfortable? (Ex. Sharpe > 0.8)
4. What is your maximum drawdown, and why? Where do you lose the most?
5. Can you explain mathematically why you win when you win?
6. Have you walk-forward tested your system? How does your system do when trained for N days and traded for N+X days?
7. What is your risk of ruin?
8. Based on (7) what is your ideal capitalization?
9. Have you looked into the kelly criterion and determined your maximum leverage?

These are just a few things I use to evaluate a system. There are many more - for example 50 trades is a statistically significant sample for an approximately normally distributed variable (i.e. log returns) but wouldn't make me feel comfortable in terms of total performance. Fat tails make me more cautious which is why I look to ruin analysis.

I will unfortunately break the news to you that I think everyone else is ignoring. If you cannot program a strategy you are doomed to failure. You are removing the ability to walk-forward and monte-carlo test your system. You are removing all ability to do any sort of real analysis on your system performance over time. Since you cannot program the system pay someone who can. It could be as simple as an excel sheet or an R script. Pay them out of your investment capital. This investment will save you many, many thousands in the long run.

Hope this helps. There is no easy answer. You are going to get a lot of garbage responses. Especially in a forex forum. If you cannot mathematically explain your edge you have no edge. End of story.

EDIT: Sidenote: You can tell a post is worth ignoring based on the amount bold in it.

no one has to explaiin anything mathematically except the number on their statement at the end of the day month year. Mkt psychology is still alive and well. when investors stop putting money in the market won't that skew your data.. blah blah blah.. backtesting means you are sifting through optimizations in the past that you GUESS with numbers and distributions will hopefully work in the future. If it were that easy I would see these guys full of themselves on tv and with Billboards spewing their greatness. I was just saying I wnet down a backtesting road once and the problem is your data sucks your executon sucks and your programming sucks and your ability to re program and optimize on teh fly sucks compared to the BIG money companies and algo traders. YOU ARE TOO LATE TO THE GAME.. get it. Thatis all I am saying and when the mkt is slowed down again.. which it will be eventually then you will say..well that was kind of a waste. I have never witnessed anyone in retail with easy language or pythong or c++ or anything actually make a system that worked all the time. SO mathematics or not you need to know how when and why to impelement your system which probably relies on typical retail indicators. The system is alrady gamed against your startegy when you start making real trades..it all changes? why is this? well because what 99 % of the community fails to realize is that your 1 share your 100 shares tha 1 lot that 10 lot it becomes a part of the system a part of the data and your own trade changes the dynamic and mvoement of the market. how can anyone paper trade or backtest without being in a real environment. You do nto udnerstand the IMPACT of your trades in the actual market. PERIOD. and you never will because math doesn't make guesses and it cannot tell you the future which is a GUESS.

 

[REDP1800]

3 Months demo profitably is ok....start with mini lot sizes and go live if you are confident.

no one has to explaiin anything mathematically except the number on their statement at the end of the day month year. Mkt psychology is still alive and well. when investors stop putting money in the market won't that skew your data.. blah blah blah.. backtesting means you are sifting through optimizations in the past that you GUESS with numbers and distributions will hopefully work in the future. If it were that easy I would see these guys full of themselves on tv and with Billboards spewing their greatness. I was just saying I wnet down a backtesting road once and the problem is your data sucks your executon sucks and your programming sucks and your ability to re program and optimize on teh fly sucks compared to the BIG money companies and algo traders. YOU ARE TOO LATE TO THE GAME.. get it. Thatis all I am saying and when the mkt is slowed down again.. which it will be eventually then you will say..well that was kind of a waste. I have never witnessed anyone in retail with easy language or pythong or c++ or anything actually make a system that worked all the time. SO mathematics or not you need to know how when and why to impelement your system which probably relies on typical retail indicators. The system is alrady gamed against your startegy when you start making real trades..it all changes? why is this? well because what 99 % of the community fails to realize is that your 1 share your 100 shares tha 1 lot that 10 lot it becomes a part of the system a part of the data and your own trade changes the dynamic and mvoement of the market. how can anyone paper trade or backtest without being in a real environment. You do nto udnerstand the IMPACT of your trades in the actual market. PERIOD. and you never will because math doesn't make guesses and it cannot tell you the future which is a GUESS.

