The first lesson you learn to approach the markets is that "no risk no gain." But in my view, the second is even more important: "Without proper risk management strategy is not able to last over time." If you understand this, then why instead of being obsessed with complex input algorithms to increase the percentage of hits, usually flower a day, not spend more time devising strategies that enable it to maintain control of their operational risk?
When designing a trading plan, there are at least three levels at which an adequate risk model must be clearly specified and measured:
1 .- System Rules: Do not conceive a quality strategy that explicitly incorporates a mechanism for closing positions for a given level of losses. This is what is known as Stop MM. It is not necessary (or even desirable) that it has a fixed monetary value. Sufficient to make it clear point from which I retired and did not keep losing.
2 .- Asset Allocation: Never evade the question: How much capital devoted to each market / system and why? Just as you do not use their capital never fatigues (except acute temporary insanity) to buy shares of a single company should not operate alone systems. A good asset allocation methodology will split the money available in various systems (as many as possible) considering: type of market, correlation between strategies and expected value (or quality) of the same. In short, money management, pure and simple as the best antidote to spread the risk in a coherent and consistent package of 'vectors' (markets / systems) of investment.
3 .- Size of the position .- How many tablets of bets placed on the table and how to increase your number when everything goes well? On this issue, whole books have been written and we've already adequately addressed in previous articles. A tip, regardless of the formula that applies to periodically review the risk-benefit and try to remain stable throughout the sequence of operations.
Well, that said, let us examine one of the evaluation and implementation models risk that has been most studied in recent years, the proposed Van K. Tharp known as multiples of R. As a first approximation I recommend reading Chapter VI of his book Success in Trading (Value Editions, Barcelona, \u200b\u200b2007) can also see the article on this website, System Quality Number (SQN) and optmización of parameres.
The starting point is the excellent idea of \u200b\u200bTharp to consider the risk-reward ratio of a system as a multiple of the risk involved in the sequence of operations. Thus, the entire sequence of profit and loss (P / L) may be expressed in terms of variable empirical value R.
Now, our main problem is to determine an estimate for 'R' (initial risk) that is realistic and consistent with the logic of the system used. To do this we have several possibilities:
1 .- When the system has an objective algorithm output losses (MM stop) with a fixed value, then this will be the value of R, regardless of which sometimes due to slippage, gaps between bars or inefficiencies of the platform, the losses exceeded the value of the stop. For example, if: MM
Stop = $ 300 → (-1R) = 300.
Suppose now the following sequence of operations: 1700, -50, 1200, -300, 150.
OP.1 = 1700/300 = 5.6 R
Op.2 = -50/300 = -0.16 R
Op.3 = 1200 / 300 = 4R
Op . 4 = -300 / 300 =-1R
Op 5 = 150 / 300 = 0.5 R
Tharp In the model of hope (E) of the system is defined as the average multiples of n R:
E = (sum R / no. ops) = 1.78.
This means that this system will have a hope of winning of 1.78 times the capital is at stake. In other words, for every $ 300 they throw on the table, earn on average $ 534.
2 .- When the system has a fixed stop, the calculation of 'R' can be approximated by estimating the average loss over the sequence of operations, provided this is sufficiently representative.
3 .- In the case of very intensive systems intraday operations, perhaps what interests us is the daily risk rating or session: In this way, the initial risk is equal to the average daily loss. Logically, the P / L applied in the number of multiples will be the result of the profit or loss per day. I found that when Cadence is the daily operations of three or more transactions per system, with the estimated value of R is obtained consistent data on the system hope.
From a long series of gains and losses expressed in terms of initial risk, you get what you called distribution Tharp multiples of R. This is the aspect of a sequence of 472 operations a system obtained intraday applied Bund futures.
Home Management R & Multiples monetary monetary management.
Multiples of R and monetary management.
TradingSys Writer (NDGA)
Tuesday, December 9, 2008
The first lesson one learns to approach the markets is that "no risk no gain." But in my view, the second is even more important: "Without proper risk management strategy is not able to last over time." If you understand this, then why instead of being obsessed with input complex algorithms to increase the percentage of hits, usually flower a day, not spend more time devising strategies that enable it to maintain control of their operational risk ?
When designing a trading plan, there are at least three levels at which a appropriate risk model must be clearly specified and measured:
1 .- System Rules: Do not conceive a quality strategy that explicitly incorporates a mechanism for closing positions for a given level of losses. This is what is known as Stop MM. It is not necessary (or even desirable) that it has a fixed monetary value. Sufficient to make clear the point from which I retired and did not keep losing.
2 .- Asset Allocation: Never evade the question: How much capital devoted to each market / system and why? Just as you do not use their capital never fatigues (except acute temporary insanity) to buy actions of a single company should not operate alone systems. A good asset allocation methodology will split the money available in various systems (as many as possible) considering: type of market, correlation between strategies and expected value (or quality) of the same. In short, money management, pure and simple as the best antidote to spread the risk in a coherent and consistent package of 'vectors' (markets / systems) of investment.
3 .- Size of the position .- How many tablets of bets placed on the table and how to increase your number when everything goes well? On this issue Entire books have been written and we've already adequately addressed in previous articles. A tip, regardless of the formula that applies to periodically review the risk-benefit and try to remain stable throughout the sequence of operations.
Well, that said, let us examine one of the models of risk assessment and implementation of the most widely studied in recent years, the proposed Van K. Tharp known as multiples of R. As a first approximation I recommend reading Chapter VI of his book Success in Trading (Value Editions, Barcelona, \u200b\u200b2007) can also see the article on this website; System Quality Number (SQN) and optmización of parameres.
The starting point is the excellent idea of \u200b\u200bTharp to consider the risk-reward ratio of a system as a multiple of the risk involved in the sequence of operations. Thus, the entire sequence of profit and loss (P / L) may be expressed in terms of variable empirical value of R.
Now, our main problem is to determine an estimate for 'R' (initial risk) that is realistic and consistent with the logic of the system used. To do this we have several possibilities:
1 .- When the system has an output algorithm objective of losses (MM stop) with a fixed value, then this will be the value of R, regardless of that sometimes, due to slippage, gaps between bars or inefficiencies of the platform, the losses exceeded the value of the stop . For example, if: MM
Stop = $ 300 → (-1R) = 300.
Now suppose the following sequence of operations: 1700, -50, 1200, -300, 150.
OP.1 = 1700/300 = 5.6 R
Op.2 = -50/300 = -0.16 R
Op.3 = 1200 / 300 = 4R
Op . 4 = -300 / 300 =-1R
Op
5 = 150 / 300 = 0.5 R
Tharp In the model of hope (E) of the system is defined as the average multiples of R n:
E = (sum R / no. ops) = 1.78.
This means that this system will have a hope of winning of 1.78 times the capital is at stake. In other words, for every $ 300 they throw on the table, earn on average $ 534.
2 .- When the system has a fixed stop, the calculation of 'R' can be approximated by estimating the average loss over the sequence of operations, provided this is sufficiently representative.
3 .- In the case of very intensive systems intraday operations, perhaps what interests us is the daily risk rating or session: In this way, the initial risk is equal to the average daily loss. Logically, the P / L applied in the number of multiples will be the result of the profit or loss per day. I have found that when the daily operational cadence is three or more operations per set, with this estimated value of R is obtained consistent data on the system hope.
From a long series of gains and losses expressed in terms of initial risk, you get what you called distribution Tharp multiples of R. This is the aspect of a sequence of 472 operations a system obtained intraday applied Bund futures.
Once the distribution, and knowing these basic facts:
Average loss ('Initial Risk') = € 85.29
Sum of R = 149.35
Average R (Expectancy) = 0.32
deviation of R = 1.62
Maximum R = 6.41
Minimum R = -7.42
can make many inferences about nature system. For example, the average R is the expectation (E) system, which allows me to calculate what the system will gain in "x" operations. So if I want to know the benefits I expect from this strategy in the next 100 transactions I have only to multiply:
Expected Profit (BE) = E * num. Ops. * Value of R
BE = 0.32 * 100 * € 85.29 = € 2,729.28
The standard deviation tells us about the variability of multiples of R. They are usually more efficient systems with low dispersion of values \u200b\u200baround the average. But this, by itself, is not a factor too significant. In fact, what matters is the ratio between mean and standard deviation, weighted by the square root of the number of operations. This is what he calls Tharp System Quality Number (SQN):
SQN = (Average R / R deviation) * Root no. Ops.
In this example,
SQN = (0.32 / 1.62) * Root (472) = 4.29.
a system is considered good when the ratio SQN> 2 and excellent if it exceeds the value of 3. However, the evolution in time of SQN is quite erratic during the first 80-100 operations, tending to stabilize as increasing their number. As discussed in my article on this subject, one of the best uses you can give the SQN is as optimization criterion. If you use the NinjaTrader platform is available remember a script that will optimize the parameters using SQN target criterion.
Another advantage of this ratio is that we can compare the quality of systems of diverse nature, as all data P / L standard in terms of R. Also, as discussed below, this estimate of quality is useful as a criterion for allocation of assets, the portfolio composition systematically.
When sorted multiples of R a class table and frequencies, we can easily find answers to some other interesting issues, as knowing what percentage of losing trades I can not find multiples of R especially bad-2R,-3R etc. In the example above, we have a 7.63% probability of appearing in future operations or above-2R and 1.91% of these exceeding the-3R.
network
Searching a program that emulates the methodology of Tharp, I found a free application in Excel MoneyExpert (designed by Matt Bowen, MTPedictor Company) is used to simulate a sequence of 100 operations on the basis of distribution of multiple R of any series of operations. Before using it, I recommend watching this video about how it works: http://www.mtptrader.com/videos/MoneyExpert.html should also read this thread: http://www.mtptrader.com/showthread.php ? p = 14619.
After entering the range and frequency of multiples of R and some initial inputs such as seed capital, risk per trade or commissions and slippage:
The blade also offers a comprehensive summary of results with all statistical ratios the series obtained. I've been playing around with this tool and, frankly, I find a good lab to get started and understand the methodology K. Van Tharp.
Finally, let's see how SQN ratio can be applied to asset allocation in a portfolio of multiple systems / markets:
Suppose we have an initial capital of 200,000 euros with which we build a diversified portfolio and, at the same time, positive weights to systems with higher expected profit. We have the following six systems we've tested in back-testing and out-sample test:
System 1; trend of continuous type, applied to FGBL:
SQN Ratio = 2.5
Max
. DD = € 3,500 per contract.
Guarantees
= 2,000 €
System 2, Break-out type, intraday, applied to EMD:
SQN Ratio = 1.95
Max. DD = € 5,600 per contract. Guarantees
= 2,500 €
System 3; antitendencial intraday applied to FDAX:
SQN Ratio = 2.62
Max. DD = € 9,850 per contract. Guarantees
= 15,000 €.
System 4, pattern recognition, intraday, applied to ZG:
Ratio = 3.72
SQN
Max. DD = € 5,800 per contract. Guarantees
= 2,800 €
System 5, trend, continuous, applied to GBP:
Ratio SQN = 2
Max. DD = € 4,200 per contract. Guarantees
= 1,500 €
System 6; microtendencial, intraday, applied to the CL:
SQN Ratio = 3.9
Max. DD = € 8,110 per contract. Guarantees
= 9,000 €
First, calculate a percentage of funding to each according to their system SQN:
System (x) = SQN (x) / (SQN1 SQN2 + + + SQN5 SQN4 SQN3 + + SQN6)
System 1 = 14% → € 28,000
System 2 = 12% → € 24,000
System 3 = 16% → € 32,000
System 4 = 22% → € 44,000
System
5 = 12% → € 24,000
System 6 = 23% → € 46,000
Logically, this asset allocation is dynamic, so we should study a timetable allocation with a frequency that is comfortable and adaptable to changes in the equity curve of the portfolio and the SQN of each system.
With these quantities and considering the maximum and DD are required, we will set the number of contracts to operate each system. In this case, choose a conservative approach:
No. Contracts = capital allocated / (DD. max * 2) +
guarantees should not be too strict, would be rounded to the nearest month, since the risk margin and the effect of portfolio diversification permit:
This So we have:
System 1 = 28,000 / 9,000 = 3 contracts.
System 2 = 24,000 / 13,700 = 2 contracts. System
3 = 32,000 / 34,700 = 1 contracts. System
4 = 44,000 / 14,400 = 3 contracts. System
5 = 24,000 / 9,900 = 2 contracts. System
6 = 46,000 / 25,220 = 2 contracts.
Two final considerations:
1) There really understand the limitations SQN ratio before use as an instrument of MM. The conscious (Media R / R deviation) is, in itself, an indicator of quality system. Its value will fluctuate dynamically as it moves the number of operations. But the multiplier Root (n) always tends to increase with time. The inclusion of this last element in the equation, reflecting the philosophy that the more operations are used in calculating the SQN, the more reliable are its results. But not always be that way. It may happen that we have fewer operations on some systems by the simple fact of having a smaller historical or because such systems have been incorporated into the actual operational later (if we use real data). When operational cadence of two strategies is similar, this will produce a bias in the results should be corrected.
2) What in principle is a methodology to allocate assets to each system in terms of quality, can easily become an alternative tool to position sizing. Enough to have everything in a simple Excel sheet and repeat the calculations on a monthly, weekly, etc.. depending on the cadence operational systems and the evolution of the profit curve. Do not forget to have updated data on the guarantees required by their broker, especially in times of great volatility how are you where you are being revised upwards or downwards continuously.
0 comments:
Post a Comment