We've been getting a lot of inquiries from equities traders interested in the TraderDNA Analyzer. Currently, the architecture of our program facilitates futures trading only. That said, TraderDNA accepts data from all products traded on the major continental futures exchanges (CBOT, CME GLOBEX NYMEX/COMEX, ICE, Eurex and Euronext/Liffe). All of our users trade products listed on one or more of these exchanges, and we will continue to add compatibility to additional exchanges as we deem necessary.
The purpose of this post is to discuss--at least briefly--what can be done manually, in lieu of using a program that will perform automated calculations and visual representations of the data, such as TraderDNA.
In my last post I introduced the concept of tracking certain external factors that could, naturally, affect the performance and thus the results of someone who spends at least part of the day in front of the markets. This is not complicated and can be done quite easily.
First you have to decide what exactly you'd like to track. I'll suggest a few factors that might be interesting.
How hungry are you at the beginning and end of the trading day?
Hunger can have a conscious or subconscious affect on your decision-making process (I will soon support this assumption and those following with empirical research as an addendum to this post).
Consider a situation in which a part of you wants to go eat and a part of you wants to stay at your desk and trade. You have no food within arm's reach of your desk and so you decide to stay in front of your screen. How does this affect your trading?
I liken this to being at a poker table and realizing that you are very hungry and thinking of something you'd really like to eat. All of a sudden part of you wants to go eat, and another part of you wants to stay and play. So you loosen your play thinking, perhaps subconsciously, that there is now something good that could come of losing your stack (the ability to go eat). You may have played diligently and followed a rigid plan for 2 hours but all of a sudden you are willing to stray from that plan and play atypically because of the perceived benefit that could come from losing.
Taking note of your level of hunger at the start and immediately at the end of the day could provide some interesting information if a correlation is made between level of hunger and results at the start and end of the day.
How tired are you at the beginning and end of the trading day?
Similar to hunger, being tired or having even the slightest desire to rest can affect how you trade and how you break the parameters of your system and decision-making.
How eager, excited, or interested are you to trade today?
This is something that could be taken note of at the start of the day. Consider that you have a survey of questions you fill out at the start of each day and this is one of the questions. The values might be "Not Eager", "Somewhat Eager", "Very Eager".
How does your level of eagerness or interest to trade affect the overall result of the day? It can be hard not to trade everyday if this is your living, but if you can make a correlation between your level of interest and the ensuing results of the day then the possibility exists that you could sit certain days out altogether so as to avoid losses to your trading account. You might find that you perform best when you show a lack of eagerness at the start, or vice versa.
Knowing ahead of time with at least some confidence that the coming day will be a loss can help you to dodge this loss or to determine if it's a good day to trade or not. You might not make this decision solely on eagerness/willingness to trade but after assessing other additional factors as well, such as your emotional well-being.
How do you feel about the state of your trading account right now?
This is another factor that can be tracked at the beginning of each day. Did you suffer a big loss on the previous day? Are you upset with the recent deterioration of your trading capital? Of course it can be difficult to be totally objective can be very hard to be objective with yourself in answering such questions but the closer you can come to giving honest/accurate answers, the better you can direct your trading. For example, would it be beneficial to sit out for at least some period of time?
Are you particularly frustrated, disappointed, or angry with a girlfriend, spouse, friend, or family member right now?
Select an answer to this question before you start trading for the day and track your answers as they relate to results of the day. Your choices might be as follows:
□ At peace with everyone right now
□ Somewhat frustrated or disappointed with one or more person(s) right now
□ Somewhat angry with one or more person(s) right now
□ Very frustrated or disappointed with one or more person(s) right now
□ Very angry with one or more person(s) right now.
What opinion about today's market direction (if any) do you have?
Does your initial opinion of the day's market direction or sentiment affect how you trade, and your results?
□ Very Bearish
□ Bearish
□ No opinion
□ Bullish
□ Very Bullish
What opinion about today's volume (if any) do you have?
Are you expecting there to be little volume today and anticipating that it might not be the best day to trade a particular market? How does your anticipation of action affect your involvement and results in that market?
____________________________________________________
Additionally, answering some questions at the conclusion of the day will help you take note of what type of day you've had. Reviewing cumulative results will help you assess on which types of days you trade best. ie, was it an up or down day, or was there a lot of chop in the market today?
If you can create standard answers (values) to the above questions and check the appropriate answers, those answers can become additional trackable factors that you can tag to each trading day. Currently you might track P/L, number of trades, contracts, etc. each day. Now you might track P/L, number of trades/contracts, starting/ending hunger level, starting/ending need for sleep, market direction, opinion of market direction, etc...
You don't need to be proficient with Excel or know VBA to keep tabs of the values in Excel, and calculations can be done manually after at least a couple weeks of tracking.
Use a different column in Excel for each question you ask yourself, as well as for P/L, number of trades, market direction, etc.
