Frosty's auto-trading bot goes live with REAL money

Frost,

as far as determining #1 vs #2, what I would probably start with is taking the #trades for each system, the average return and the standard deviation of the returns for each and run a t-test to see if the there is a statistically significant difference between the average return of system #1 vs. #2.

Even better, if you have side by side statistics for each day for both the systems, you can do a paired t-test which will tell you if the differences between the 2 systems are significant.

Probably the biggest assumption for this is 'will the distribution of the returns moving forward be the same as the distribution of the returns of the trades already observed?'....probably not, but its the best thing that we've got, and the better you can understand your risk the more likely you'll do better than a lot of other people here.


-rob
 
Quote from robbob:

Frost,

as far as determining #1 vs #2, what I would probably start with is taking the #trades for each system, the average return and the standard deviation of the returns for each and run a t-test to see if the there is a statistically significant difference between the average return of system #1 vs. #2.

Even better, if you have side by side statistics for each day for both the systems, you can do a paired t-test which will tell you if the differences between the 2 systems are significant.

Probably the biggest assumption for this is 'will the distribution of the returns moving forward be the same as the distribution of the returns of the trades already observed?'....probably not, but its the best thing that we've got, and the better you can understand your risk the more likely you'll do better than a lot of other people here.


-rob
More...

here's an old thread that might help you along in these matters.

btw, the thread's OP also recommends giving the most weight to the most recent series of data, and suggests a couple'a years worth of trades is more than enough to get to where you have to go.

Activate/Deactive System?

Regards,

JJ
 
-$144 today......
Method #1 would have lost -$194

Another bad day.... I cant give up on the system, as soon as I do is when it will start to perform well again
 
Frost,

how many contracts are you trading, and what sort of drawdown can you expect in percentage terms relative to the current amount of equity in your account?

-rob
 
Frosty:

A bit of advice: by the very act of logging your daily trades, you are selecting anecdotal "evidence" in favor of statistical evidence in order to determine your faith (or lack thereof) in your system.

In other words, you are needlessly putting your emotions on a roller-coaster by a) choosing only a very small sample of what could be years of and thousands of data points (if you keep running the system), and b) attaching your ego and reputation to these anecdotal results by putting them on display here for all the world to see.

This talk of abandoning your system(s) would stop right away if you stepped back and let your systems run as they were intended to: fully automatically.

Now, I thank you very much for making your findings and endeavors available to the rest of us as a learning resource, but it might be best for you if you stop watching your system day in and day out and simply put an automatic alarm on your system to tell you when it's met your objective indications of failure (such as a 33% drawdown in your entire account equity , or whatever is acceptable to you)

If you still wish to convey your results to the rest of us, you could always program it to automatically post results or charts at some website or perhaps in this forum (though I don't know if bots are allowed here).

Or, better yet, it may be best for you to run the system, make a million bucks, and <i>then</i> post the results, along with your play-by-play analysis (and maybe the full source code to your system--lol). But letting your play-by-play calls mess with your game is doing you great harm psychologically, and maybe financially if you decide to give up on or tweak an otherwise good and sound system.
 
dchang0,

Good post..... makes a lot of sense, and you are probably right that keeping this play by play everyday is probably going to hurt me in the long run....

It may be a bad time to bring this up... but I may have a strategy that is FAR FAR more accurate in terms of risk/reward and win%.. only downside is it suffers from what my method #2 does... that is I dont have enough data to backtest on except what I have captured over past 4 months.. so hard to say if this strategy will continue to perform well...Actually I dont know how it even performs yet, running simulations now... basically what happened is I was analyzing some key differences between strategy 1 and strategy 2... and a light bulb went off..... so I have been trying to figure out the best way to implement it.. but I am fairly confident that it may turn out to be multiple times better than what I am currently trading....

The strategy is VERY different from my current batch of strategies...... it will share only a common reasoning behind why entering the trade, everything else about how it works is pretty new...

Wont know until these simulations are done however..... I have had ideas like this before that seemed GREAT... but backtested like crap.... so I am eagerly awaiting the results...

