[Edit: 8/17/2011 - Once again, Greg has made my "Grouping of Trades and Tags" section below obsolete. Please see the comments section below or his blog post here!]
I finally completed updating the tags of all my historical trades in the free online trading journal at Tradervue.com last week. I went back over 4 months of trades since I started day trading the beginning of April 2011, and am now in the process of cleaning up the tags. With nearly 700 journal entries accounting for over 1000 trades, I know there is likely some tedious cleanup work remaining.
However, there's a point of diminishing returns and based on my requirements, I don't intend to or need to have perfect data. Just something good enough for relative comparisons and insights. It's not like I'm using this data to file my taxes. My primary job is a trader, not a statistician.
Some caveats on the data and results
- Data quality - garbage in/garbage out? My data is not perfect as described above, but I believe it's more than good enough to provide useful and valuable information. To any other trader, my database is likely of very limited value, since it's not a reflection of who they are and their actions. But to me, this data is priceless, it is me.
- Selection bias in the data. In other words, this database is a reflection of my real life ability to see and execute the setups, for better or for worse. So the accuracy or average profit of a particular setup in my journal does not reflect the optimal outcome of that setup -- again, only my ability to see and execute it properly (which hasn't been consistently great). In the future as I improve, the nature of this dynamic database will shift and grow with me.
- For example, there are instances where I took a setup that was grade "B" and would be considered profitable based on backtesting. But in real life, I totally chased it, executed it poorly, and possibly lost money. So that particular B grade trade will show up and be calculated on my reports as if it was a losing setup. It's not the fault of the setup, I just botched it. As my execution improves, this will become less of a factor.
- Grouping of trades and tags. Tradervue currently groups trades together into a single record if they occur within a certain timeframe. Therefore, if you're scalping (or in my case, overtrading) a stock, it's likely that you'll have multiple trade pairs in one journal record. As a result, the tags in that one particular journal record might contain tags of multiple setups, grades, outcomes, etc.
- This is not a problem if I'm just reading my journal entries, but when I query tags for certain statistics, I will not be getting the most accurate results. A query for A grade trades might also include D grade trades, and the average P&L outcome of all the trades in the group will be included in both queries. However, these calculations seem to wash out as the sample size grows.
- Reports filter one tag at a time. The ability to query tags on Tradervue is currently setup to filter based on one tag at a time. So there are some instances where I will bring the data into Excel or Access for more sophisticated queries. In those instances, there may be some results I discuss that you may not be able to easily reproduce from Tradervue alone, at least at this time. Let me know if you would like to know more details on how I do this and I can write a future post.
- Gross P&L only. Although this might change in the future, the figures from Tradervue are currently gross profit/loss, and therefore exclude commissions. For longer term traders with a minimal number of trades, this is not a big deal. However, I have over 1000 trades over 4 months. And I am essentially a scratch trader according to gross P&L (-2R), but include commissions and all I can say about my net P&L is...OUCH!
Based on R-Multiples
In order to keep everything on a relative basis, I have converted the figures into R-Multiples. Trader Mike posted an article on R-Multiples back in 2006, and this concept works best for me as a day trader. I'm currently risking a very conservative 25 to 40 basis points (.25% to .40%) of my portfolio per trade, since the current name of the game is to survival while I learn. So I'll use a weighted average of 30 basis points in my calculations as my current R. In other words:
R = .003 * my portfolio valueAs I become more consistently profitable in the future, I will raise the R-Multiple and balance it with the level of draw down I will eventually be comfortable with. And I will likely consider incorporating tiers, so that if my level of conviction is high for a particular trade (e.g. A+ setups), then I will establish the trade with something like a 2R risk.
The shocking...but really, not so shocking results
Reminder, see caveats earlier regarding the data...think relative not absolute regarding these results. Based on how I tagged my trades in Tradervue, here are the results to date:
GRADE # RECORDS % PROFITABLE TOTAL R AVG R PER TRADE
A 24 92% +61R +2.54R
B 182 63% +167R +.91R
C 206 33% -32R -.16R
D 483 16% -141R -.89R
F 171 25% -70R -1.42R
Note: As mentioned in the caveats, a trading record may contain more than one paired trade, and therefore more than one "grade." So the sum of all records, as well as the sum total R, will not square with my actual figures to date.
TYPE # RECORDS % PROFITABLE TOTAL R AVG R PER TRADE
Revenge 475 20% -136R -1.12R
Rogue 87 22% -68R -.78R
I could likely just stop here with my "no-duh" analysis to become a consistently profitable trader -- simply take only A and B setups. Only taking those trades would have resulted in 206 trades (~12 trades/week) being 65% accurate with +228R gross profits over 4 months. Very respectable. With better trade selection and execution, I have no doubt taking only A and B setups in the future will result in 70+% accuracy with even greater average profits (and risk/reward) per trade.
Taking C, D and F setups combined resulted in -243R. Now this part, I admit, I did find shocking. Taking these setups actually lost more than setups A and B combined. And if commissions were included, then the gap is even greater. I have to remember that taking lower quality setups, even if I risk "only" 30 basis points, are toxic to my P&L. Death by a million paper cuts.
- When reviewing my A & B trades, it's really quite a confidence booster. There's no question I have the ability to find good setups and to execute properly....
- ....then why the heck am I taking all those C, and especially D and F trades?!? I've got some serious work to do. I'm upset by this, and I'm going to channel these emotions to further fuel positive change.
- Only taking A & B quality setups are truly good to your P&L's health
- Taking C, D and F setups produce losses greater than profits from A and B combined
- Revenge trades don't work. Surprise...Not! But I am surprised at the level of punishment these trades had on my P&L
- Rouge trades (trades that are not part of my defined strategy) don't work as an aggregate. Surprise...Not!
- Everything seems to rank order well, so my initial grade assignments worked relatively well
- However, the ranking of C setups might need adjusting. I was expecting about 40-50% accuracy and breakeven gross profits. Therefore, perhaps some grade C trades should actually be rated D.
- I was disgusted with nearly all D and F trades, so I wasn't as careful grading between them.
- My risk management is solid. I don't mess with my stops nor do I risk too much on any one trade. But I do have a problem with stops being too close when I adjust them.
I really need to understand those setups that feel nearly effortless and produce great results, then really focus on them. I also need to learn to avoid, or at least minimize, all the others. In other words, really focus on my strengths, and let them completely overpower my weaknesses.
And perhaps I also need to better understand why I occasionally fall into the overtrading/rogue mode, learn to lookout for those triggers, then smash those patterns. Or who knows, maybe I should embrace this by evaluating other trading strategies and methods that better fit with my natural tendencies and strengths?
In the days and weeks ahead, while making sure I avoid the analysis paralysis trap, I'll continue to dig into the data for more advanced insights, and especially determine which setups have worked best for me. My work has only just begun.