Julian’s Journal

[ScottMcNab]

Julian Cohen wrote:

I feel for you Scott. I just entered my trades and found I’m holding it in my ROC system as well so I got double F*&%$ed. HA HA

Never know, it might go up tomorrow

Fingers crossed

[TrentRothall]

Scott McNab wrote:

Trent Rothall wrote:

Have you tried setting it at say 50% for a day? Just exit on the open if needed, shouldn’t be too bad?

Should work…I’m not sure what effect selling 5% of daily volume at the close 50% below would be…. I need to look to find some stats on what average % of daily volume transacted in closing auction…if it was 20% and 1/4 of that was 50% lower it would impact system adversely I imagine?

Not sure i quite follow? Aren’t you just putting the order in 50% below the entry price so that you are participating in the closing auction? If it’s 5,10, 20 or 50% it shouldn’t make a difference should it… (as long as IB allow it of course!)

[ScottMcNab]

Trent Rothall wrote:

Scott McNab wrote:

Trent Rothall wrote:

Have you tried setting it at say 50% for a day? Just exit on the open if needed, shouldn’t be too bad?

Should work…I’m not sure what effect selling 5% of daily volume at the close 50% below would be…. I need to look to find some stats on what average % of daily volume transacted in closing auction…if it was 20% and 1/4 of that was 50% lower it would impact system adversely I imagine?

Not sure i quite follow? Aren’t you just putting the order in 50% below the entry price so that you are participating in the closing auction? If it’s 5,10, 20 or 50% it shouldn’t make a difference should it… (as long as IB allow it of course!)

I am thinking the final price when closing auction complete is going to be lower than it would be otherwise if I place a large sell order in the closing auction way below the last traded price…ie I might skew the results that were achieved in backtest so that live results would not be similar…in a stock which traded 20% of total daily volume in closing auction, 5% of daily volume way under market price could be significant ?

[TrentRothall]

I just think the 5% of volume would get the next available price no matter where the limit is set? Might need to ask Nick about the auction, i could be miles off!!

[JulianCohen]

I can’t believe I did it again! I set up my ASX orders on the API at 2pm Sunday so I could get it done early and forget about them, so of course they got cancelled at 3:45pm as that’s what the API is supposed to do.

A man who has committed a mistake and doesn’t correct it is committing another mistake.
Confucius

[SaidBitar]

are you using TWS?
Normally TWS restarts once in the 24 hours this is pain and gain
the gain in this you will not do this mistake anymore and the pain logging in every day

[JulianCohen]

I’m using Gateway for the API. I put a recurring reminder in my calendar for every Sunday so I won’t do it again. It’s just on weekends or exchange holidays that I do it.

[JulianCohen]

February ’17

Combined ASX ROC and SPY ROC MOC 3.62%
Combined ASX MR and SPY MOC 6.22%
Combined ASX MOC and RUA MR 5.22%
WTT US 6.02%
WTT ASX -5.28%

Overall Account 4.95%

Again US is performing very well, ASX is not, but it is in line with my MCS so carry on carrying on!

[JulianCohen]

Following Nick’s updated Selection Bias lectures I have gone back over my six MR systems and all but one are running at 97% of all entries so I’m not worried about the selection bias. However my best performer in the Russell 1000 is running at lower than 70% and therefore I have to conclude that its performance so far has been much more based on luck than on judgement. Gutted, but better to find out now.

So I have been experimenting for the last hour or so and it looks like the Russell isn’t going to work for me. I tried the Russell excluding the S&P 500 but it is looking like the S&P 500 is going to be the universe I will settle on. I increased the stretch in order to get the entries to 85% and I think I’ll settle for that. Just running a big MCS on it now to see how the variance will be.

It feels strange to take a system that looks to be performing so very well in backtests and then ‘hack’ at it until the performance is barely touching 20% BUT I now understand so much more than I did a year ago, even six months ago….I know that what I am doing will hopefully give me a system that will perform in reality close to the backtests.

[Nick Radge]

Interesting. I actually switched my MOC system up to the Russell recently

[JulianCohen]

Nick Radge wrote:

Interesting. I actually switched my MOC system up to the Russell recently

I couldn’t get a good performance with that system on the Russell. It was probably too simple, so I switch universe. In fact I ended up changing one entry criteria and got something I was happy with so I’m going with that on S&P.

Interesting you changed from NASDAQ100 to Russell on the MOC. That’s quite a magnitude of change.

[Nick Radge]

I was getting bored sh*tless with the NASDAQ. It needed to do more trading and wasn’t happening. Changing a few things didn’t make much difference so switched universe, tightened it all up and get to feed the habit a little more :unsure:

[JulianCohen]

Nick Radge wrote:

I was getting bored sh*tless with the NASDAQ. It needed to do more trading and wasn’t happening. Changing a few things didn’t make much difference so switched universe, tightened it all up and get to feed the habit a little more :unsure:

Yup I know what you mean.

[JulianCohen]

I just found a way to reduce selection bias on one of my systems. I took the number of trades from 61% to 85% with Historic Volatility.

Here’s the code for HV in case you want it:

Code:

// Historical Volatility
//———————————————————————————
lenPeriod = Param(“Periods”,20,10,1000,1);
HV = 100 * StDev(log(C/Ref(C,-1)),lenPeriod) * sqrt(252);

I used

Code:

HV > 20 AND HV < 30

as a condition.

Worked for me so if you are having trouble reducing signals give it a go anfd let me know how it works. It does of course affect CAGR but that is to be expected. Mine went from around 30% to 23%

[TrentRothall]

Does the spread between Min & Max MCS tighten very much once you have reduced the selection bias that much Julian?

I guess that is the desired result??

[Author]

Posts