© Copyright 2007 David Adler
All rights reserved
All analysis generated with the TraderDNA Analyzer.
Monday, December 3, 2007
Monday, October 22, 2007
Performance Metrics: Digging Deeper than P/L
We've developed a variety of complex measures that we've incorporated into TraderDNA. Similarly to your trading software capturing and keeping track of your P/L while you trade, our technology is able to capture data elements with which we use to calculate complex measures that help to paint the entire picture of how you or your system traded. These measures can be grouped into 8 categories which encompass the areas that we believe have the greatest, most direct, and most profound bearing on an outright trader's P/L.
Drift
Drift metrics are useful in determining how much risk a trader takes in open positions, and conversely how much profit potential the trader sees in open positions. If a trader relies upon the level of a his P&L alone as an indication of how well he or she is doing, he misses a vital statistic that indicates how many negative or positive ticks his open positions experience before they are closed. An example will clarify the uses of drift metrics.
If a trader takes a long position at 100 and then allows the position to lose several ticks to 95 before closing it out for a small profit at 102, in P&L terms the trader has performed averagely well. However, the negative 5 ticks that the trader allowed the position to drift is reflected in the Negative Drift value.
On the positive side, if a trader takes a long position at 100 and the market trades up to 108 and then soon after the trader covers his position at 105, the entire profit of 8 ticks that was available is expressed in Positive Drift.
Negative Drift: Amount of risk taken in open positions
Positive Drift: Amount of profit (either realized or unrealized) available in open positions.
_______________________________________________________
Lost Opportunity
The Lost Opportunity measure is useful in determining how much profit was left on the table. Lost Opportunity Drift is calculated by taking the difference in the Positive Drift of a position and the realized P/L, or where the position was covered.
As an example, a trader takes long position at 100 and the market trades up to 108. Soon after the trader closes the position for a profit at 105. The Lost Opportunity value is 3 ticks (108-105).
Lost Opportunity Drift: Amount of unrealized profit that was lost in a position.
___________________________________________________
Shape of Trade
The Shape of Trade is a very powerful visual representation of 7 primitive measurements, namely:
Average Negative Drift
Average Positive Drift
Average Lost Opportunity
Average P/L
Average Time to Negative Drift
Average Time to Positive Drift
Average Time in Trade
___________________________________________________
Consecutive Winners, Losers, Scratches
Being able to determine when winning and losing streaks occur, the size of the streaks and under which circumstances they occur is of significant value.
Consecutive Winners: The number of consecutive, or back-to-back winning trades.
Consecutive Losers: The number of consecutive, or back-to-back losing trades.
Consecutive Scratches: The number of consecutive, or back-to-back scratched trades.
_____________________________________________________
Stacking
Stacking refers to a measurement that determines how well a trader recognizes a potentially good trade, having placed an order in the marketplace and supported his decision by increasing leverage and exposure by adding additional orders in the same format.
Conversely, negatively stacking measures if, when, and to what extent the trader is leveraging his losses and adding to a losing position.
Negative StackUp: The number of instances in which the trader added to a losing position.
Positive StackUp: The number of instances in which the trader added to a winning position.
Negative StackUp Quantity: The number of contracts the trader added to losing positions.
Positive StackUp Quantity: The number of contracts the trader added to winning positions.
_______________________________________________
Time Measures and Intervals
We provide various time-related measures.
Time in Losers: Time interval taken to accept a loss.
Time in Winners: Time interval taken to take a profit.
Time in Scratches: Time interval taken to scratch a trade.
Time Since Last Trade: Time between closing a position and opening a new position.
Time Since Last Loss: Time between closing a losing trade and initiating a new position.
Time Since Last Win: Time between closing a winning trade and initiating a new position.
Time Since Last Scratch: Time between closing scratched trade and initiating a new position.
Time to Positive Drift: Time taken to reach the best possible profit point in the trade.
Time to Negative Drift: Time taken to reach the maximum point of risk in the trade, or the worst possible loss amount in the trade.
______________________________________________
Quantity
Quantity reflects either the number of trades, the number of round turns, or the number of contracts traded, depending on which measure is desired. Additionally, quantity values can be isolated by winning trades, losing trades, and scratched trades.
Quantity: Number of trades (defined by round turns)
Win Quantity: Number of winning trades
Loss Quantity: Number of losing trades
Scratch Quantity: Number of scratched trades
Quantity (contracts): Number of contracts traded
Win Quantity (contracts): Number of winning contracts
Loss Quantity (contracts): Number of losing contracts
Scratch Quantity (contracts): Number of scratched contracts
_______________________________________________
Profit/Loss
TraderDNA provides the flexibility to look at an overall P/L metric, as well as more specific P/L metrics. The ability to split P/L earned/lost by losing and winning trades is an important function in examining the size of winning and losing trades, side-by-side.