ONE last thing.. just so you know the CME has a test ability for the DIRECT access customers were they can "BACK TEST" or trade demo but in real time to see exactly what there orders would have done a millisecond ago in the market. hsitory is history.. for us.. we think in long terms of time.. like I amde a trade this morning and it will show up on my statement tonight. However, direct access and self clearing get the match information instantaneously and it is microsecond history that they can use for the next microsecodn trade.. they know what FCM you clear at ot BD and they know it is retail traders doing one lots underfunded.. so know that also

 

[newwurldmn]

Well, I saw many cases at ET that people doubled or tripled their money within a short period of time, a couple weeks to a couple months, only to gave all of profits back and more. However, I would consider it a real or reliable edge or talent in the case of Optionsinvestor who turned $2000 into more $200,000, a 100X or 10000% return in about 8 or 9 months.

He was outed as a fraud.

 

[gaussian]

no one has to explaiin anything mathematically except the number on their statement at the end of the day month year. Mkt psychology is still alive and well. when investors stop putting money in the market won't that skew your data.. blah blah blah.. backtesting means you are sifting through optimizations in the past that you GUESS with numbers and distributions will hopefully work in the future. If it were that easy I would see these guys full of themselves on tv and with Billboards spewing their greatness. I was just saying I wnet down a backtesting road once and the problem is your data sucks your executon sucks and your programming sucks and your ability to re program and optimize on teh fly sucks compared to the BIG money companies and algo traders. YOU ARE TOO LATE TO THE GAME.. get it. Thatis all I am saying and when the mkt is slowed down again.. which it will be eventually then you will say..well that was kind of a waste. I have never witnessed anyone in retail with easy language or pythong or c++ or anything actually make a system that worked all the time. SO mathematics or not you need to know how when and why to impelement your system which probably relies on typical retail indicators. The system is alrady gamed against your startegy when you start making real trades..it all changes? why is this? well because what 99 % of the community fails to realize is that your 1 share your 100 shares tha 1 lot that 10 lot it becomes a part of the system a part of the data and your own trade changes the dynamic and mvoement of the market. how can anyone paper trade or backtest without being in a real environment. You do nto udnerstand the IMPACT of your trades in the actual market. PERIOD. and you never will because math doesn't make guesses and it cannot tell you the future which is a GUESS.

Did you have a stroke while writing this?

no one has to explaiin anything mathematically except the number on their statement at the end of the day month year.

True. Except for when you want to be taken seriously. Every major hedge fund manager can mathematically explain their results. Look into https://en.wikipedia.org/wiki/RiskMetrics. RiskMetrics are used to mathematically quantify risk. Risk is yet another thing that is necessary to be explained in order to be taken seriously by anyone who matters.

backtesting means you are sifting through optimizations in the past that you GUESS with numbers and distributions will hopefully work in the future.

Backtesting confirms an edge under the correct assumptions. For example, taking in to account the fatter tails than a normal distribution that are experienced with observed returns over a long enough time frame. Walk-forward analysis, a form of backtesting, is far more statistically reliable and reveals weaknesses in your assumptions faster. Paired with ruin analysis you can very quickly reveal "fat tailed" problems in your model. If you have ever taken an actual probability and statistics course at the college level you'd learn about proper model fitting practices. "The past may not indicate the future" is purely legal non-sense and is not realistic in any way. If it wasn't we wouldn't be able to model processes using statistics, financial time series included. The efficient market hypothesis is deeply flawed. The idea backtesting doesn't work is based entirely on that flawed hypothesis.

I was just saying I wnet down a backtesting road once and the problem is your data sucks your executon sucks and your programming sucks and your ability to re program and optimize on teh fly sucks compared to the BIG money companies and algo traders.

I don't play the same game at the same table as the big money traders. My algorithms are profitable, mathematically explainable, and backtested. I think you suck at it. Don't speak for everyone else.

SO mathematics or not you need to know how when and why to impelement your system which probably relies on typical retail indicators.

"Typical retail indicators"? This is how I know this is amateur hour rambling.

and you never will because math doesn't make guesses and it cannot tell you the future which is a GUESS.

Math makes guesses. The branch is called probability theory (https://en.wikipedia.org/wiki/Probability_theory) and it's the cornerstone of modern computational and quantitative investment practices. The past is generally a good enough indicator of the future to indicate an edge. If the efficient market hypothesis was true (which is what you are implying here) we wouldn't have a statistically significant portion of highly successful traders using the past to indicate the future. There are entire branches of mathematics dedicated to the analysis of signals and regimes.

I'm upset with myself giving you the time of day when you can't even be bothered to spell correctly or use punctuation but there you go.

 

[REDP1800]

Did you have a stroke while writing this?



