Supalgo

[themickey]

This is about my latest algo creation for trading stocks on the ASX, but it could be applied to any stock exchange. It is a continuation from: https://www.elitetrader.com/et/threads/active-investor-gets-out-the-machine-gun.319908/

The purpose of the journal?
A year or two ago we had on ET a big time TA sceptic and my journey began then to post short journals on ET to disprove that individual, I think I have either 3 or 4 journals now on ET, all have been profitable by a large margin, better than market returns.
Anyhow this time not attempting to prove anything, other than to show encouragment to other traders that a retail trader can with experience pull off a good ratio of wins & profit without too much effort, that is once you have become familiar with general stock market behaviour. Also this is my contribution to ET on what is intended to be a meaningful positive experience.

Although I have owned Amibroker (AB) and downloaded daily data for approx 16 years, I havn't always used it. When I was a wage slave I may have traded a few times a week as an EOD trader & strictly using AB. Then I gravitated to another method the last couple of years using a combination of web sites and running googlespreadsheets and cloud. This as I no longer am a wage slave so more time on my hands.
My method now is a live breakout stock trader, recently purchased intraday data, this gives me the luxury of catching potential & actual breakouts as they happen ahead of the wage slave EOD traders.
Therefore now using a combination of AB, Cloud, and web sites, the creation of Supalgo is self designed & written into AB by myself, the culmination of years of trading ideas put into a series of formulas. It is two algos into one algo.
First part scans the latest ASX (or any market data) and strikes out dud stocks and then ranks the winners. Winners leave an indelible mark of quality, the algo then ranks into an order of acceptance these quality picks.
The 2nd part of the algo finds those ripe for buying, in otherwords, part A of the algo determines 'what' while part B determines 'when'.
In AB an algo is called an 'Exploration' when initiated/run. Running an exploration which is formula heavy may take about 20 seconds to scan 2500 odd stocks. Both partA and partB of the algo run as one, it number crunches everything at once and spits out a results which may consist of 50 results from a field of 2500. I just then look at maybe the top 15 contenders in the results ranking and then by eyeballing the chart to look for further clues on what best to buy. It's this last part, looking at the chart which I determine would be very difficult to code into an algo.
Currently Supalgo is; 4 pages long, 1650 words, 9900 characters, 238 lines.

For some ET traders, supalgo would be interpreted as SupaSlowNoGo or SuperSlothNoShow, because my trading style in general is don't pull trigger until there is a very high probability of a good win, then it's "go in big or go home". Big being a relative term, 'big' equals 'small' compared to a professional. What I'm attempting to say is, at times there will be no trading, just sitting.

I will update the journal whenever a trade is executed, near as live as possible and a weekend summary of results.
Journal will run until it doesn't, have no intention of forever as I get bored talking to myself.

 

[themickey]

A brief about my background, Tulip Muncher (Dutchy) by birth, raised as a Kiwi in Aetearoa- Land of the long white cloud (NZ) and lived the past 20+ years in the Land down under Oz (Australia)
Apprenticehip in a shipping company as fitter machinist. NZ.
Went to Oz on a hitchhiking trip, worked in NSW Newcastle State Dockyards on ship construction.
Went back to Kiwi worked some years in pulp & paper industry in the forests of central north island.
Spent many years deer hunting as a weekend sport.
Became a hotel barman for a year or so.
Worked in Rotterdam as a bellhop in a major hotel.
Back to NZ employed machining moulds for bottles in Auckland, super crap & dirty job.
Married first time.
Got a job working in powerstations, about 15 years. Attained boiler & turbine tickets.
Hunting as hobby.
Came to Australia, marriage broke up almost immediately, spent 5 years as mobile mechanic working on waterblasters, steam cleaners, installing car hoists, lathes, milling machines, attained electrical registration.
Hobby, began trading.
Went back to Kiwi, worked in a number of electrical jobs including fibre optics installations.
Hobby; dating women & trading.
Got my foot in the door in engineering sales, sized, bought & sold control valves and oil & gas burners, purchased from Germany.
Came back to Oz, worked 10 odd years in valve sales.
Hobby, dating and underwater snorkeling & trading.
Went to Darwin, more valve sales.
Moved to Queensland, sales of steel & pipe into mining.
Hobby attempting to automate futures trading.
Moved to WesternAust, more pipe, steel & valve sales.
Got a job with a Texas based subsea company, engineering planner.
Married 2nd time.
Worked for German Co supply mining equipment, also was workshop planner.
Retired from workforce.
Hobbies, gardening, trading, building prototype quadcopter drone for underwater photography.

 

[themickey]

Below is a history of recent trades, the journal will continue where I left off from my previous 'machinegun' journal. The reason for the continuation from a previous journal is these are real trades with real money, I'm not about to start yet another spreadsheet.