P/L: Amount of profit or loss
Win P/L: Amount of profit in winning trades
Loss P/L: Amount of loss in losing trades
P/L Per Contract: Amount of profit or loss, per contract traded
© Copyright 2007 David Adler
All rights reserved
All analysis generated with the TraderDNA Analyzer.
Drift
Drift metrics are useful in determining how much risk a trader takes in open positions, and conversely how much profit potential the trader sees in open positions. If a trader relies upon the level of a his P&L alone as an indication of how well he or she is doing, he misses a vital statistic that indicates how many negative or positive ticks his open positions experience before they are closed. An example will clarify the uses of drift metrics.
If a trader takes a long position at 100 and then allows the position to lose several ticks to 95 before closing it out for a small profit at 102, in P&L terms the trader has performed averagely well. However, the negative 5 ticks that the trader allowed the position to drift is reflected in the Negative Drift value.
On the positive side, if a trader takes a long position at 100 and the market trades up to 108 and then soon after the trader covers his position at 105, the entire profit of 8 ticks that was available is expressed in Positive Drift.
Negative Drift: Amount of risk taken in open positions
Positive Drift: Amount of profit (either realized or unrealized) available in open positions.
_______________________________________________________
Lost Opportunity
The Lost Opportunity measure is useful in determining how much profit was left on the table. Lost Opportunity Drift is calculated by taking the difference in the Positive Drift of a position and the realized P/L, or where the position was covered.
As an example, a trader takes long position at 100 and the market trades up to 108. Soon after the trader closes the position for a profit at 105. The Lost Opportunity value is 3 ticks (108-105).
Lost Opportunity Drift: Amount of unrealized profit that was lost in a position.
___________________________________________________
Shape of Trade
The Shape of Trade is a very powerful visual representation of 7 primitive measurements, namely:
Average Negative Drift
Average Positive Drift
Average Lost Opportunity
Average P/L
Average Time to Negative Drift
Average Time to Positive Drift
Average Time in Trade
___________________________________________________
Consecutive Winners, Losers, Scratches
Being able to determine when winning and losing streaks occur, the size of the streaks and under which circumstances they occur is of significant value.
Consecutive Winners: The number of consecutive, or back-to-back winning trades.
Consecutive Losers: The number of consecutive, or back-to-back losing trades.
Consecutive Scratches: The number of consecutive, or back-to-back scratched trades.
_____________________________________________________
Stacking
Stacking refers to a measurement that determines how well a trader recognizes a potentially good trade, having placed an order in the marketplace and supported his decision by increasing leverage and exposure by adding additional orders in the same format.
Conversely, negatively stacking measures if, when, and to what extent the trader is leveraging his losses and adding to a losing position.
Negative StackUp: The number of instances in which the trader added to a losing position.
Positive StackUp: The number of instances in which the trader added to a winning position.
Negative StackUp Quantity: The number of contracts the trader added to losing positions.
Positive StackUp Quantity: The number of contracts the trader added to winning positions.
_______________________________________________
Time Measures and Intervals
We provide various time-related measures.
Time in Losers: Time interval taken to accept a loss.
Time in Winners: Time interval taken to take a profit.
Time in Scratches: Time interval taken to scratch a trade.
Time Since Last Trade: Time between closing a position and opening a new position.
Time Since Last Loss: Time between closing a losing trade and initiating a new position.
Time Since Last Win: Time between closing a winning trade and initiating a new position.
Time Since Last Scratch: Time between closing scratched trade and initiating a new position.
Time to Positive Drift: Time taken to reach the best possible profit point in the trade.
Time to Negative Drift: Time taken to reach the maximum point of risk in the trade, or the worst possible loss amount in the trade.
______________________________________________
Quantity
Quantity reflects either the number of trades, the number of round turns, or the number of contracts traded, depending on which measure is desired. Additionally, quantity values can be isolated by winning trades, losing trades, and scratched trades.
Quantity: Number of trades (defined by round turns)
Win Quantity: Number of winning trades
Loss Quantity: Number of losing trades
Scratch Quantity: Number of scratched trades
Quantity (contracts): Number of contracts traded
Win Quantity (contracts): Number of winning contracts
Loss Quantity (contracts): Number of losing contracts
Scratch Quantity (contracts): Number of scratched contracts
_______________________________________________
Profit/Loss
TraderDNA provides the flexibility to look at an overall P/L metric, as well as more specific P/L metrics. The ability to split P/L earned/lost by losing and winning trades is an important function in examining the size of winning and losing trades, side-by-side.
P/L: Amount of profit or loss
Win P/L: Amount of profit in winning trades
Loss P/L: Amount of loss in losing trades
P/L Per Contract: Amount of profit or loss, per contract traded
© Copyright 2007 David Adler
All rights reserved
All analysis generated with the TraderDNA Analyzer.