Even if this batch of simulations goes poorly, I am confident that I am onto something... I just don't know how close I am.... Hopefully it turns out to be something, I will post some results in the morning if anything is worthwhile....

Unless I receive absolutely STELLAR results I will still continuing trading method #2 as I have been as I know that strategy is good.. I have no idea how this strategy will work... plus I will probably need to take a few days on the sim account with it to make sure no quirks in how it works... anyway, I am probably jumping the gun here as for all I know it may have a negative expectancy when the results come back ;)

PS... sorry about the mad man ramblings..... this thread probably does not portray me in the best light.. however, I am very real and upfront with everything I post here, and want to share what is going on in my head as I work on getting a profitable automated system

PSS..... I just had an idea which may allow me to backtest this on all of my data going back all 4 years... with a slight modification... will have to wait on that.. but could be good news... still waiting on results btw...
 
Quote from frostengine:


Wont know until these simulations are done however..... I have had ideas like this before that seemed GREAT... but backtested like crap.... so I am eagerly awaiting the results...
More...

How long do your sims take?
 
Not very long.. around 20 min, depending on what I need it to do....

The results came back only so-so.... It has room to possibly be good... As for as risk/winning % it looked a little better.. total profit was down however.... and equity curve really no smoother....but thus far looks like I am still missing something...

Something to work on, but nothing for me to get too excited about just yet
 
Quote from austinp:

<i>"You need to stop autotrading and seriously learn how to successfully trade the markets by your own hand FIRST. Anyone with 6-12 months of discretionary trading under their belts could have picked up on just how strong the downtrend was going to be well within the first hour of the trading day.

It's very surreal what you're trying to accomplish. It's like you're trying to experience what it's like to live in Paris without ever physically being there. You're guessing at what it might be like instead of knowing what it IS like."</i>

That is hands down one of the most point-blank posts I've yet to read here. 100% true... every word of it. In order to construct and operate a robust, lasting mechanical system, one must fully understand the market AND the system tools operation thru all market conditions.

Too many non-traders attempt to mask inexperience by taking the computerized route. Sadly, that does not work, ever. Until one understands exactly how & why the system operates = behaves in all market conditions, drawdowns cannot be survived.

An experience surgeon could build a robotic machine to perform surgeries mechanically. An engineer and/or computer programmer cannot... they must first accquire surgeon skills before ever dreaming of building a mechanical model to replace said skills.

It is impossible to know what we do not know. Successful system writers must have discretionary trading skills = experience, that part of the equation cannot be circumvented at all.

Hope this helps :>)
More...
While real-life experience is always helpful, it's not always necessary. Your analogy above is hand picked... think of other analogies where the overall movements of a system is pre-defined.

I know a very successful programmer who built a computer program to play chess that defeated the first Deep Blue using commodity PC hardware and open source clustering software (this happened 2.5 years after Deep Blue was created, think Moore's Law, but even then a very significant acheivement). He never played a game of chess in life. But he understood academically the universe of possible chess moves, game theory, and how to evaluate success/failure within the chess universe.

I know the c-robots programming champion (this goes back 18-years ago) who never played a video game in his life. He was a computer scientist and physicists and understood the universe of motion and game theory.

I know a poker bot programmer who has never played a live ring game in his life, but he built me a bot that earns 0.75 PTBB per hour. With rake rebate it was a nice little bot farm. Lots of maintenance to beat the continual Party upgrades and anti-bot software and floating around dozens of accounts... but it worked. This programmer never played a hand of poker in his life. He simply applied basic +EV theory. The bot can't beat the higher limit games (yet) where the decision tree is much more complex, but for any low limit 5/10 and below it could easily identify +EV scenarios and take in a small earn hour-after-hour, 24-hours a day, multiple accounts. Party is shut down for US customers and now it is being converted to Stars.

In the universe of trading, the # of possible actions is very straightforward. Buy or sell (to open or to close). There are many ways to analyze the markets, but it is fairly straightforward to see that even a simple model can be successful if it can identify truly +EV events and trade them... automatically... even without prior live trading experience.

I also disagree that *discretionary* trading experience is required. There are many traders here who have only traded systems, manually, and can be very successful with an automated method.
 
You must log in or register to reply here.
Back
Top