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

[JimmyJam]

Hey Frosty,

Decided to check-in and see how you're doing with the Brain Trust, and I see it's business as usual.

Quick, somebody tell me what kind of day yesterday was ...

a) doesn't get aboard obvious uptrends and
b) doesn't take obvious downtrends and
c) overtrades the chop ...

... if you guessed Option C you would be correct.

You know it's funny. When I was updating the spreadsheet this afternoon during the trading session, based on probability and the negative equity curve that the system shows (regardless, of the mindless number crunching that backtesting it represents) I figured it would be for a loss, I just didn't know how much.(chart.xls).

Without what is appearing to be more and more a lucky break trade that made the initial 1,400+, the system would be rapidly approaching 50% drawdown, with no real knowledge of when you could expect a turn-around.

Here's a scary thought: for every $1,000 dollars the system makes, it losses $1,200 - $1,400.

JJ

P.S. While we've discussed the problems, it remains to be seen what the actual solutions are ... perhaps you can start here and see how other system developers crafted their profitable trading systems: http://www.futurestruth.com/

 

[basis]

Quote from JimmyJam:

Quick, somebody tell me what kind of day yesterday was ...

How is that useful? Who cares what kind of day yesterday was? Last time I checked, no one can trade it.

 

[syswizard]

Quote from jonnyman:

In all honesty if your system does not have the capability to deal with low-volatility trading, developing even a rudimentary filter should be a high priority.

This is a great suggestion. In fact, Keith Fitschen in the "Developing Trading Systems" DVD mentions this as a critical step after the basic buy/sell logic has been established.
I remember reading somewhere that system traders back in the 1999-2000 period had to filter-OUT high volatility periods. Now it appears they've got to filter-out low volatility periods. Bottomline: markets change...no question about that. The issue is: Can/will your system adapt to the change ?

 

[frostengine]

Today should prove to be interesting.... For some reason the vnc server on my trading machine appears to be dead.... but the computer seems to still have connectivity... this means it should still be trading all day, but while I sit at work I am unable to use vnc to connect to and see what is going on.. to see what trades it made.. how much its up/down... change trades... anything...... So essentially until 5 central time today I will have NO idea how the day has gone... hopefully I don't get home and see a BIG negative.....

 

[Gonz]

Quote from frostengine:

Today should prove to be interesting.... For some reason the vnc server on my trading machine appears to be dead.... but the computer seems to still have connectivity... this means it should still be trading all day, but while I sit at work I am unable to use vnc to connect to and see what is going on.. to see what trades it made.. how much its up/down... change trades... anything...... So essentially until 5 central time today I will have NO idea how the day has gone... hopefully I don't get home and see a BIG negative.....

Today's market opening, very trendy. if yr system is trend-following.. should be profitable.

 

[JimmyJam]

Quote from basis:

How is that useful? Who cares what kind of day yesterday was? Last time I checked, no one can trade it.

... that's the whole point.

Frost's coin-flipping trade-o-matic can't discern the chop days, so that's when it tends to generate its most consistent losses.

JJ

 

[frostengine]

jimmy,

Believe it or not but deciding before the fact if the day is choppy or not is not that easy... plus a lot of the big trends normally comes out of a period where it was chopping prior to it... am I suppose to tell the bot to not take trades if its choppy? Looking back at a day in hindsight its very easy to see that day was choppy, but real time not so easy.. unless you know of some great method but thus far I have found nothing to adequately filter out the trades that have a low chance of succeeding as opposed to the ones which turn out to be my +1000 trades...

 

[traderob]

Quote from frostengine:

jimmy,

Believe it or not but deciding before the fact if the day is choppy or not is not that easy... plus a lot of the big trends normally comes out of a period where it was chopping prior to it... am I suppose to tell the bot to not take trades if its choppy? Looking back at a day in hindsight its very easy to see that day was choppy, but real time not so easy.. unless you know of some great method but thus far I have found nothing to adequately filter out the trades that have a low chance of succeeding as opposed to the ones which turn out to be my +1000 trades...

Frosty,
anyone that could discern whether the market is going to range or trend would suck up all the world's weath within a couple of years.

 

[nkhoi]

check out Mark Fisher, the logical trader book, better than average method to divine whether it's chop day or not.

 

[jonnyman]

Quote from roberk:

Frosty,
anyone that could discern whether the market is going to range or trend would suck up all the world's weath within a couple of years.

My sentiments exactly. I was going to say that it's not difficult to forecast chop - it's almost IMPOSSIBLE.

Yesterday is useless for anything but regrets and "what if's".

Yes, developing a filter is somewhat challenging, but so is finding an edge in the first place, and so is being able to program an automated system.

Yes, you are going to miss some trades that originate in chop and end up running in your favour, but the objective is to offset that by eliminating enough chop losing trades to emerge more profitable in the long run.

Look at your equity curve and the corresponding charts. Look at the chop days and your trades. Is there a pattern? I can tell generally whether our machine would have made/lost money in a day just by looking at a daily chart.

Once you know how your bot performs in the worst situations and why then you can move onto an analysis of those charts. Is there some way of preventing/reducing those high frequency losing trades with a universal rule/logic that can be added to the machine? This is one way of going about it - not very scientific, but I find that sometimes if I just study the charts it can illuminate things.