Monday, October 1, 2007
Explaining Your P/L
When assessing the performance of a trading system, the benchmark by which we most often measure the efficacy is P/L. This makes perfect sense and P/L is the first measure we should consider. It should be used as a starting point though, from which to drill deeper in order to discover the reasons behind the P/L value(s).
If we use P/L alone as an indication and summation of what has happened, we will only know so much. Yes, we will have the big picture, but we will be missing the more important information: what specifically formed the big picture; what made my P/L what it was. An example will clarify this...
Now I can reach a more detailed conclusion that my trading was poor in the S&P and Brent, but was profitable in the other 3 markets I trade. This conclusion, however, is still too broad. I want to know specifically why my trading was poor. If I have been better at trading the Crude than I've been at trading the S&P, I want to be able to explain that difference. What did I doing differently in these markets? WHY have I performed worse in the S&P?
© Copyright 2007 David Adler
All rights reserved
All analysis generated with the TraderDNA Analyzer.
If we use P/L alone as an indication and summation of what has happened, we will only know so much. Yes, we will have the big picture, but we will be missing the more important information: what specifically formed the big picture; what made my P/L what it was. An example will clarify this...
This represents the total amount (in USD) won or lost over a period of time (cumulative P/L). Judging from this number alone, you might deem my trading unsuccessful since, afterall, I lost $8,430.00. The goal though, should be to draw specific, rather than broad, conclusions. There are countless ways I can explore this number by splitting it up and dissecting it. For example, I trade 5 markets...
Now I can reach a more detailed conclusion that my trading was poor in the S&P and Brent, but was profitable in the other 3 markets I trade. This conclusion, however, is still too broad. I want to know specifically why my trading was poor. If I have been better at trading the Crude than I've been at trading the S&P, I want to be able to explain that difference. What did I doing differently in these markets? WHY have I performed worse in the S&P?
This is where P/L can only take us so far. It is important to discover the cause for these P/L values in the first column above.
We've developed a variety of complex measures that we've incorporated into TraderDNA. Similarly to your trading software capturing and keeping track of your P/L while you trade, our technology captures data elements from which we calculate exotic measures that help to paint the entire picture of how you or your system traded. My next post will describe these measures in more detail. For now, I'll show you how a few of them might be used with the example above.
Here I'm finding out more about my trading in various markets. I can use the first two columns as a reference when I'm interpreting the last 4 columns. The third column (Neg. Stackup) shows me the number of times I added to losing positions. The 4th and 5th columns (Neg. Drift and Pos. Drift), show me how much risk I'm taking on my average trade and how much profit opportunity I'm seeing in my average trade. The last column (Lost Opportunity Drift) shows me how much money I am leaving on the table when I have opportunities to take profit.
Here are 4 more measures: average time in losing trades, average time in winning trades, average number of consecutive winners, and average number of consectutive losers. These charts help explain more of the big picture and help to refine the conclusions we came to above.
© Copyright 2007 David Adler
All rights reserved
All analysis generated with the TraderDNA Analyzer.
Monday, September 10, 2007
Charting Up Your P/L and Taking a Critical Look at Your Equity Curve
There's valuable insight to be gained by charting up your cumulative P/L over a period of time. You may or may not already do this.
How accurate a picture do you have of the last 3 months of your trading? Do you remember when your slumps occured? What about your really good days or weeks? Do you recall a week or two in which you were up and down the whole time? Do you know whether you've been up, down, or have hoverred around baseline over the past year?
The manual way to do this is to gather your P/L statements from each trading day, and run a cumulative calculation of each day. So if you made 200.00 the first day and the next day lost 350.00, your net cumulative P/L for the two days would be -150.00.
The chart below represents about 3 months of trading. The x-axis represents the date and the y-axis represents cumulative P/L. Each dot represents a trade. Note that on days with small P/L ranges, many dots overlap eachother.
This trader is doing something right. For one, he has been cumulatively profitable after trading every day for the past three months (9/1/06 to 12/7/06). He steadily makes money and has no swings. We can tell from the chart that he has handled his drawdowns well and recovered from every single one of them (highlighted in yellow).
Until October 5th, his trading was generally break-even. I want to know what changed on October 5th. I might even run analysis on the two weeks prior to 10/5 and the two weeks after 10/5 and look for the significant differences so I can pinpoint exactly what changed (other than P/L). I also want to know what happened on November 14th, the day he made the most profit in the shortest amount of time.
Here's a different example...
And another one...
This trader got off to a poor start and had an incredible improvement. Although, looking at the entire picture he's taking form as someone who's had a few relatively substantial swings in P/L.
Assuming the market dynamic didn't dramatically change during this period, we can conclude that his trading or his emotions, or possibly both, are at fault. Just for clarification, by "market dynamic" I am referring to how a particular market trades---namely the depth of market (# of bids/offers at each price increment), the volatility, and the average daily range. This guy happened to be trading the YM, ER, and ES, so...
