Yamato
Legendary member
- Messages
- 9,840
- Likes
- 246
money and risk management
Today I need to study money and risk management, because I realize that I need to automate them better.
But it's hard. I've been working on this for years and I still have not found a good way to automate it, so that I don't have to constantly monitor how many contracts which of my systems are being allocated. One of the reasons why money management is so hard, besides the fact that I am bad at formulas, is that it's hard to appraise my systems' performance.
The most basic principle would be:
1) Let all profitable systems trade 1 contract.
But then of course, if capital is limited, we should favour the best systems, so I created formulas to appraise which are the most profitable systems and implemented this money management:
2) Let the best systems trade more than the others.
Then I realized that it's not just a matter of how much profit, but also how much they could lose during a drawdown, so:
3) Let the best systems trade more than the others, and measure performance by profit AND drawdown.
Then I realized ROA wasn't enough, and I included Profit Factor.
4) Let the best systems trade more than the others, and measure performance by profit, drawdown AND profit factor.
Then I realized I also want to know how often and how deeply a system incurs drawdowns, so I measured this with my simple formula of average deviation, which is not too bad, I guess. Since alone it would not be a good indicator, I divided profit by average deviation.
5) Let the best systems trade more than the others, and measure performance by profit, drawdown, profit factor AND average deviation.
Now, it would seem my appraisal of performance is complete, and maybe it would, but, with a small capital of a few thousands, strange things happen, like the best systems not trading because risk management doesn't allow them, which brings me to another ingredient: risk management, which in my own vocabulary I use to define my control over losses. I never allow a system to incur a loss bigger than x% of total capital. While I would define "money management" as deciding how much you should allocate to each system based on the "good" side of a system's performance (mostly the profit it makes), I consider "risk management" as something to protect me from excessive risk. Traders use and misuse these two terms in so many ways, and this is the only way I could make sense of them, for now.
So this brings me to the present definition of my risk and money management formulas, which I still don't find satisfactory nor self-sufficient:
6) Let the best systems trade more than the others, and measure performance by profit, drawdown, profit factor and average deviation. Make sure no system can lose more than x% of capital.
In the past I've downloaded, via emule, dozens of .pdf books on money management, but of course they need to be read, which is not easy at all, considering each book doesn't approach it with a quick answer, but needs to be at least 100 pages long (as all books are) so it starts discussing money management from the middle ages, and basically is filled up with a lot of god damn bull****. But that's what you have to do to get your book published: if you can say it in 10 pages, they won't publish your book, so you have to fill it up with bull****. It's like a natural selection all reversed: only the people willing to write bull**** get published. The others are probably writing on forums, but it's quite hard to find their stuff among the sea of smilies and bull**** in forums.
So maybe I will try to see what I can get out of wikipedia, which is probably the best combination of both worlds (academic bull**** world and forum bull**** world).
Obviously I went here first:
http://en.wikipedia.org/wiki/Money_management
But it's no good. Except that it tells me to read this book, which is a good one actually and I have it:
Until now, the best wikipedia entry I remember reading on money and risk management was this entry:
http://en.wikipedia.org/wiki/Kelly_criterion
But let's check out "risk management", just in case:
http://en.wikipedia.org/wiki/Risk_management
As I read this stuff, i need to remind myself constantly that I don't give a **** about being an academic and that I should stick to the original point of making money. It's hard to remember this, when everyone writing this stuff is busy making himself look good and knowledgeable. These people don't care to be useful to me: they just want to show off their knowledge. Like the "look elsewhere" user.
Anyway, the wikipedia entry is no good. So we're left with kelly criterion as far as wikipedia.
You see, my problem is that whereas it's easy to see which systems are the best, and I've done that and even automated it, it's not easy, when you have enough capital to allow your good systems (and the others) to trade more than 1 contract, to automate the allocation of these extra contracts.
Maybe I should completely separate the risk from the money management part...
Let's first discuss kelly. Kelly is great, but, with 40 systems, it gets too complicated to use it, at least with my low skills at using formulas.
So let's see if I am at least doing what he considers to be right. Other than the kelly formulas, the most important stuff I found on that entry is:
This is telling me to stay on the safe side in case I am not sure of my edge and drawdown (and I am not sure), because, if I underbet I will simply increase my capital more slowly. Whereas, if I overbet, I will blow out my account.
Forget the books, as they are too long and too hard to make sense of them. It's best to do it by myself in depth, than to read books superficially. I know what I want - I will figure it out by going back to my excel workbooks.
Let's take care of risk management (as I intend it) first. We want to separate the two completely.
RISK MANAGEMENT
The way I mean "risk management" is just to make sure that if a system incurs its worst possible loss, it won't exceed x% of capital. That "x" will vary depending on how much capital I have: I will define that "x" arbitrarily, even though I am aware that Larry Hite and others have said it should be no more than 1%. But I don't have their money.
I have estimated the worst possible loss by each system as an average of max drawdown (back-tested and forward-tested) and max loss during forward-testing, just in case my forward-testing sample isn't large enough.
I will now also add a rule that says if the foward-testing trades are less than 20, no more than 1 contract should be allowed.
