Frosty's trading bot goes live part 2

Quote from optioncoach:

The point is basically the same, he should focus on those two days in particular to see if there is something he can fix.
More...

Focusing only the negative days would not be useful imo. This is yet another reason I believe something more insidious is at work... Filter the 4-digit outliers(?) and there is still a measurably significant inverse...
5 loss days, -1710 tot, -342 avg
7 win days, +1350 tot, +192 avg

My back of the napkin calcs do not include sim mode days.

Even with small sample, with no longer-term results, these kind of numbers suggest something is not right. As earlier, imo, it has to do with exits since position sizing is not a factor.

Osorico :)

Quote from frostengine:

It is a rather small sample size, but I can at least look to see if there is any glaring piece of information consistent between the losing days thus far. Here is a breakdown of everyday:

6/18 +130.4
6/19 -374.8
6/20 -1200
6/21 +320.4
6/22 -199.2

6/25 +150.8
6/26 -539
6/27 +35.6
6/28 -369
6/29 -1334

7/2 +115 but on sim mode
7/3 +70 but on sim mode
7/5 +220.8
7/6 +245

7/9 +250
7/10 -229


Now I will go and look at charts from each of those days and see if anything stands out
More...
 
In response to the stop loss. Yes, I have done those tests very early on. I have always had VERY bad results with any stop loss. That is why I do not use a stop and hence the reason its possible to have huge losing days. Also, the bot does routinely go below -$500 to end positive.. just like out of the 14 days it has ran, 8 days the gain has gone for more than +$500 at some point in they day only to end up negative on a lot of the days. Or to end up being only a few hundred positive by the end... it has some wild swings.

Next i'm sure someone will suggest why not just stop the bot when its at +$500.. again bad results with that as well.
 
Quote from frostengine:

In response to the stop loss. Yes, I have done those tests very early on. I have always had VERY bad results with any stop loss. That is why I do not use a stop and hence the reason its possible to have huge losing days. Also, the bot does routinely go below -$500 to end positive.. just like out of the 14 days it has ran, 8 days the gain has gone for more than +$500 at some point in they day only to end up negative on a lot of the days. Or to end up being only a few hundred positive by the end... it has some wild swings.

Next i'm sure someone will suggest why not just stop the bot when its at +$500.. again bad results with that as well.
More...
But if it is a "trend following" bot, shouldn't it stop, maybe stop and reverse, if the trend has changed?

As for no stop loss, I, personally, wouldn't risk more than 1% of the account on any trade. If you've always had bad results with any stop, the bot isn't competitive, not even viable, in my opinion.

It seems like an "I know I'm right" bot. But if that is the case, why does it ever exit? Why not wait until the "trend" resumes?

I suppose that we have too little information to speculate. Would it hurt to describe the bot's algorithm?
 
i would say in my experience (i'm well documented n00b).. i set every new ATS w/ 3 parameters in place at the get-go.
-buy/sell signal
-hard stop
-hard profit target
just my way i guess.
i like it when my system will perform well w/ tighter stops. imo means less risk.
also means my entries are doing well. i hate looking back and seeing an entry, followed by price wandering all over the map.. then finally a target or stop. just looks like randomness..
 
Frosty: You might consider doing a comparison of how the bot has performed in high volatility vs. low volatility environments. It could give you some valuable insights.
 
I am not saying focusing on those 2 bad days is the end all of analysis but those two are some significant outliers that have led to a large net loss. One can look at those days and try and ascertain:

Were these heavy chop days where a trend following system gets eaten up?

Is there an hole in the system where it misses certain moves and you can optimize the parameters?

Why on other days was maximum loss lets say $300 on a handful of days but well over $1,000 on others? Is it just the nature of the system where you will get these drawdowns or is there something to fix?

Another poster compared high vol and low vol or wide range days to compare results.

I think there is plenty that can be gleaned from the short time frame data and one should not let bots run rampant and just assume they will come around.

Best of luck.
 
(1) This is not directed at anyone in particular but it sounds like some people are commenting in this thread who have no experience with developing an automated trading system.

Its easy to say, "Why not just following the trend and when the trend changes reverse with it". Try coding "Follow the trend" in software - if it were that trivial everyone with a CompSci degree would be rich.

I think part of the problem is that humans look at a chart and see exactly what they want to see (the trend is obvious in hindsight) but when you try to code specific rules you quickly learn that the patterns that you see on charts cannot be so easily quantified with hard and fast rules.

(2) There is a lot of speculation about why some days were big losers and other days not but Frosty has not released the individual trade profit/loss on those days, never mind the specific entry/exit points that you would need to really understand what is going on.

I don't blame him for not sharing those details but trying to offer suggestions on how to "fix" his ATS when you don't even know even basic details on what it is doing (like the individual trade stats) seems pointless.
 
Quote from GTS:

There is a lot of speculation about why some days were big losers and other days not but Frosty has not released the individual trade profit/loss on those days, never mind the specific entry/exit points that you would need to really understand what is going on.

I don't blame him for not sharing those details but trying to offer suggestions on how to "fix" his ATS when you don't even know even basic details on what it is doing (like the individual trade stats) seems pointless.
More...

Indeed.
 
Great conversation across the board.

I'm glad to see that we're at least entertaining the idea of putting more stringent risk mananagement parameters in place, because without them Frosty's trading bot is never going to successfully be able to trade more than one contract.

And while I do agree with GTS that designing a working, profitable trading bot is no easy task, it actually is being done by successfully by hybrid programmer/traders who have mastered both skill sets ... and I'll bet you dollars to donuts, those bots are all about risk management too, infinitely more so than human discretionary (or even system) traders.

That's what I would infer, and that's what I've seen on the board, but of course, someone like traderdragon2 would actually have to chime with their 2 cents to confirm.

Good trading,

Jimmy Jam
 
Quote from JimmyJam:

Great conversation across the board.

I'm glad to see that we're at least entertaining the idea of putting more stringent risk mananagement parameters in place, because without them Frosty's trading bot is never going to successfully be able to trade more than one contract.

And while I do agree with GTS that designing a working, profitable trading bot is no easy task, it actually is being done by successfully by hybrid programmer/traders who have mastered both skill sets ... and I'll bet you dollars to donuts, those bots are all about risk management too, infinitely more so than human discretionary (or even system) traders.

That's what I would infer, and that's what I've seen on the board, but of course, someone like traderdragon2 would actually have to chime with their 2 cents to confirm.

Good trading,

Jimmy Jam
More...

Good insight. Successful bots HAS ALOT TO DO WITH or emphasize about strong risk mananananananagement.

Though, this is to an extent of having the approapriate and sound strategies (doesn't necessarily have to be about having an "edge")
 
You must log in or register to reply here.
Back
Top