Frosty's trading bot goes live part 2

Quote from GTS:



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...

I've posted about this a few days ago. I don't think Frosty is looking for a solution.

As a matter of fact, I'm guessing, he doesn't see any problems arising because the model has been working for a week. It's very similar to a discretionary trader dealing with a losing position, hoping that the market (not direction but tendency, in this case) will turn out favorable. Of course, it may turn out favorable... but it's a matter of being risk adverse...

I'm not sure what Frosty wants to do in the future with his career. But if he is to become a professional trader ( I used to run my own hedge fund but sold it to a bank... and now I am trading bank capital because I suck at marketing and collecting OPM...), the decision he's made is not risk adverse.

Again... there is chance that his model may work in the future. But due to the outlyer drawdowns he's had, the model tends to be too risky. Especially, when it's the only system he seems to be trading...
 
Another post...

There has been some talk about sample size. The usual approach is to keep things flexible.

1. The general / conceptual tests should be tested in a large sample size with years of testing. For example, testing if MA crossover works or not. (There is no decisive measure of what works, depends on the type of model you're developing)

2. Tweaks can be done with small sample size. Small sample size (recent data... think of Exponential MA where recent data has stronger weight) can be done.

One advice for this is... if the models have adaptive parameters... is to stay out of statistics but to focus on quantifying the markets for a measure, not the trade stats.

Max contingencies = Statistical.

Mean contingency = Markets.
 
Quote from nkhoi:

today was kind of ranging in the morning however 20 EMA was below 34 EMA for the whole time, hm...
More...
a mirror of yesterday, yester it ranged then it went down, today it went up first then it ranges. Note the position of 2 MAs, wonder what happen if you filter thru them?
 

Attachments

Well still green for the week. Not much happened today...

Today +$30.80
Week to Date: +$52
Since going live: -$2861.2
Total days: 15
Avg per day: -$190.74

Up Days: 8
Down Days: 7
 
Ferget what i said about TD2 (prolly on vacation), TSGG is doing a pretty good job of showing, not why the bot doesn't work, but rather why it has a lessor probability of ultimately being successful following its current trading parameters.

But within the logic being shown, is the way to create a very successful trading model, for whatever type of trader you might be (hint: Osorico has it down pat on the discretionary side).

I just shoot for my measly $100+ per day, per contract, as weekly average ... once you start stack'in 'em, I'll be making more doing this than I ever could at a J.O.B.

Good trading,

Jimmy Jam
 
Quote from frostengine:

Today -$239
More...
Frosty, any chance you would be willing to post the specific entries/exits for today's trades?

I know it will only invite a rash of useless comments from the peanut gallery but I think at this point you really don't have anything to lose.
 
Quote from frostengine:

Today -$239
Week to Date: -187
Since going live: -$3100.2
Total days: 16
Avg per day: -$193.76

Up Days: 8
Down Days: 8
More...

I think it's safe to say that you are NOT trend following. If the bot was, today should have been a major pay day. Today is what a trend following bot should have pounced on.
 
I already know what caused today's issue. Its the same thing that plagued the strategy 6 months ago on days like this. yes today looks like a powerful up day and the bot should have no problem following the strategy....

However, the problem comes in with:

1. The bot is not allowed to start trading until 9:00 central. By 9 today at least in the russel the majority of the day's range was now gone. For the next several hours the russel moved mostly in a 4 pt range up and down. No real direction, with a slight edge upward. This caused a few chop trades.

2. Now around 1ish central time it began a down move that was much longer lasting than both the up and down moves of the previous hours. The bot interpreted this as a new trend developing. And consequently took a short position.

3. shortly after entering this short position the market started rocking up. Was moving fairly quickly upward and by the time the bot realized it was in the wrong direction, it was already down some money....

So basically that is what happened today. The first reason listed is the biggest reason of all. Not starting until after 9:00 central when the majority of the move was gone and then getting stuck in several hours of chop is what produced the loss
 
Brownsfan,

Read my last post. It clearly explains why on the outside today looked like an easy day for trend following, but when you actually try to capitalize on it during the day. It was not so easy....

Granted had I been trading YM or something instead the day would have worked out better.
 
You must log in or register to reply here.
Back
Top