...thanks for replying. By "opportunity" do you mean the fraction of allowed trade times which result in trades? E.g., for a daily system, three trades per week produces 0.6 opportunity? Is the value 2 for expectancy the number of points? NQ or ES? Sorry, I think in dollars per month per contract traded. Trying to relate your experience to my own.
You are using an out of date browser. It may not display this or other websites correctly.
You should upgrade or use an alternative browser.
You should upgrade or use an alternative browser.
Systemized profit-taking?
- Thread starter illiquid
- Start date
Sorry, should have been more specific.Quote from hypostomus:
...thanks for replying. By "opportunity" do you mean the fraction of allowed trade times which result in trades? E.g., for a daily system, three trades per week produces 0.6 opportunity? Is the value 2 for expectancy the number of points? NQ or ES? Sorry, I think in dollars per month per contract traded. Trying to relate your experience to my own.
Opportunity = trades per year;
Expectancy = [1+(avgWin/avgLoss)]*%Win -1
Therefore Opp*Exp = $ return pa for dollar risked.
...thanks. If I am doing the algebra right, your expression for expectancy boils down to
[(1-%Win)*avgLoss + %Win*avgWin]/avgLoss,
which is expectancy per trade/average loss, or reward/risk.
Therefore I think you mean your "2" to apply to R/R, as trades per year times expectancy per trade divided by average loss is annual expectancy divided by average loss, which for you is more than 2 (for me, that's about right).
Sorry to drill in. I'm an Asperger's syndrome suffer (remote, pedantic, socially inept).
Anyway, if I'm interpreting right, IMO 2:1 R/R is pretty damn good for intraday. In the original context of my question, I too find that if a system is that good "raw", it won't be improved by stops and takes. Unfortunately most of the ideas I come up with are not that good. I only have one system that good, and it doesn't trade often enough to be worth the trouble.
[(1-%Win)*avgLoss + %Win*avgWin]/avgLoss,
which is expectancy per trade/average loss, or reward/risk.
Therefore I think you mean your "2" to apply to R/R, as trades per year times expectancy per trade divided by average loss is annual expectancy divided by average loss, which for you is more than 2 (for me, that's about right).
Sorry to drill in. I'm an Asperger's syndrome suffer (remote, pedantic, socially inept).
Anyway, if I'm interpreting right, IMO 2:1 R/R is pretty damn good for intraday. In the original context of my question, I too find that if a system is that good "raw", it won't be improved by stops and takes. Unfortunately most of the ideas I come up with are not that good. I only have one system that good, and it doesn't trade often enough to be worth the trouble.
Damn that is complicated! My math is bad at the best of times, so the algebra is wasted on me, sorry. The 2 in my prev post refers to Opp*Exp, which means for each dollar I risk I stand to get 2 back (in addition to my original dollar) each year. Hope that clears it up.Quote from hypostomus:
...thanks. If I am doing the algebra right, your expression for expectancy boils down to
[(1-%Win)*avgLoss + %Win*avgWin]/avgLoss,
which is expectancy per trade/average loss, or reward/risk.
Therefore I think you mean your "2" to apply to R/R, as trades per year times expectancy per trade divided by average loss is annual expectancy divided by average loss, which for you is more than 2 (for me, that's about right).
Sorry to drill in. I'm an Asperger's syndrome suffer (remote, pedantic, socially inept).
Anyway, if I'm interpreting right, IMO 2:1 R/R is pretty damn good for intraday. In the original context of my question, I too find that if a system is that good "raw", it won't be improved by stops and takes. Unfortunately most of the ideas I come up with are not that good. I only have one system that good, and it doesn't trade often enough to be worth the trouble.
Position trader, EoD only (tried intraday, too tough for me).
Quote from Grob109:
to reinforce the commentary made so far, their areways to define this arena of failure.
[...]
I am providing wake up calls to a lot of people on a lot of levels.
the pragmatic ones like provideing graphs in advance of the actions that do follow is more or less a wake up call using a blunt force insturment.
This post is one where some people really get to enjoy a really positive "eureka" moment. If a person gets to the point of enquiry on how ot optimize taking profits on trend endings, then they are sitting in a place for a magnificent and major breakthrough. Enjoy. And have a nice look at something pervasive and elegant.
