Three questions about trading


Brett Steenbarger wrote Three Good Questions to Ask About Your Trading in TraderFeed yesterday.

A good snippet from his is preamble "when profits are good, ask yourself what can you do to make them outstanding."

The 3 questions are:

1. How many distinct setups do you trade?

As with the Correlation of Market Movements your signals may correlate. Are you signals for distinct setups based on the same math? If so, and they're providing the same signals, then they should go. For example, you may find that the underlying math of several oscillators is similar enough that you only need 1 or 2 of them.

2. What are the outcomes of your largest vs. smallest trades?

Are you allocating for risk correctly?

3. Is the distribution of your long and short trades consistent with the market's direction at the next largest time frame?

This is another correlation question. If you break the market up into time segments and label them as up, down, and sideways, you can then perform a correlation analysis using that data against the trades that you took to see if you are trading with or against the prevailing market trends.

Let me expand on this last point and give you the tools. Using Excel and the =CORREL() function you can find out if your trading bias ties in with the market trend. You can do this on any time frame. I would suggest using 1's, 0's, and -1's, in two columns in your spreadsheet to represent longs, flat, and shorts.

Let's say that you're a day trader and you wanted to enter the data for 1 day's worth of trading. The market trends up in the morning, sideways during the middle of the day and down in the late session. In each of these sessions you open and close a long trade followed by a short trade. So your data looks like this:
Market	You
1 1
1 -1
0 1
0 -1
-1 1
-1 -1

If you perform a correlation on this data you will see that it's zero. This means that your trading has no correlation to the market direction.

I will attach a spreadsheet to this posting that will allow you to enter figures and compare market direction with your actual trading direction.
I have not entered a date/time column because you can use it on any time frame. You can easily enter a date/time column if that will help you enter your data. Insert it to the left.

Try changing the values in the "Your Trade Direction" column and watch how the correlation figure changes.

Click link to access uploaded file:
TradeMarketCorrelation.xls
This broader concept you are touching on here is quite important, DT. When I saw the other thread I wanted to add my two cents in (you know, always trying to bring up food for thought that tends to make things more complex, yet perhaps more relevent, and tends to frustrate people:-), but I'm always strapped for time.

Anyway, I do a lot of what I call 'intermarket dynamics', where I try to assess the money flow from one area to another. A major 'asset allocation swap' is, in my opinion, a large version of what I try to get a handle on. A lot of the time I feel money just flows, and I try to see from where to where, to assist the formation of my trade premises (I am not a 'scalper' for ticks, so this is useful for me). So, what has this to do with this topic?

Let's say the ZN doesn't correlate real high with one of the market indices, as an example. I feel there is a much higher correlation than a simple straight up, all the time correlation. I think it generally can run in 'lock-step', or it can run in 'inverse lock-step', to use the terms of the talking heads on TV. When it is in one of these modes I feel the correlation is much higher.

If rates are going down (hence the ZN up), the market likes that, and it goes up, generally. They are positively correlated at this point. But say, as an example, something negative happens geopolitically, and the market starts to tank. Does the ZN go down, and hence rates up? No, generally there is a 'flight to quality' and they buy treasuries. This also goes with the idea that if the market is tanking on a big concern, the Fed may step in and lower rates. So in this case there is a negative correlation. In both cases the correlation is high. Add them up and look at both times, and like your example, the correlation might show zero. I feel this also applies in more everyday money flow scenarios.

I address this simply with experience in reading and assessing money flow. As a 100% discretionary trader (albeit one with a very comprehensive and well-defined methodology), I can't 'back-test' or code anything I do, and I can't come up with any way to separate out the money flows on a quantitiative basis. I have to assess every day and situation as a separate entity. This is like DT's comments on each holiday time being unique. I find every setup every day to have its own unique character, perhaps like every hand in poker. Same game, same rules, but a different layout and opponents, each in a unique frame of mind at that time.

I think what I am getting at is that I feel there is a lot more correlation than most think, but it is much more complex than simply 'this market correlates with this other market', right across the board. The intraday mini example really helps show my point, in that the three standard periods we all know about are so different that I feel they must be assessed separately, and even that may be too simple of a model.

I hope these ramblings bring something worthwhile to the discussion. I'm never sure when I post a specific idea that what I present that is of value to me and my 'Trading Plan' will be perceived as potentially useful to others.
Thanks Jim, your comments are always good.

Even though you're 100% discretionary with your trading, I believe that you could still measure many aspects of it. You would of course need to quantify your decisions into compartments and give each a number.

For example, you could rate your mood on a scale of 1 to 10. You could also rate your alertness, how well you slept the previous night, family issues etc. on similar scales.

You could also rate your decisions for entry a trade. For example, rate the percentage of the decision that was feeling and the percentage that was technical that triggered your entry and record that with each trade along with a number of other metrics that were available at trade entry. You would of course store the trade result in the same row with this data. You don't need to store the profit but can also give the trade a scale from (say) 0 (being the worst outcome) to 5 (being the best outcome).

Once you have enough of this data, you can run a correlation of each input data against your trades' results and you can see which most closely correlates to the results. This would allow you to analyze both emotional (intangible) aspects as well as objective technical aspects against your trading results and give you an insight to which is working better/worse.

I think I'm also starting to ramble now :)