The Lawrence Chan Blog

An interesting video on learning.

This is an interesting concept worth exploring some more. To be specific, I wonder if it is possible to break down the core elements in trading into parts that newcomers can learn more quickly and effectively.

One thing I know for sure though is that blank staring at the computer screen is not training yourself to understand a market. That does not contribute to your 10,000 hours to become an expert in trading unless you really focus and pay attention to what you are observing.

2 comments

I have seen this sign so many times over the years. Today is the first time I think of taking a picture of it.

Just pause for a moment and think about this:

1. This is a manually updated board

2. Custom made

3. Totally functional

My thoughts:

1. One of the administrative staff did this

2. On her own personal time too

3. Maybe the hospital paid for the material cost, maybe not

4. Not even an electronic version of the board can do a better job

What does this sign has to do with trading? Not much.

1. Some of the most useful things we can do to improve our trading may not come from better entry or better exit. It can be as simple as having a quick check on your open position whether it is doing what you expected it to be doing at any point in time.

2. If something works well and serves the purpose, do not try to fine-tune an existing tool unless there is obvious advantage and benefit from the overhaul that worth your time and effort.

3. Keep it simple, stupid

No comment yet

I have been offered many times to sell my mechanical day trading systems. I turned down almost all of them. Most of these potential clients are retail traders. I rejected them not because of the amount of money offered is not good enough. The problem is that I feel obligated to support them as my clients until they are proficient in using the trading models. I just do not have the time to commit to the idea at this point in time.

I work with hedge funds and professional traders for a long time and I still have trading models licensed to some. It is much easier to communicate the risk factor and other complex details with them comparing to retail traders who do not see mechanical trading quite the same way. Hopefully this article will give everyone a general idea where the hurdles are. More importantly, I like to explain what it takes to operate your own mechanical trading models.

What You Think You Know May Not Be Precise Enough

Many people think they can find patterns in the charts. Then they start to see the ones they recognize with recurring outcomes. This is the normal process of pattern recognition by human. We learn by studying the example cases. It is also the necessary step for any trader to train themselves into proper chart reading. This process, however, does not work well in defining trading rules suitable for mechanical models.

Patterns not exact enough to be described mathematically or programmatically will give you way more matches in the historical data than you think. The ones you think that should match may not even be included in the set of matched cases if you cannot provide a definition that fully describe what you think it should be. Too many rules will produce next to no matches and exclude the ones you think that should be recognized. Too few rules may result in a set of matches that you deem incorrect.

Being computer programming savvy does not necessary help in this situation because it is the ability to describe the pattern in an abstract way that really gives you the precise definition necessary. Programming capability is just one of the entrance level requirements. It is not enough to solve your problem if you are looking for recurring patterns to produce profitable trading models.

In short, you have to have an understanding of the data with some kind of framework. It can be statistically driven. It can also be chart structure driven. If the framework has merits, you will be able to produce trading rules based on the framework with exploitable edges.

Risk Management In Mechanical Models

One of the area where mechanical models outshine strategies designed for human is that they can go real detail in terms of risk management. For human traders, it is better to keep the money management rules as simple as possible because humans in general are not good at micro-managing things. Nudging a position continuously is very difficult to do and takes a lot of mind power.

For example, good day traders often demonstrate the ability to scale out of a position at certain targets and then being able to add back the size taken out later when the market pullback. Great traders, however, can add the position back on even when there is no pullback. Within seconds should the scenario changed to warrant full size position, they can jump in without hesitation. They can also go flat within a moment when they detect the signs of danger. Normal people simply cannot change their minds fast enough to deal with the subtle changes in the market.

Mechanical models, or bots, do not have this problem. If you can isolate the conditions for all out melt up (or melt down), you can make your bot pile on. When the scenario started to develop against your position, as human we tend to wait a bit longer while the bot you created will cut the position off as soon as the criteria to bail out is in place. Your mechanical counter-part does not have the psychological burden you have.

Bots are also better at dealing with complex sideway / congestion market environment. They can go microscopic and making scalps faster than you can enter the orders. For humans doing the same, they would probably be able to do it for a few days but their performance will deteriorate very quickly when their minds are exhausted from the activity.