True. Except for when you want to be taken seriously. Every major hedge fund manager can mathematically explain their results. Look into https://en.wikipedia.org/wiki/RiskMetrics. RiskMetrics are used to mathematically quantify risk. Risk is yet another thing that is necessary to be explained in order to be taken seriously by anyone who matters.



Backtesting confirms an edge under the correct assumptions. For example, taking in to account the fatter tails than a normal distribution that are experienced with observed returns over a long enough time frame. Walk-forward analysis, a form of backtesting, is far more statistically reliable and reveals weaknesses in your assumptions faster. Paired with ruin analysis you can very quickly reveal "fat tailed" problems in your model. If you have ever taken an actual probability and statistics course at the college level you'd learn about proper model fitting practices. "The past may not indicate the future" is purely legal non-sense and is not realistic in any way. If it wasn't we wouldn't be able to model processes using statistics, financial time series included. The efficient market hypothesis is deeply flawed. The idea backtesting doesn't work is based entirely on that flawed hypothesis.



I don't play the same game at the same table as the big money traders. My algorithms are profitable, mathematically explainable, and backtested. I think you suck at it. Don't speak for everyone else.



"Typical retail indicators"? This is how I know this is amateur hour rambling.



Math makes guesses. The branch is called probability theory (https://en.wikipedia.org/wiki/Probability_theory) and it's the cornerstone of modern computational and quantitative investment practices. The past is generally a good enough indicator of the future to indicate an edge. If the efficient market hypothesis was true (which is what you are implying here) we wouldn't have a statistically significant portion of highly successful traders using the past to indicate the future. There are entire branches of mathematics dedicated to the analysis of signals and regimes.

I'm upset with myself giving you the time of day when you can't even be bothered to spell correctly or use punctuation but there you go.

it is obvious all of the defenders of automated trading on this site actually are clueless as to what it takes to trade with a good algo. you cannot trade with a PC from home and expect good results no matter what your math..period.

 

[Ldeyana]

The new system should be tested as long as it takes until you are quite confident that it really works without mistake. I find it very important, because your system, as I have understood, cannot be backtested. Try also on Demo account to use larger sums.

 

[wrbtrader]

The new system should be tested as long as it takes until you are quite confident that it really works without mistake. I find it very important, because your system, as I have understood, cannot be backtested. Try also on Demo account to use larger sums.

He stated his system has "clearly defined" rules (his exact words). Therefore, it can be backtested via a code or manually.

The rules I use for this system are clearly defined. There is no instinct here, just hard data...

I think by his 4th post in this thread, he seems to have already concluded what his next steps will be.

What I was looking for was a little more "scientific" answer to the questions. A month will only generate a theoretical maximum of 40 trade observations. Realistically, it will generate 20 trade observations. Hardly enough to deem a system safe in my "opinion" ( I never studied deep into stats). I so far have about 50 observations. So the conclusion I have now is this. 1)The more observations the better. 2) Instead of dumping a lump of money into the account, I will probably just add in small chunks of money in every month.
Know when the system stops working is the easiest part as I have prior trading stats to compare it against.

Thus, seems like he's just going to add small chunks of money until he accumulates enough "observations" to go along with his prior trading stats although I've never heard of a "safe system".

wrbtrader

 

[shatteredx]

There's no "scientific" answer to this question.

William Eckhardt said in 2011 that he requires 1800+ trades (including backtest) before he goes live. I think Robert Pardo said somewhere that he requires 1000 trades.

I say if it's been working for a month, put a little money in it and trade it. Add a little money every month it keeps working until it can grow easily with its own profits.

 

[REDP1800]

There's no "scientific" answer to this question.

William Eckhardt said in 2011 that he requires 1800+ trades (including backtest) before he goes live. I think Robert Pardo said somewhere that he requires 1000 trades.

I say if it's been working for a month, put a little money in it and trade it. Add a little money every month it keeps working until it can grow easily with its own profits.

how does one backtest and try and account for what impact there own system will have in the mkt and how the nkt itself changes daily

 

[Overnight]

how does one backtest and try and account for what impact there own system will have in the mkt...

Now I KNOW you are reaching. You are looking to not just backtest a system you have, but you are looking to backtest a system that anticipates how the market will react to trades you make in the market, before you actually make the trades. Within what time continuum do you exist? Huh, wha?

You are not just smoking the pot and meth, you are also definitely on the wormwood and shrooms.

How the hell can you backtest a system to take into account trades you have never taken? Whoa!