[...]
After much thinking and tweaking, I have reached a solution:
1) I have finally given up on my plan to entirely automate risk and money management. Better to do less but well, than do much more but not well.
2) I have entirely separated risk management from every other column (no connections, no "dependents" functions). Now it calculates just this:
=-s!$B$20*s!$B$25/mm!C2
Capital multiplied by % of capital allowed to be lost by biggest loss and then divided by biggest expected loss.
If we have a capital of... say 100k and multiply by... say a 5% of capital allowed to be lost by biggest loss, we find out we can only risk 5k with each system. Then we divide it by a max loss of... say 2.5k and we get 2, which means only 2 contracts can be allocated to that system, if we don't want it to cause us a potential loss bigger than 5k. Now this is all automated and it limits our contracts, as it has to do with limiting our risk.
3) Now, on to money management. This is the toughest part, and it's the part I've kept arbitrary, or rather "manual" and "discretionary". It's hard because it's hard to appraise each system's performance. I have set up 2 columns:
A) one is the (automated) rating of my systems, based on my custom index I've spoken about in previous posts. Such index relies Return On Account, Profit Factor, Average Deviation. Each system's index value in turn gets weighed against the best system's value and then multiplied by the highest amount of capital I want to allocate to the best system:
=s!$B$23*(P2/s!$B$24)/B2
% of Capital allocated to best system times how well a system is doing compared to the best system divided by margin
Say we have a capital of 100k, and we decide to only invest 50% of it on our best system (how is it possible with 40 systems? systems do not all trade at once). We then have to multiply that 50k by how good a system's performance is compared to the best system, let's say our hypothetical system on OIL is only 50% as good, in terms of the mentioned Index. That CL system will then only get 25% of capital. Then we divide 25k by CL's overnight margin of 5k, and we get a result of 5, which is the suggested amount of contracts to be traded.
B) I then decide with my discretionary manual column if everything seems right and I am willing to let the CL system trade 5 contracts. Let's say I do, and have written a 5 to confirm that. That 5 will stay there no matter how much capital increases, so the system will only be allowed to trade 5 contracts until I check on it again and manually change that figure. The A) column, with the Index allocated contracts is only advice to me, but it doesn't affect the trading. The trading contracts get decided by two columns: the one with my manual choice and the risk management column, which in this case would limit contracts to 2 probably, because that very good CL system happens to lose too much.
Today I need to study money and risk management, because I realize that I need to automate them better.
But it's hard. I've been working on this for years and I still have not found a good way to automate it, so that I don't have to constantly monitor how many contracts which of my systems are being allocated. One of the reasons why money management is so hard, besides the fact that I am bad at formulas, is that it's hard to appraise my systems' performance.
The most basic principle would be:
1) Let all profitable systems trade 1 contract.
But then of course, if capital is limited, we should favour the best systems, so I created formulas to appraise which are the most profitable systems and implemented this money management:
2) Let the best systems trade more than the others.
Then I realized that it's not just a matter of how much profit, but also how much they could lose during a drawdown, so:
3) Let the best systems trade more than the others, and measure performance by profit AND drawdown.
Then I realized ROA wasn't enough, and I included Profit Factor.
4) Let the best systems trade more than the others, and measure performance by profit, drawdown AND profit factor.
Then I realized I also want to know how often and how deeply a system incurs drawdowns, so I measured this with my simple formula of average deviation, which is not too bad, I guess. Since alone it would not be a good indicator, I divided profit by average deviation.
5) Let the best systems trade more than the others, and measure performance by profit, drawdown, profit factor AND average deviation.
Now, it would seem my appraisal of performance is complete, and maybe it would, but, with a small capital of a few thousands, strange things happen, like the best systems not trading because risk management doesn't allow them, which brings me to another ingredient: risk management, which in my own vocabulary I use to define my control over losses. I never allow a system to incur a loss bigger than x% of total capital. While I would define "money management" as deciding how much you should allocate to each system based on the "good" side of a system's performance (mostly the profit it makes), I consider "risk management" as something to protect me from excessive risk. Traders use and misuse these two terms in so many ways, and this is the only way I could make sense of them, for now.
So this brings me to the present definition of my risk and money management formulas, which I still don't find satisfactory nor self-sufficient:
6) Let the best systems trade more than the others, and measure performance by profit, drawdown, profit factor and average deviation. Make sure no system can lose more than x% of capital.
In the past I've downloaded, via emule, dozens of .pdf books on money management, but of course they need to be read, which is not easy at all, considering each book doesn't approach it with a quick answer, but needs to be at least 100 pages long (as all books are) so it starts discussing money management from the middle ages, and basically is filled up with a lot of god damn bull****. But that's what you have to do to get your book published: if you can say it in 10 pages, they won't publish your book, so you have to fill it up with bull****. It's like a natural selection all reversed: only the people willing to write bull**** get published. The others are probably writing on forums, but it's quite hard to find their stuff among the sea of smilies and bull**** in forums.
So maybe I will try to see what I can get out of wikipedia, which is probably the best combination of both worlds (academic bull**** world and forum bull**** world).