Hi Grob109,
You sound like the ghoast of Jack Hershey.
nononsense
...sorry to be so dense. Now I get it. In my experience that is excellent. The system I currently trade yields only about 1.1 to 1, but I trade it anyway because it trades 3 days out of 4 and nets more than competing systems which have an R/R of about 2 but trade 1/3 as often.
In making a system evaluation, do you ever look at other measures of risk? For example, the "smoothness" of the equity curve? A poster elsewhere here (I forget where) said he calculates the standard deviation of the equity curve.
Do you have a rule that tells you when the system is broke? Like exceeding backtested drawdown? Equity curve failing to make a higher high? Absent long experience, I am flying by the seat of my pants here doing what seems logical.
You know any good books on system development? I have ordered Robert Pardo's book based on Amazon reviews, but the pickings seem to be slim. Thanks. - Mike
In making a system evaluation, do you ever look at other measures of risk? For example, the "smoothness" of the equity curve? A poster elsewhere here (I forget where) said he calculates the standard deviation of the equity curve.
Do you have a rule that tells you when the system is broke? Like exceeding backtested drawdown? Equity curve failing to make a higher high? Absent long experience, I am flying by the seat of my pants here doing what seems logical.
You know any good books on system development? I have ordered Robert Pardo's book based on Amazon reviews, but the pickings seem to be slim. Thanks. - Mike
Hypostomus, in your recent reply to me, you astutely observe, "I think it is axiomatic that if you are backtesting you plan to trade the system mechanically, even automate it."
Is there some simple software available on which I can test combos of entry price trades/limits/stops? I can of course do this manually .. a quicker route would be better though.
I have my own data so I can input market days as a price line (an index value) with moves above as + (plus) and moves below as - (minus). This would be for the Dow.
Any comments would be greatly appreciated.
Is there some simple software available on which I can test combos of entry price trades/limits/stops? I can of course do this manually .. a quicker route would be better though.
I have my own data so I can input market days as a price line (an index value) with moves above as + (plus) and moves below as - (minus). This would be for the Dow.
Any comments would be greatly appreciated.
...I am humbled that YOU would ask ME. The way I do it is very painful. I try to keep my candidate systems limited to five rules. Two of them are time rules and I expect them to have weak influence on expectation. The other three are the entry rule, the stop rule (which may also trigger a reverse if it's a crappy system), and the take profit rule (also if it's a crappy system). I try to keep it this relatively uncomplicated because it's the only hope I have of optimizing it without winding up with a solution at a minor optimum, rather than at the major one. Being a quant I think in multidimensional statistics.
In E-Signal I first iterate through values for the entry parameter for a rough optimization with no stops and no takes. Then with the entry rule semi-optimized, I iterate through the take profit parameter for no stop. Then I iterate through the stop parameter for no take profit. A narrowed trade space for the three parameters then is evident, which I further iterate to get the absolute best (overfitted, haha!) set of rules. I do this every weekend for my favorite system and for two backup systems, not only to see if the systems are behaving (aren't just random flukes), but also to reoptimize for any changes in volatility.
The people I talk to who REALLY know how to trade generally find this bizarre, because they have something that has worked like a hose since Livermore was bucketshopping. They also are unimpressed with the expectation, but hey, it buys lunch.
If this all sounds painful, overly complicated, time consuming, and masochistic, you are absolutely right. I recommend instead finding a hose.
In E-Signal I first iterate through values for the entry parameter for a rough optimization with no stops and no takes. Then with the entry rule semi-optimized, I iterate through the take profit parameter for no stop. Then I iterate through the stop parameter for no take profit. A narrowed trade space for the three parameters then is evident, which I further iterate to get the absolute best (overfitted, haha!) set of rules. I do this every weekend for my favorite system and for two backup systems, not only to see if the systems are behaving (aren't just random flukes), but also to reoptimize for any changes in volatility.
The people I talk to who REALLY know how to trade generally find this bizarre, because they have something that has worked like a hose since Livermore was bucketshopping. They also are unimpressed with the expectation, but hey, it buys lunch.
If this all sounds painful, overly complicated, time consuming, and masochistic, you are absolutely right. I recommend instead finding a hose.
..thanks for the lead. That's one I was not aware of. I'll check it out. Need help big time with some really fundamental questions on assessing the goodness of a system. - Mike