Down Days

[Nick Radge]

On highly volatile days like today it pays to go back and closely look at what the system has done in the past (which you should have already done anyway).

Personally I still remember the 2010 Flash Crash like it was yesterday.

No, it wasn’t comfortable.

But trading can’t have the upside returns if it were comfortable.

The payoff is coping with the discomfort.

Because the vast majority can’t.

Since 2007 I see my own system has had 5 days similar today.

Not nice.

But no surprise.

Running a Monte Carlo bootstrap I fell in the 35th percentile, so below the average expected decline, but a long way from worst possible scenario.

Considering this is now the largest single day decline in history, I’m okay with that.

Up until today, that prior day was the 2010 Flash Crash.

I finished that month -5.5% or so.

And as I have stated before, I was okay with that as well.

[TimothyStrickland]

Good point Nick, I was fine Friday and even today but when the futures market and asian markets opened and we are down again, its the first time I’ve been quite uncomfortable in awhile trading. What good is a system if we don’t use it, even when things get dicey.

-Tim

[JulianCohen]

Tim Strickland wrote:

Good point Nick, I was fine Friday and even today but when the futures market and asian markets opened and we are down again, its the first time I’ve been quite uncomfortable in awhile trading. What good is a system if we don’t use it, even when things get dicey.

-Tim

One thing to bear in mind is that Asia often takes it’s cue from the US markets the day before. So they are falling today due to the US fall yesterday.

[Nick Radge]

Quote:

So they are falling today due to the US fall yesterday.

mmm…nah. They’re falling because the US night session is getting smashed again.

[Nick Radge]

So new equity highs in my MOC system today.

That’s why, when you have a proven system, you stand up and swing the bat.

#next1000trades

[JulianCohen]

Mine will have a loooong way to go to get back to new equity highs.

However I just had a good look at the code and found something that would have made a lot of difference to last night.

I’ll investigate further and report back.

[TrentRothall]

what did you drop the other day julian?

[JulianCohen]

Trent Rothall wrote:

what did you drop the other day julian?

My MOC dropped 5%

The code correction I just discovered would have made 15% last night. There’s a good learning experience for you! I have to check on the selection bias of the adjustment though…

[Nick Radge]

I’m struggling to understand why last night didn’t recoup what was lost the night before?

[JulianCohen]

Nick Radge wrote:

I’m struggling to understand why last night didn’t recoup what was lost the night before?

I had this in my entry criteria:

BollBPerc = (C – BBBot)/(BBTop – BBBot);
Cond4 = BollBPerc >= 0.5;

This kept me from getting too many entry signals yesterday. It was meant to give me “good quality” entries and improved the selection bias.

I guess this is one way to find out what works and what doesn’t eh!

[RobGiles]

thanks to an API failure I couldn’t trade last night. I was down 9.7% the day before.

[Nick Radge]

Quote:

thanks to an API failure I couldn’t trade last night. I was down 9.7% the day before.

We’re not sure if its an API failure or not. Two missed MOC orders out of 40 in those conditions is hardly a failure.

I had 6 MKT orders failed to execute but it didn’t stop me taking the trades last night.

[Anonymous]

How did you still make the trades Nick?

Did you, for example, do it manually by being awake and monitoring the market at open, noticing they didn’t fill at open then quickly click a few buttons and do them manually in the first few minutes of trade?

[Nick Radge]

Julian Cohen wrote:

I had this in my entry criteria:

BollBPerc = (C – BBBot)/(BBTop – BBBot);
Cond4 = BollBPerc >= 0.5;

This kept me from getting too many entry signals yesterday. It was meant to give me “good quality” entries and improved the selection bias.

I guess this is one way to find out what works and what doesn’t eh!

Well, I’m not sure about that. What does the longer term backtest and MCS look like with it removed?

If you had it in there for better quality signals, then where are you left now by removing it?

[JulianCohen]

I’m just doing it all now to be sure.

When I put it in it was a very small difference to the results from 2010-2017 and the selection bias was about 1% improved so I used it.

Without it yesterday I had 150 entries. With it I had 28

So 99% of the time it does what I thought it would do, but that one time that makes a big difference to the P/L it didn’t

[Author]

Posts