Fully Automated Stocks Trading

[globalarbtrader]

Reading recommendations:

William Eckhardt: A Chapter in "The New Market Wizards"
Howard B Bandy: "Mean Reversion Trading System: Practical Methods for Swing Trading"
Larry Connors: "Short Term Trading Strategies That Work"
Laurens Bensdorp: "Automated Stock Trading Systems"
Cesar Alvarez: https://alvarezquanttrading.com/

William Eckhardt. 1 out 2 guys who created the famous Turtles experiment. No systems there. But he speaks about timeless truths of systematic trading. My personal favorite.

Comp. Sciense prof. Howard Bandy gives examples of systems in AmiBroker's code. Not sure if they still work, but they inspired me to look into MR in the first place and my first strategies were based on his work. I've read all his books but this one is the most practical in my opinion. Looking back - his strategies are a bit naive as they use low trade count and have very high win rate. Easy to overfit. But he has lots of right ideas / examples of research.

Larry Connors - a famous trader. He is not purely mechanical, like Linda Raschke, but uses lots of formal research to power his decisions. Sometimes his findings still work sometimes they don't. But it is a good value. Lots of the stuff he mentions will have value as individual components of a system.

Cesar Alvarez worked for Larry Connors and now on its' own. He is focused on AmiBroker and periodically publishes research on his blog. It is an interesting read. He dissects and measures individual edges fairly often. Great example of how to do it.

Finally, Laurens Bensdorp. Biggest value, in my opinion, is the overall method he describes. That can be summed up - holy grail of trading is trading multiple non correlated strategies. After reading his book there is a decent chance that will be internalized. The way his strategies are described is a high quality. Decent description of how to look at them together.

In general, don't expect that any strategy published in any book will work. Mostly they don't. Sometimes they "somewhat” work. Sometimes they work, but over market types which are not what we're seeing now and you will feel like they don't work, but they actually do. After reading a bunch of a good ones you get a sense of what a system looks like, gain knowledge / skill required to validate, find something that work but maybe not as great as you want, go thru attitude/expectations adjustments, pickup individual ideas and then finally put something together that has a potential.

I've probably went thru couple of hundred of other ones. Mostly garbage.

Val

"holy grail of trading is trading multiple non correlated strategies"

Put this on my gravestone, please.

GAT

 

[.sigma]

"holy grail of trading is trading multiple non correlated strategies"

Put this on my gravestone, please.

GAT

right....

the risks (greatest) of any portfolio is variance and correlation. Thus both are encapsulated in the covariance matrix.

the question I have it, of course depending on the personal goals of the individual and how that person views idiosyncrasies in his risk tolerance, how many uncorr strategies should you have?

 

[globalarbtrader]

right....

the risks (greatest) of any portfolio is variance and correlation. Thus both are encapsulated in the covariance matrix.

the question I have it, of course depending on the personal goals of the individual and how that person views idiosyncrasies in his risk tolerance, how many uncorr strategies should you have?

As many as possible.

GAT

 

[leon7]

Reading recommendations:

William Eckhardt: A Chapter in "The New Market Wizards"
Howard B Bandy: "Mean Reversion Trading System: Practical Methods for Swing Trading"
Larry Connors: "Short Term Trading Strategies That Work"
Laurens Bensdorp: "Automated Stock Trading Systems"
Cesar Alvarez: https://alvarezquanttrading.com/

William Eckhardt. 1 out 2 guys who created the famous Turtles experiment. No systems there. But he speaks about timeless truths of systematic trading. My personal favorite.

Comp. Sciense prof. Howard Bandy gives examples of systems in AmiBroker's code. Not sure if they still work, but they inspired me to look into MR in the first place and my first strategies were based on his work. I've read all his books but this one is the most practical in my opinion. Looking back - his strategies are a bit naive as they use low trade count and have very high win rate. Easy to overfit. But he has lots of right ideas / examples of research.

Larry Connors - a famous trader. He is not purely mechanical, like Linda Raschke, but uses lots of formal research to power his decisions. Sometimes his findings still work sometimes they don't. But it is a good value. Lots of the stuff he mentions will have value as individual components of a system.

Cesar Alvarez worked for Larry Connors and now on its' own. He is focused on AmiBroker and periodically publishes research on his blog. It is an interesting read. He dissects and measures individual edges fairly often. Great example of how to do it.

Finally, Laurens Bensdorp. Biggest value, in my opinion, is the overall method he describes. That can be summed up - holy grail of trading is trading multiple non correlated strategies. After reading his book there is a decent chance that will be internalized. The way his strategies are described is a high quality. Decent description of how to look at them together.

In general, don't expect that any strategy published in any book will work. Mostly they don't. Sometimes they "somewhat” work. Sometimes they work, but over market types which are not what we're seeing now and you will feel like they don't work, but they actually do. After reading a bunch of a good ones you get a sense of what a system looks like, gain knowledge / skill required to validate, find something that work but maybe not as great as you want, go thru attitude/expectations adjustments, pickup individual ideas and then finally put something together that has a potential.

I've probably went thru couple of hundred of other ones. Mostly garbage.

Val

Val, good day. Thank you for reading list. Agree on most authors/ideas. It's a discovery for me Laurens Bensdorp's work. What about AI, part of it ML, do you use it in your trading? Howard Bandy is highly educated in this field, also wrote a lot in this topic recently. Also what about idea of systematically rotating capital to better performing strategies? It's like a system for systems management, with own rules and position sizing...