A problem currently is I'm quite fully committed holding numerous positions, (16) and while supalgo may find me stocks ripe for buying I'm rather drained of new cash until I sell, so will be passing on a number of new trade contenders, there is a bit of money in the kitty but not a lot.
On the list below, trades placed last week which Supalgo found were positions, 1, 35, 45 48

 

[drm7]

Interesting bio! One of the things that I've noticed about trading stocks using a model is that the high degree of correlation makes risk management very difficult. First of all, if the market is strong, then you will get a ton of breakouts all at once - especially in the same sector (e.g., mining and energy right now). Plus, once you take all those positions, if the market (or sector) turns down, all of your positions will go against you at the same time. So, if you only risk 1% per losing trade, and you have 15 positions, it's really like one trade with 10%-15% risk!

Some people try to adjust by having equal numbers of long and short positions, but the overall market direction still has a huge influence. One other way (which is a bit more effective, but cuts your upside) is to short the underlying index in proportion to the beta of your positions. So, if you have some mining stock with a beta of 2, you would short 2 units of the ASX as long as you were long that particular stock.

 

[traider]

Below is a history of recent trades, the journal will continue where I left off from my previous 'machinegun' journal. The reason for the continuation from a previous journal is these are real trades with real money, I'm not about to start yet another spreadsheet.


Based on what I read, algo is long only. Have you taken into account the fact that in a bull market any algo that picks random stocks to long will make money?
How do you filter you stocks after it is identified by an algo. Lots of false breakouts and mean reversion happen for stocks too.

 

[themickey]

Interesting bio! One of the things that I've noticed about trading stocks using a model is that the high degree of correlation makes risk management very difficult. First of all, if the market is strong, then you will get a ton of breakouts all at once - especially in the same sector (e.g., mining and energy right now). Plus, once you take all those positions, if the market (or sector) turns down, all of your positions will go against you at the same time. So, if you only risk 1% per losing trade, and you have 15 positions, it's really like one trade with 10%-15% risk!

Some people try to adjust by having equal numbers of long and short positions, but the overall market direction still has a huge influence. One other way (which is a bit more effective, but cuts your upside) is to short the underlying index in proportion to the beta of your positions. So, if you have some mining stock with a beta of 2, you would short 2 units of the ASX as long as you were long that particular stock.

Yes, high degree of correlation for reasons; In this instance because the Resources sector now appears to indicate a re-resurgence back upward after 5 years of heading south which ended in 2016, banks and other sectors appear a little pricey, the PE's within resources are still comparitively low. Therefore PE ratios are part of the Supalgo plan.
So the decision to be heavily weighted resources was a decision made pre Supalgo era, from now on I think my picks will be more diversified because a machine is now involved finding contenders.
So you are correct, I would get a ton of breakouts in resources stocks but my theory is, catch those breakouts, grab the money and exit post haste at the first sign of trouble. That's what I had been doing with gold last couple years, continually in and out, scores of times. However gold is quite difficult to trade, a very wild beast, I think day traders are too good for me and the plunges are dramatic.
Regarding shorting, have looked into quite carefully and decided I would rather exit and sit out rather than short. My strength is not as a shorter.

False breakouts, yes happens from time to time, but mostly get it right as I usually trade in conjunction with the sector, ie, copper breakout, all copper stocks go with it. Sector breakouts beget stock breakouts.
Mean reversion, my algo has that coded in. It is part of the coded plan.

Remember, most of my picks on the spreadsheet were pre supalgo, last week's picks was a iron ore scalp for 3 days, 2 off gold stock positions and an IT type stock - Trademe Group which is based in NZ.

 

[themickey]

Sold on open due to strong USD and Trump Tarrifs as well as profit taking.
OZL copper position, 2500 units x $9.70
SFR copper position 3000 units x $9.35
WHC coal position 5000 units x $5.65
RIO aluminium/iron ore 300 units x $82.45

Bought near open
CAJ CAPITOL HEALTH, 16000 x $0.33
MNY MONEY3 Corp, 2500 x $2.03
Small speccy positions

Attempting to rotate into more defensive positions atm and away from metals.

 

[themickey]

Just bought BGA, Bega Cheese, 700 x $7.50
Sold SAR, Gold position, 12500 x $2.14

 

[themickey]

Bought just now QMS X 9000 X $1.16 (Media company)
Will update spreadsheet on weekly trades end of each week.

I'm going in on small sizes atm while I bed the algo in, I need confidence in it before taking on larger size, by going in small I also get the chance to place a higher number of trades in order to test the win rate.
I won't make a lot of money with small positions but that is not the objective, the objective is to test the system at this moment of time.
So far it is looking ok but early days yet.

 

[themickey]

End of week update, first week of new algo posted positions, slow going, in actual fact $4k regression on the week.
Positions 52 & 53 are losing due to the amount of brokerage fees, I calculate both buy & sell brokerage amounts into the total cost price.