As a summary, high-end mechanical models with complex money management algorithm can often do better than human traders including their creators. The way these bots operate are not compatible with normal people psyche. If you cannot understand the bot you deployed, you will have huge difficulties in managing the bot when it is risking your money everyday. It can be extremely stressful if you cannot trust your bot doing its job.

First Step – Find An Edge That You Can Understand

Tradable edge exists in many forms. It is not necessarily something complex. It can be something as simple as an observation you find interesting and recurring often enough that you think it has an edge. It can also be an idea or trading setup you find from other sources like a book on trading or a discussion online. As long as it is something you can understand, it will give you a good starting point to work from.

Then you have to go through the process of identification and clarification of your observation. The goal is to define the recurring pattern with rules that are precise enough to allow you to conduct backtesting. You want to gather objective measures of the outcomes from the pattern so that you can tell if it is an exploitable pattern where you can profit from the pattern with controlled risk.

Backtesting can be done by hand or by code. Both method works. Depending on the type of patterns you are trying to gather information, sometimes it is better to go with manual checking if you are very familiar with the historical charts yourself. It gives you way better feedback as you can visually memorize the setup from the charts and learn about the possible variations. It helps you in gaining insights into the pattern.

Doing backtesting with code will force you to come up with precise rules to describe the pattern. It can be difficult as many concepts may require you to built complex algorithms to identify them. It can be many layers of concepts to be tackled first before you can work on the pattern you originally come up with. The advantage, of course, is that the pattern you identified this way can be eventually incorporated into bots.

Second Step – Keep The Money Management Rules Close To What You Do Manually

Do not start your mechanical models with money management rules that are too complex to follow by yourself. You will have a hard time tracking the behaviour of your model if you have difficulty in recognizing the trades it takes. It takes time to train yourself into understanding what the bots are doing.

It is like giving instructions on a subject you are pretty good at. Although you may know the subject well, giving instructions about the subject to someone else is a completely different story. Your ability to tell if someone else is leaning from your instructions adds another layer of difficulty. Until you learn to teach efficiently and able to evaluate the results objectively, you will have a hard time teaching the more advanced topics.

Over time, you will be able to add more complex money management rules to the bots that will not intimidate yourself.

Third Step – Learn The Nuisance Of Operating A Bot

It should be pretty obvious now why I think it is a bad idea for inexperienced traders to trade complex mechanical systems. It takes time to learn to trade mechanically, just like learning to trade discretionarily.

By starting out with bots (or rigorous manual strategy) having simpler rules, you will gain the necessary experience before moving onto more complex models. It is also a good idea to observe simple bots working in real-time. It does not matter whether the bot is playing with real money or not because at this point the goal is not to make money with bots yet.

The goal is to train you in becoming a proficient mechanical trader or bot operator.

You need to learn to handle all the details of having a bot doing its job. You need to know how to check if the bot is operating correctly. You also need to know if it is not functioning properly, how to take over. You need a trading plan built around the model you have created with clear guidelines on capital commitment, disaster control where you would pull the plug and stop the bot from trading, etc.

Summary

Comparing to discretionary trading, which you have to be your own coach and psychologist while you are performing like a sportsman, mechanical trading turns you into the manager and mechanic of your racing team while the bots compete. Although the roles we play are different, the objective of making money from the markets stays the same. They are just different approaches to tackle the same issue.

One comment

As I mentioned last time, Ninja is bullish for a run higher.

Here is an update to the setup.

Green arrow marks the perfect retest of the quarterly STOPD level from 1st quarter high. This happens often when previous test was a false breakout.

As of today price has traded above red down trend line triggering a weekly level breakout to the upside.

Light green line is a copy of the green line projection to realign the projected time window to get to the target.

Light green target zone is the projected price target for this move.

Red zone should not be challenged again if this move higher is for real.

No comment yet

Very interesting night with Nikkei gapped up and zoomed higher to +120 points.

Due to the fact that it is still an ongoing crisis with Syria I suspect such wild swings will continue for some time.

No comment yet