There's one last thing I noticed with this graph. He seems to have either conscously or subconsciously identified levels of support, in terms of his P/L track. I drew red lines in these places. In both instances, when he broke below that level, things got worse at an accelerated pace.
In summary, this is a conceptually simple graph that can tell us a lot.
© Copyright 2007 David Adler
All rights reserved
All analysis generated with the TraderDNA Analyzer.
How accurate a picture do you have of the last 3 months of your trading? Do you remember when your slumps occured? What about your really good days or weeks? Do you recall a week or two in which you were up and down the whole time? Do you know whether you've been up, down, or have hoverred around baseline over the past year?
The manual way to do this is to gather your P/L statements from each trading day, and run a cumulative calculation of each day. So if you made 200.00 the first day and the next day lost 350.00, your net cumulative P/L for the two days would be -150.00.
The chart below represents about 3 months of trading. The x-axis represents the date and the y-axis represents cumulative P/L. Each dot represents a trade. Note that on days with small P/L ranges, many dots overlap eachother.
This trader is doing something right. For one, he has been cumulatively profitable after trading every day for the past three months (9/1/06 to 12/7/06). He steadily makes money and has no swings. We can tell from the chart that he has handled his drawdowns well and recovered from every single one of them (highlighted in yellow).
Until October 5th, his trading was generally break-even. I want to know what changed on October 5th. I might even run analysis on the two weeks prior to 10/5 and the two weeks after 10/5 and look for the significant differences so I can pinpoint exactly what changed (other than P/L). I also want to know what happened on November 14th, the day he made the most profit in the shortest amount of time.
Here's a different example...
And another one...
This trader got off to a poor start and had an incredible improvement. Although, looking at the entire picture he's taking form as someone who's had a few relatively substantial swings in P/L.
Assuming the market dynamic didn't dramatically change during this period, we can conclude that his trading or his emotions, or possibly both, are at fault. Just for clarification, by "market dynamic" I am referring to how a particular market trades---namely the depth of market (# of bids/offers at each price increment), the volatility, and the average daily range. This guy happened to be trading the YM, ER, and ES, so...
There's one last thing I noticed with this graph. He seems to have either conscously or subconsciously identified levels of support, in terms of his P/L track. I drew red lines in these places. In both instances, when he broke below that level, things got worse at an accelerated pace.
In summary, this is a conceptually simple graph that can tell us a lot.
© Copyright 2007 David Adler
All rights reserved
All analysis generated with the TraderDNA Analyzer.
Monday, August 20, 2007
How a Poor Start Can Affect the Rest of Your Day
How does your performance in the beginning of the trading day, affect how you trade and how does it affect the ensuing results for the rest of the trading day?. I am referring to getting a good start, or getting a poor start (in terms of P/L) to the day. Take a look at the following example, taken from a discretionary trader. This is actually quite common...
Below is a representation of a string of 19 trading days in March. The colored left column represents the date; the days highlighted in green were winning days and those highlighted in red were losing days. The third column represents the total P/L earned or lost during the specific hour of day (indicated in the middle column).
Out of the 19 trading days, 11 were losing days and 8 were winning days. Two significant trends appear:
© Copyright 2007 David Adler
All rights reserved
All analysis generated with the TraderDNA Analyzer.
Below is a representation of a string of 19 trading days in March. The colored left column represents the date; the days highlighted in green were winning days and those highlighted in red were losing days. The third column represents the total P/L earned or lost during the specific hour of day (indicated in the middle column).
Out of the 19 trading days, 11 were losing days and 8 were winning days. Two significant trends appear:
1) 8 of the 9 days in which the first trading hour was negative (represented in yellow), have turned out to be losing days. I might conclude the following: For the past 4 trading weeks, when I've been down after my first hour of trading, 89% of the time the rest of the day ended up as a losing day.
2) Of the 8 days in which the first TWO hours were negative, all 8 days have ended up as losing days. From this I might conclude that over the past month, when the first two hours of my trading day were both negative (in terms of P/L), 100% of the time, I have ended the day in a loss.
© Copyright 2007 David Adler
All rights reserved
All analysis generated with the TraderDNA Analyzer.
Monday, July 30, 2007
How Do You Trade During Different Times of Day?
How do you trade during different hours of the day, and how does your P/L differ as the day progresses? Do you know? Have you identified certain hours or parts of the day when you are generally profitable, and others when you are not? Finding out such information can be a very powerful discovery, and successfully modifying your approach based on what you find can have immediate results on your bottom line.
If I look back at one week of my trading during which I traded everyday from 8am to 3 pm, I'll have 5 days of data to look at for some pattern. Let's just assume that my Hour of Day P/L analysis of the past 5 trading days shows that in every one of the 5 days, my losses were greatest from 8-10am. Well, that's intriguing, but may not bear much significance, statistically speaking.