Obviously I went here first:
http://en.wikipedia.org/wiki/Money_management
But it's no good. Except that it tells me to read this book, which is a good one actually and I have it:
Balsara, Nauzer J. (1992). Money Management Strategies for Futures Traders
Until now, the best wikipedia entry I remember reading on money and risk management was this entry:
http://en.wikipedia.org/wiki/Kelly_criterion
But let's check out "risk management", just in case:
http://en.wikipedia.org/wiki/Risk_management
As I read this stuff, i need to remind myself constantly that I don't give a **** about being an academic and that I should stick to the original point of making money. It's hard to remember this, when everyone writing this stuff is busy making himself look good and knowledgeable. These people don't care to be useful to me: they just want to show off their knowledge. Like the "look elsewhere" user.
Anyway, the wikipedia entry is no good. So we're left with kelly criterion as far as wikipedia.
You see, my problem is that whereas it's easy to see which systems are the best, and I've done that and even automated it, it's not easy, when you have enough capital to allow your good systems (and the others) to trade more than 1 contract, to automate the allocation of these extra contracts.
Maybe I should completely separate the risk from the money management part...
Let's first discuss kelly. Kelly is great, but, with 40 systems, it gets too complicated to use it, at least with my low skills at using formulas.
So let's see if I am at least doing what he considers to be right. Other than the kelly formulas, the most important stuff I found on that entry is:
Reasons to bet less than Kelly
A natural assumption is that taking more risk increases the probability of both very good and very bad outcomes. One of the most important ideas in Kelly is that betting more than the Kelly amount decreases the probability of very good results, while still increasing the probability of very bad results. Since in reality we seldom know the precise probabilities and payoffs, and since overbetting is worse than underbetting, it makes sense to err on the side of caution and bet less than the Kelly amount.
This is telling me to stay on the safe side in case I am not sure of my edge and drawdown (and I am not sure), because, if I underbet I will simply increase my capital more slowly. Whereas, if I overbet, I will blow out my account.
Forget the books, as they are too long and too hard to make sense of them. It's best to do it by myself in depth, than to read books superficially. I know what I want - I will figure it out by going back to my excel workbooks.
Let's take care of risk management (as I intend it) first. We want to separate the two completely.
RISK MANAGEMENT
The way I mean "risk management" is just to make sure that if a system incurs its worst possible loss, it won't exceed x% of capital. That "x" will vary depending on how much capital I have: I will define that "x" arbitrarily, even though I am aware that Larry Hite and others have said it should be no more than 1%. But I don't have their money.
I have estimated the worst possible loss by each system as an average of max drawdown (back-tested and forward-tested) and max loss during forward-testing, just in case my forward-testing sample isn't large enough.
I will now also add a rule that says if the foward-testing trades are less than 20, no more than 1 contract should be allowed.
[...]
After much thinking and tweaking, I have reached a solution:
1) I have finally given up on my plan to entirely automate risk and money management. Better to do less but well, than do much more but not well.
2) I have entirely separated risk management from every other column (no connections, no "dependents" functions). Now it calculates just this:
=-s!$B$20*s!$B$25/mm!C2
Capital multiplied by % of capital allowed to be lost by biggest loss and then divided by biggest expected loss.
If we have a capital of... say 100k and multiply by... say a 5% of capital allowed to be lost by biggest loss, we find out we can only risk 5k with each system. Then we divide it by a max loss of... say 2.5k and we get 2, which means only 2 contracts can be allocated to that system, if we don't want it to cause us a potential loss bigger than 5k. Now this is all automated and it limits our contracts, as it has to do with limiting our risk.
3) Now, on to money management. This is the toughest part, and it's the part I've kept arbitrary, or rather "manual" and "discretionary". It's hard because it's hard to appraise each system's performance. I have set up 2 columns:
A) one is the (automated) rating of my systems, based on my custom index I've spoken about in previous posts. Such index relies Return On Account, Profit Factor, Average Deviation. Each system's index value in turn gets weighed against the best system's value and then multiplied by the highest amount of capital I want to allocate to the best system:
=s!$B$23*(P2/s!$B$24)/B2
% of Capital allocated to best system times how well a system is doing compared to the best system divided by margin
Say we have a capital of 100k, and we decide to only invest 50% of it on our best system (how is it possible with 40 systems? systems do not all trade at once). We then have to multiply that 50k by how good a system's performance is compared to the best system, let's say our hypothetical system on OIL is only 50% as good, in terms of the mentioned Index. That CL system will then only get 25% of capital. Then we divide 25k by CL's overnight margin of 5k, and we get a result of 5, which is the suggested amount of contracts to be traded.
B) I then decide with my discretionary manual column if everything seems right and I am willing to let the CL system trade 5 contracts. Let's say I do, and have written a 5 to confirm that. That 5 will stay there no matter how much capital increases, so the system will only be allowed to trade 5 contracts until I check on it again and manually change that figure. The A) column, with the Index allocated contracts is only advice to me, but it doesn't affect the trading. The trading contracts get decided by two columns: the one with my manual choice and the risk management column, which in this case would limit contracts to 2 probably, because that very good CL system happens to lose too much.
Last edited: