Julian’s Journal

[LEONARDZIR]

Matt,
I am curious. Are you running a hedge fund. 40,000 shares of a 20 dollar stock is 800k per stock position in an MOC system. Maybe I misinterpreted your message.
Len

[Anonymous]

Len, no not running a fund. Just doing under my own business name for my own personal gain. I don’t manage anyone else’s money. My account is with IB Australia, but I don’t trade any Australian systems because quite frankly I can’t find systems that work well enough compared to what I am doing in US market.

I wasn’t doing 40,000 units on a $20 stock back then. At that time I was setting my lower price limit quite low ($5). Positions were more like $200K per position, so the 5 dollar stocks were the culprits that caused needing to order 40,000 units. The $50 stocks were 4,000 units needed, so those were generally fine to fill in either one hit or just a few orders broken up.

The optimisation traps I fell into in those early days were the good old ones where less positions always generally gave more favourable returns and lower drawdowns, but I was just looking at stats and falling in love, not thinking enough of the bigger picture. When you looked at the equity graph there were periods of relatively little returns and then during the volatile times is when you would see a nice profit spike. At the time I was thinking “Hey this is perfect, it complements my trend system so well, because this is the system that can make money when market is volatile and my trend system can handle making the profits in the boring times”. I thought I was onto something great in terms of low correlation between systems. It was impossible to trade though. Psychologically it was impossible. Almost no trades for weeks on end then one day suddenly a dozen fills. Not only did it feel like I was sitting around doing nothing for weeks on end, I would be allocating cash to the system where in three weeks of low volatility I would get maybe one or two trades, and in that same three weeks my trend system could have made good money, so it was also annoying me that it made allocation of cash between systems very difficult because I thought I was “wasting money” having it allocated to a system that seemed to perform hardly any trades.

The other issue is the one we are discussing here. Many of the profits on that 15 position system, in those big lumpy gain days, were coming from just a handful of trades. If they didn’t get filled fully, or just partially, all the trades giving those juicy overall returns were not able to be realised. I eventually saw the light and realised I needed to radically increase the number of positions such that there was much less dependancy on a profitable system making a return on only a small number of trades. I now trade closer to 100 positions, on track to over two thousand trades per year, and all the returns are coming from the boring as hell +/- 5% trades. Yes, there are still a handful of long tail trades, but in a backtest of 15 years and 25,000 trades there are only perhaps 500 trades outside the core +/- 5% distribution. As a check, upon completion of the backtest, I dump it into a spreadsheet, sort for the largest winners outside the 5% and delete them. Then sort largest losers outside the 5% and delete them, then total the profit column and see what the difference is. If my total gain over 15 years was showing 350% net with all trades before deletion included, and after this little deletion exercise of these highly profitable and highly lossy tails ends up being something like 330%, then I’m happy because it is obvious the core return is coming from the +/- 5% and not the tails.

In my previous system, I’d be more like 3,000 trades over a 15 year backtest, and still 500 or so trades in the tails. I would delete them using this method and hey presto my gains over 15 years would drop from 350% to 60%. Obviously this was terrible but at the time I couldn’t see the forest for the trees.

The reason why I fell into this trap earlier is because I have just come from over 15 years in manufacturing for the retail/FMCG industry and a German employer. There, the game was about efficiency maximisation, cost reduction programs, trimming losers and maximising profits on profit makers, a global supply chain where 4 decimal place costings made a difference. This mentality which was so ingrained in everything I did forced me down a particular line of thinking during the optimisation process, like “I can make the same amount of money with half the amount of trades, perfect!”. And also “I can trim my commission costs from 0.68% of my profits to 0.45% of my profits”. And “I can process and manage 15 orders compared to 100 orders”. All these things pushed me down the optimisation line in the pursuit of efficiency to a very stupid place in the design of my systems, believing that I had the bases covered well and was running everything “efficiently”.