Generally in small data sets (sample sizes), even very large relations cannot be considered reliable (significant), whereas in relatively large samples, much smaller relations between variables will be significant. Point being, the more data we have to measure (or trading days, in this example), the more significant our findings will be.
I'd be very happy if I were able to find out that out of the last 2 months (40 trading days), on 34 of the 40 days, I've lost money from 8-9am. Of course I wouldn't be happy about consistently losing money during that time of day, but rather I'd be pleased to have discovered this. Why? Because I can now do one of two things, both of which will likely improve my P/L: I could stop trading during this time altogether, or I could look into this hour in more detail and compare it to other hours to see exactly what I am doing differently and what is causing the losses from 8-9am.
The first choice above has its limitations. For example, if I stop trading altogether from 8-9am who's to say 9-10am wouldn't now become the new losing hour? Maybe it's more an issue of momentum, rather than how my system is interacting with the market during that specific hour. I want to find out.
In this example I've created a bar graph showing average P/L per trade, per the hour of day, shown in the top graph for only losing days, and the bottom graph for only winning days. The charts show data from 54 trading days (33 winning days and 21 losing days).
Creating one graph for each type of day allows me to sort the data depending on the outcome of the day, and in turn makes the analysis more specific. I'm able to see what generally happens during winning days, and what generally happens during losing days.
If I look back at one week of my trading during which I traded everyday from 8am to 3 pm, I'll have 5 days of data to look at for some pattern. Let's just assume that my Hour of Day P/L analysis of the past 5 trading days shows that in every one of the 5 days, my losses were greatest from 8-10am. Well, that's intriguing, but may not bear much significance, statistically speaking.
Generally in small data sets (sample sizes), even very large relations cannot be considered reliable (significant), whereas in relatively large samples, much smaller relations between variables will be significant. Point being, the more data we have to measure (or trading days, in this example), the more significant our findings will be.
I'd be very happy if I were able to find out that out of the last 2 months (40 trading days), on 34 of the 40 days, I've lost money from 8-9am. Of course I wouldn't be happy about consistently losing money during that time of day, but rather I'd be pleased to have discovered this. Why? Because I can now do one of two things, both of which will likely improve my P/L: I could stop trading during this time altogether, or I could look into this hour in more detail and compare it to other hours to see exactly what I am doing differently and what is causing the losses from 8-9am.
The first choice above has its limitations. For example, if I stop trading altogether from 8-9am who's to say 9-10am wouldn't now become the new losing hour? Maybe it's more an issue of momentum, rather than how my system is interacting with the market during that specific hour. I want to find out.
In this example I've created a bar graph showing average P/L per trade, per the hour of day, shown in the top graph for only losing days, and the bottom graph for only winning days. The charts show data from 54 trading days (33 winning days and 21 losing days).
Creating one graph for each type of day allows me to sort the data depending on the outcome of the day, and in turn makes the analysis more specific. I'm able to see what generally happens during winning days, and what generally happens during losing days.
-The first thing I notice here is that on my winning days I am getting a good start to the day and generally trading well in the first 3 hours. As the winning day continues, each hour I’m able to keep my losses smaller than they are in the corresponding time frames in the losing days.
-I see that losing days generally get off to a poor start and tend to get worse as the day progresses, with the exception of the 2:00 hour.
-The 2:00 pm hour has been a good hour for me consistently. I might look into this in more detail and look at different measures in this hour compared to all other hours grouped together so I can understand why I am generally profitable at this time of day.
-What’s interesting is that in looking at the bar charts, my P/L trend looks very similar in both my winning days and my losing days: I get progressively worse (in terms of P/L) as the afternoon approaches and continue to suffer losses, despite my averagely good performance during 2:00-3:00pm.
- The main difference seems to be that I start the day well in my winning days and generally keep my losses smaller throughout the day.
© Copyright 2007 David Adler
All rights reserved
All analysis generated with the TraderDNA Analyzer.
Monday, July 9, 2007
Winning Days vs. Losing Days
What do I do differently on my winning days and losing days?
Ask yourself that question and then try to answer it.
Why did you answer the way you did? Did you think of recent winning and losing days and remember what stood out during those days? Did you remember a conclusion you had reached in the past about your trading? Or maybe you looked back at detailed notes you had taken over the course of numerous trading days.
To accurately answer this question you need to take an extensive and objective look at winning days, and losing days separately. The easiest and quickest way to do this is to group together all of your trade data from every winning day, and then group together all of your trade data from every losing day.
The more data you've collected/saved, the more statistical significance your analysis will have. That said, two weeks of trading consists of 10 days with which you can run analysis, and is at the very low end in terms of the sufficient amount of data needed. 2 months of trade data (40 days) is a lot better, and 4 months (80 days) is even better.
We're working on some very interesting things right now at TraderDNA. We've recently added a feature that can group all winning days together, and all losing days together from a designated time period, so we can run analysis on the two groups of days. I'll show you the value of this...
In this view we're showing the average P/L per losing trades on losing days, losing trades on winning days, winning trades on losing days, and winning trades on winning days. Here I can easily see that my losers are bigger on my losing days than they are on my winning days. My winners are larger on my winning days than on my losing days. Ok, fine. This is what I'd expect.
Here's a different trader...
This guy looks the polar opposite. His losers are actually slightly larger on his winning days than they are on his losing days. Conversely, his winners are smaller on his winning days. The point of bringing up this second example is to show that different traders might have different causes for winning and losing days.
Now I'd like to look at the first trader above again in more detail. I want to know WHY his losers are larger on losing days than they are on winning days. It'd also be nice to know exactly why his winners are larger on winning days.
I want to know more about the difference in these two types of days. Average P/L per trade can only tell me so much.
So let's look at other measures than P/L. I am searching for significant differences in these values (measures) between both types of days (winning, losing)...
Average Negative Drift represents the average amount of heat(risk, downside) incurred during every open position. Average Positive Drift represents the average profit potential seen in every open position. Average Lost Opportunity Drift represents the difference between Avg. Positive Drift and where I covered my trade.
An example is useful here. Let's say I took a long position in the market at the price of 100. The market traded down to 96 and I am still in my position. Then the market starts to come back and trades up to 108. I cover my position at 106. My Negative Drift is 4 (100-96), my Positive Drift is 8 (108-100), and my Lost Opportunity Drift is 2 (108-106). I will discuss these measures in more detail in future posts but for now let's continue with this example.
This guy's trading looks eerily similar with respect to the amount of risk and potential profit he sees in his open positions, and with respect to his amount of Lost Opportunity Drift. This tells me a couple things. First off, it tells me his reasons for getting in and out of trades, or the parameters that dictate his decision when and where to trade, and how much downside to take before taking a loss, are nearly identical on both winning days and losing days. From this I can conclude that the amount of Neg. Drift, Pos. Drift, and Lost Opportunity are not determining factors in whether he has a winning or losing day. It also indicates that his parameters are static and his style of trading does not vary from day to day.
So let's look further. We're searching for significant differences in these measures over the two types of days.
Here I can see the average number of consecutive losing trades, the average number of contracts traded per trade, and the average number of contracts on scratched trades (avg. size of scratched trades), separated by winning days and losing days.
The chart shows that he traded slightly larger size on my losing days, but by an amount that is probably not significant in the context of things (middle column).
The last column is the most interesting to me because it shows that he's scratching bigger trades on his winning days. This suggests that this might have some impact on the overall outcome of the day. I could drill down and look at the scratched trades in more detail but I'd like to keep this example general and continue on.
The fact that he generally has more consecutive losing trades on losing days (first column) is nice to know, but I’m more interested in knowing why.
Average Time Since Last Win, and Average Time Since Last Loss represent the amount of time spent out of the market after a winner and loser, respectively, before putting on a new trade of any kind.
The first column indicates that on losing days, after his losing trades he enters a new trade much sooner than he does on his winning days.
He takes more time before he trades again after a winner during his winning days than he does on his losing days (indicated in the second column).
He spends slightly more time in his losing trades on losing days than he does on winning days. He spends longer in his winning trades on winning days.
Going forward, if I'm him I would want to be conscious of the fact that if I rush to back into the market after a winner it might contribute (at least to some extent) to the day turning out in a loss. I would also want to remember that on losing days, for whatever reason, I spend less time out of the market after a losing trade than I do on winning days, for whatever reason. Let's have a look at a couple more measures...
This chart shows the average number of times I add additional contracts to a losing trade (Negative Stackup) and the average number of times I add additional contracts to a winning trade (Positive Stackup), during winning days, and losing days.
The first column tells me he adds additional contracts to winning positions more often during losing days than he does during winning days. This suggests he's performing better when he doesn't add additional contracts to winning positions, or when he adds additional contracts to winners less frequently.
To me, the second column reveals the most important piece of information of all: In losing days, he adds to losing positions at a rate of nearly twice as often as he does in winning days.
The conclusions he comes to might not be a very big surprise to him, but more importantly, the data brought to surface facts about his trading that likely contributed to the outcome of the day. His ability to remain conscious of the information learned above and apply effective changes to his trading based on what past data has told him should increase his bottom line, statistically speaking.
© Copyright 2007 David Adler
All rights reserved
All analysis generated with the TraderDNA Analyzer.
Ask yourself that question and then try to answer it.
Why did you answer the way you did? Did you think of recent winning and losing days and remember what stood out during those days? Did you remember a conclusion you had reached in the past about your trading? Or maybe you looked back at detailed notes you had taken over the course of numerous trading days.
To accurately answer this question you need to take an extensive and objective look at winning days, and losing days separately. The easiest and quickest way to do this is to group together all of your trade data from every winning day, and then group together all of your trade data from every losing day.