 

[Stephbaker]

This is so informative

 

[ValeryN]

Val, good day. Thank you for reading list. Agree on most authors/ideas. It's a discovery for me Laurens Bensdorp's work. What about AI, part of it ML, do you use it in your trading? Howard Bandy is highly educated in this field, also wrote a lot in this topic recently. Also what about idea of systematically rotating capital to better performing strategies? It's like a system for systems management, with own rules and position sizing...

Not in any automated sense.

There is truly too many things that can work in trading to tap into everything. This is mostly why. Resource constraints.

If I would ever to start a fund, I would probably put a guy on researching applicability of ML/AI to trading. Meanwhile, I am curious to learn if someone else successfully applied that on retail level. Biggest question would be what exactly they are applying it for.

Val

 

[ValeryN]

I will explain what this chart is about in a minute.

Many times I heard that trading has emotional asymmetry. For example on negative days we often feel not great, well, because the money was lost. On positive days we worry about keeping the profits. Then edges don’t last forever and a trader need to be evolving. The list of negative is actually quite long. But I find even listing all those negative things is… well… quite negative, so I won't. Those should be enough to make a point.

Both Larry Hite and William Eckhardt referred to trading as a net negative game emotionally. Listening to the best traders out there and hearing that they struggle with those, I seriously thought about it, and, over the time came up with a list things to help with that. Here are some

1. Put together 100% automated executions (inspired by Ed Seykota)
2. All routine reports generated are “emotionless”. No PnL or $s there
3. No checking “how things are going” until daily model vs live report is generated
4. No checking of market in general during trading day
5. No checking account balance till end of week or better till end of month
6. Take a moment to appreciate up moves in equity and positive events

While I am not oblivious that those things don’t last + luck being part of it, to counter the negative emotions of trading I do take a moment today to celebrate 13 up days in a row. The chart attached is of SPY for the same period. All my systems worked together extremely well and managed to ultimately pull the equity up every day. Not sure what was my longest winning streak in mechanical trading, this could be one.

This is especially pleasing as SPY is at a net loss, market thru quite a few different types of days, plus we’ve seen a volatility spike.

Neither month or a year are over yet and anything can happen from now till then. But there is no point to worry about that every day.

Today, I am sitting on ~80% profits YTD and Today is a great day.

Val

 

[andycct]

The book Thinking, Fast and Slow by Daniel Kahneman echos your net negative sentiment. We have problems with loss aversion and trouble acting on probability because of it.

+1 and -1 = zero, but emotionally, we need more like +1.5 and -1 to feel like zero.

One solution he puts forward is heuristics (rules of thumb). While they may be inaccurate, if based on data, they are often more accurate than our judgement in predicting future response of messy systems.

 

[ValeryN]

The book Thinking, Fast and Slow by Daniel Kahneman echos your net negative sentiment. We have problems with loss aversion and trouble acting on probability because of it.

+1 and -1 = zero, but emotionally, we need more like +1.5 and -1 to feel like zero.

One solution he puts forward is heuristics (rules of thumb). While they may be inaccurate, if based on data, they are often more accurate than our judgement in predicting future response of messy systems.

Very good points. I heard of similar ratios from a psychologist.

Val

 

[ValeryN]

Update on penny stocks:

As I was looking how my experimental penny stock system was doing, it reminded me of this quote - "Trading is not about being a baller and crushing the market so you can go sip a drink in the swimming pool. Trading is a grind of applying that thin edge over and over with discipline as nearly perfect as you can muster" (Adam Grimes)

Most mechanical edges are indeed so thin that are commonly discarded.

Finding such a small edge and harvesting it over as many trades as possible is probably the second closest thing to holy grail of trading, after trading multiple non-correlated systems.

This is one example -

My penny stocks system not only requires perfect execution but a patience way beyond what most people would give it. None of my systems are anywhere close to HFT domain, but they far ahead of what a human trader will be able to execute. Plus when it comes to penny stocks 1-2 cents can kill a system or make it a great success. Most traders will roll their eyes and pass on those systems, and the ones who wouldn't - will likely drop it on a first sign of trouble. But I am a bit more patient.

First few trades didn't look too good as I was seeing some slippage that wasn't supposed to be there. Slippage is never good but this one especially as it was coming from differences in backtesting data source vs consolidated tape. That doesn't happen much with higher priced stocks. But I knew I will be getting rebates due to little execution tricks. My guesstimate was around +0.01$ positive slippage over every 4-10 trades executed. So I though to myself - things might cancel out. Plus volatility in this universe is crazy and I can use truly micro-sized positions, that makes individual trade risk very low in terms of account % impact.

So it's been a bit less than a month. ~50 trades later the new system set new highs 2 times and made a bit of money, though it's somewhat irrelevant. Patience did pay off so far and model vs live executions diff came down to -55.28$ (accounting for partials, rebates, commissions and slippage). Meaning live returns were ~55$ less than model. Considering my account size this difference is negligible. So that part looks promising.

More time will be required to see how much I impact the market and if model is degrading. Probably will take 3-6 months to find out.

IB must be very pleased with this development as my commissions tripled. I now consider their alternative commission model, could cut costs by ~30%. Need to crunch some numbers.

Meanwhile in the world of trading sensations and educators - great comeback by Will Karaman.

Val