I have learned now that squeezing the last drops of efficiency out of a system makes no sense if in the end you can never trade it in real time and if all the metrics are not well balanced and working in harmony. I do see a place for optimisation and make sure I now do it in a way that I feel is a good balance of razor sharp precision and blunt instrument brute force, however it does need to be performed in the right ways with the right balance. I won’t say “Curve fitting is bad and should never be done” because we all need to chase a reasonably good curve fit for our systems to feel comfortable it is a workable system and that we will be able to realistically trade it in real time.

It is actually quite funny because I used to have such bitter arguments in my previous employer, me accusing them with metaphorical examples like trying to run a cute little 50 seat French restaurant with the back of house management systems of a formula one team. And, in the end, I did the exact same thing when I was totally at my own control with my own money and chased the efficiency path at the behest of all other practicality and reason.

[JulianCohen]

It is so true that you get out of trading exactly what you want, only we never realise what it is that we actually want. It is a credit to you that you have seen through your biases so early in your journey and corrected them.

[JulianCohen]

December’19

Short-Term Systems

US MR: 6.62% (Annual -5.1%)
Combined US MOC: 2.38% (Annual -3.93%)

Long Term Systems

S&P 500 Momentum: 0.75 % (Annual 5.85%)
NASDAQ Momentum: 8.53% (Annual 4.79%)
Long Term NASDAQ: 7.6% (Annual 10.10%)

ASX Growth -1.7% (Annual 10.66%)

Total Account: 2.13% (Annual 2.45%)

Overall a forgettable year as far as CAGR but I believe I have set myself up now for a cracking 2020 with the addition of three new MOC systems. Time will tell. I might have to rewrite the mantra to “next 50,000 trades” though as I am placing up to 270 orders a day.

[JulianCohen]

384 trades last night on 4 MOC systems! Took two hours to load into STT.

And I was thinking of adding two more MOC systems that I am stress testing right now.

[Anonymous]

Brilliant!
Don’t hesitate. Add ’em.

[Anonymous]

p.s. are you happy to share how many max positions you are running in each system?

[JulianCohen]

Sure. They are 80 positions at 5%. I found that to be the sweet spot.

[Anonymous]

Thanks.
I’m doing similar (100 positions).

[JulianCohen]

January ’20

Short-Term Systems

US MR: -1.00%
Combined US MOC: 7.49 %

Long Term Systems

S&P 500 Momentum: 2.2%
NASDAQ Momentum: 2.94%
Long Term NASDAQ: -4.08%

ASX Growth 4.93%

Total Account: 3.15%

[Anonymous]

Nice result there! if only they could all be 3% months!

[JulianCohen]

Thanks for your help with the MOC systems Matthew. Running 5 systems now.

[Anonymous]

Wow I didn’t know that I actually helped in the end!

What things did you try differently that ultimately made the difference to what you are now running? What changes were good improvements and what changes were pointless?

[JulianCohen]

Well keeping $10 as a minimum price made a huge difference. It actually made me keep going with the idea as when I was testing in real time with $1 minimum I got too many missed trades. I was about to bin the whole idea until you mentioned changing the minimum in a post.

Cranking up the number of positions so that the missed trades didn’t matter so much, and the chance of missing trades due to excessive size was lessened.

Not worrying about whether an idea made “trading sense’ before testing it helped me create more systems that worked, even though my biases thought that they wouldn’t.

Anything that was pointless….I don’t think anything is really pointless as if you don’t try something you can’t be sure it won’t work. I might try something and it doesn’t work for me, but you might try it in a slightly different way that I hadn’t thought of, and it works for you.

[JulianCohen]

Today I did my monthly checkup and the MOC is not looking as rosy as first thought.

Primarily the R2000 systems underperformed in real time. One system has only been running a few days but the other which has been running since Dec is up 9.39% in backtesting but only up 5% in real trading.

5% is not to be sniffed at BUT it’s half the backtest and that doesn’t bode well.

I’ll continue in real time for this quarter and see how it performs but it has a red flag on it already….

The other one that has only run since 27th Jan is up 1% in real time as opposed to 2.19% in backtesting….say no more

[Author]

Posts