Frosty's trading bot goes live part 2

[dcraig]

Quote from JimmyJam:

Learn how to trade.

JJ

Aside from being fairly arrogant, you are assuming that automated systems must attempt to emulate a human trader.

This is not necessarily true. There are many ways of dealing with risk and drawdown, and close stops are not the only way.

One great potential advantage of automation is that it allows diversification that is not available to a discretionary trader. If you are trading, say, ten different systems with capital equally allocated, then a 20% drawdown in one system is neither here nor there.

Of course developing several good uncorrelated systems and the infrastructure to reliably trade them is no simple matter. But it is certainly a promising direction for research that can be overlooked if one obsesses about emulating a discretionary day trader.

 

[JimmyJam]

Quote from dcraig:

Aside from being fairly arrogant, you are assuming that automated systems must attempt to emulate a human trader.

Nope, I'm not being arrogant. I've seen Frost do this go-around before (with the naive well-wishers like yourself and walterjennings in tow).

I'm just recommending another path of action for Frost before he blows out his automated trading account again.

This is not necessarily true. There are many ways of dealing with risk and drawdown, and close stops are not the only way.

But you got to admit it's a pretty good suggestion, or better yet, maybe you'd be willing to let'em trade your account with this system?

Yeah, thought so.

One great potential advantage of automation is that it allows diversification that is not available to a discretionary trader. If you are trading, say, ten different systems with capital equally allocated, then a 20% drawdown in one system is neither here nor there.

You hear that Frost, all you need to do is come up with 9 more systems and the money to trade them ... boy, my recommendations are looking better by the minute.

Of course developing several good uncorrelated systems and the infrastructure to reliably trade them is no simple matter. But it is certainly a promising direction for research that can be overlooked if one obsesses about emulating a discretionary day trader.

On this we are in agreement, "developing several good uncorrelated systems and the infrastructure to reliably trade them is no simple matter".

It would be so much better if he would just learn to trade.

Jimmy Jam

 

[BigBubba]

Quote from EricP:

If you haven't already, please be sure to backtest/papertrade your system on the data for the past week. Assuming that the paper trading shows the same performance as actual (and your backtesting methods were valid), then you would appear to just be having a run of bad luck. However, if paper trading showing better performance, then you should investigate why, and adjust your assumed backtesting slippage values, etc, to ensure that your adjusted backtesting performance still warrants activation of the system.

Best of luck.

sry, but is this simply a test to verify that the system is executing the entry properly as coded. in other words.. to detect possibly errors in the signals? otherwise, i'm not sure i follow what this exercise is for..
thx
bb

 

[fickletrader]

Quote from JimmyJam:

He was able to do that because, well, he probably knows how to trade.

So he took the basic frame work that you mentioned in your other thread and designed a bot around it that would mimic the habits of a good trader (OK, here's a hint for free - a good intra-day trader won't allow a position to move against by over $1,000 multiple times.).

JJ

Jimmy chill with this, it isn't constructive. Have you ever programmed anything? How about an enterprise trading application? From your messages, I'd guess you have no idea how software models can work. While I concede that trading experience is important for the software developer creating an ATS, there is so much more to it than growing a big swinging dick, like you seem to think.

Frosty is well into creating his masterpiece. Forget about blowing out accounts and trying to eyeball tops and bottoms, that is gibberish. Go read the nuclear phynance forum for a few hours and you'll realize how silly you sound. It wouldn't hurt Frosty to spend more time reading it too. His project feels rushed to me, but his path will ultimately lead to success if he can study sufficiently, do enough groundwork up front, and be patient enough to nurture his project for as many years as it takes.

 

[TraDaToR]

I think people are a bit too hard with Frosty.

What is a 3000 $ DD on an ER2 system? nothing... I had a 7000$ one before shutting the system just to see the bot climbing back and making new highs....

When you are trading an ER2 trend following system, any tests that shows lower than a 5000 $ DD is totally overoptimized and stupid. To me, Frosty is unlucky.

But, would I trade Frosty's bot? Never... You are making 3 trades per day on let's say a 50 $ average trade, paying about 50 $ of commish and slippage per day to trade one instruments that is constantly evolving by gaining participants, noise...

Even if your system will become profitable in the end, just think about the money you pay to the market and how the odds were against you... reproductibility of the past cannot overcome these kind of expenses.

 

[trader225]

I just realized how everything could go very wrong. Is it bar/candle trading? Waiting for a bar/candle to complete before making a decision? 5 minute bars/candles?

On fed day, look at a 5-min chart. There is more than one 5 min bar where, even if you were in a trend and on the right side, waiting for a bar to complete could have been expensive.

 

[EricP]

Quote from BigBubba:

sry, but is this simply a test to verify that the system is executing the entry properly as coded. in other words.. to detect possibly errors in the signals? otherwise, i'm not sure i follow what this exercise is for..
thx
bb

Yes, that's the purpose. To ensure that the paper trading is valid when compared with actual trading with the system being tested. Where there any 'real' (losing) trades that the paper trading did not take? Did paper trading exit at the same point as the 'real' trades. Did the real trading generate a higher level of slippage than was assumed in the paper trading, etc? When a paper trading system works beautifully, and then the real trading immediately following the backtesting period goes awful, you need to convince yourself that something obvious is not wrong. The first way to check for that, IMO, is to verify that the paper trading performance during this period is equally as bad as the real trading.

 

[GTS]

Frosty, just let it keep running.

This is the exact same situation as last time where you just happened to pick a drawdown period to start running live and then pulled the plug right before it started its next profitable move up.

You've done the proper backtesting, optimization and live vs sim comparison - there is nothing to do except let it play out.

Not sure why many here are second guessing you and asking your questions that have already been answered earlier in this thread. And then there are losers like JJ that are simply jealous that you are working on something that they cannot do themselves so of course they attack it.

I think your decision to disable sending trades to the exchange next week is a mistake because of murphys law - you know it will end up being profitable and you will miss out.

I hope you reconsider just letting it run live.

 

[JimmyJam]

Quote from fickletrader:

Jimmy chill with this, it isn't constructive. Have you ever programmed anything? How about an enterprise trading application? From your messages, I'd guess you have no idea how software models can work. While I concede that trading experience is important for the software developer creating an ATS, there is so much more to it than growing a big swinging dick, like you seem to think.

Frosty is well into creating his masterpiece. Forget about blowing out accounts and trying to eyeball tops and bottoms, that is gibberish. Go read the nuclear phynance forum for a few hours and you'll realize how silly you sound. It wouldn't hurt Frosty to spend more time reading it too. His project feels rushed to me, but his path will ultimately lead to success if he can study sufficiently, do enough groundwork up front, and be patient enough to nurture his project for as many years as it takes.

Going into unsustainable drawdown isn't constructive either if you aren't financially or emotionally equipped to handle it.

Frosty is pretty much creating the same thing that he was creating a year ago, with similar performance issues.

I'm not talking about tops or bottoms (but that is one useful way to define highs and lows btw), I'm talking about defining trend, entering a position at a good time, putting prudent risk measures in place and taking profit.

The bot doesn't do any of those things well, mainly because Frost doesn't know how to tell it to ... and for the record fickletrader, I'm being a lot more positive than the cheerleaders sitting on the sidelines making trite suggestions while Frosty's bot is taking'em to the cleaners.

Good trading,

Jimmy Jam

 

[BlindLemonBoosh]

Why are we getting all excited over 10 days of results? Guys, get a grip. In order to assess its performance, the bot needs quite a bit more time to run.