The more data you've collected/saved, the more statistical significance your analysis will have. That said, two weeks of trading consists of 10 days with which you can run analysis, and is at the very low end in terms of the sufficient amount of data needed. 2 months of trade data (40 days) is a lot better, and 4 months (80 days) is even better.
We're working on some very interesting things right now at TraderDNA. We've recently added a feature that can group all winning days together, and all losing days together from a designated time period, so we can run analysis on the two groups of days. I'll show you the value of this...
In this view we're showing the average P/L per losing trades on losing days, losing trades on winning days, winning trades on losing days, and winning trades on winning days. Here I can easily see that my losers are bigger on my losing days than they are on my winning days. My winners are larger on my winning days than on my losing days. Ok, fine. This is what I'd expect.
Here's a different trader...
This guy looks the polar opposite. His losers are actually slightly larger on his winning days than they are on his losing days. Conversely, his winners are smaller on his winning days. The point of bringing up this second example is to show that different traders might have different causes for winning and losing days.
Now I'd like to look at the first trader above again in more detail. I want to know WHY his losers are larger on losing days than they are on winning days. It'd also be nice to know exactly why his winners are larger on winning days.
I want to know more about the difference in these two types of days. Average P/L per trade can only tell me so much.
So let's look at other measures than P/L. I am searching for significant differences in these values (measures) between both types of days (winning, losing)...
Average Negative Drift represents the average amount of heat(risk, downside) incurred during every open position. Average Positive Drift represents the average profit potential seen in every open position. Average Lost Opportunity Drift represents the difference between Avg. Positive Drift and where I covered my trade.
An example is useful here. Let's say I took a long position in the market at the price of 100. The market traded down to 96 and I am still in my position. Then the market starts to come back and trades up to 108. I cover my position at 106. My Negative Drift is 4 (100-96), my Positive Drift is 8 (108-100), and my Lost Opportunity Drift is 2 (108-106). I will discuss these measures in more detail in future posts but for now let's continue with this example.
This guy's trading looks eerily similar with respect to the amount of risk and potential profit he sees in his open positions, and with respect to his amount of Lost Opportunity Drift. This tells me a couple things. First off, it tells me his reasons for getting in and out of trades, or the parameters that dictate his decision when and where to trade, and how much downside to take before taking a loss, are nearly identical on both winning days and losing days. From this I can conclude that the amount of Neg. Drift, Pos. Drift, and Lost Opportunity are not determining factors in whether he has a winning or losing day. It also indicates that his parameters are static and his style of trading does not vary from day to day.
So let's look further. We're searching for significant differences in these measures over the two types of days.
Here I can see the average number of consecutive losing trades, the average number of contracts traded per trade, and the average number of contracts on scratched trades (avg. size of scratched trades), separated by winning days and losing days.
The chart shows that he traded slightly larger size on my losing days, but by an amount that is probably not significant in the context of things (middle column).
The last column is the most interesting to me because it shows that he's scratching bigger trades on his winning days. This suggests that this might have some impact on the overall outcome of the day. I could drill down and look at the scratched trades in more detail but I'd like to keep this example general and continue on.
The fact that he generally has more consecutive losing trades on losing days (first column) is nice to know, but I’m more interested in knowing why.
Average Time Since Last Win, and Average Time Since Last Loss represent the amount of time spent out of the market after a winner and loser, respectively, before putting on a new trade of any kind.
The first column indicates that on losing days, after his losing trades he enters a new trade much sooner than he does on his winning days.
He takes more time before he trades again after a winner during his winning days than he does on his losing days (indicated in the second column).
He spends slightly more time in his losing trades on losing days than he does on winning days. He spends longer in his winning trades on winning days.
Going forward, if I'm him I would want to be conscious of the fact that if I rush to back into the market after a winner it might contribute (at least to some extent) to the day turning out in a loss. I would also want to remember that on losing days, for whatever reason, I spend less time out of the market after a losing trade than I do on winning days, for whatever reason. Let's have a look at a couple more measures...
This chart shows the average number of times I add additional contracts to a losing trade (Negative Stackup) and the average number of times I add additional contracts to a winning trade (Positive Stackup), during winning days, and losing days.
The first column tells me he adds additional contracts to winning positions more often during losing days than he does during winning days. This suggests he's performing better when he doesn't add additional contracts to winning positions, or when he adds additional contracts to winners less frequently.
To me, the second column reveals the most important piece of information of all: In losing days, he adds to losing positions at a rate of nearly twice as often as he does in winning days.
The conclusions he comes to might not be a very big surprise to him, but more importantly, the data brought to surface facts about his trading that likely contributed to the outcome of the day. His ability to remain conscious of the information learned above and apply effective changes to his trading based on what past data has told him should increase his bottom line, statistically speaking.
© Copyright 2007 David Adler
All rights reserved
All analysis generated with the TraderDNA Analyzer.
Subscribe to:
Posts (Atom)
