止損線可能就是你的生命線
止損線可能就是你的生命線
一天,他已見到有12隻火雞在他的籠子裡。接著有一隻火雞溜出來,剩下11隻。他想:“剛才有12隻的時候我應該拉繩子關籠子的。也許再等一下,它還會走進去。”當他等這第12隻進籠子時,又有2隻火雞走出了籠了。他又想:“我應該滿足,有11隻也好,只要再有1隻火雞跑回籠子,我就拉繩子。”當他在等機會時,又有3隻火雞跑了出來,結果他落得個雙手空空。他的問題就在於他無法放棄原來已快到手的火雞數目,想像其中一些還會跑回去。這種態度正是典型的投資人無法認賠賣出的情況。他不斷期望期貨能恢復原價。教訓是:為了減低市場風險,停止數你的火雞。
一筆交易必須設立一個止損點。“止損線可能就是你的生命線”。在市場面前,不可能永遠判斷正確,人人都會遭受失敗,為此我們要面對失敗。
除了投資交易這一職業,世上也許沒有其他職業需要你承認錯誤。索羅斯深諳風險的重要,因此他成為馳騁世界的金融大鱷。李森固執自己的經驗和教條,一味追加投資,妄圖扭轉大勢,便搞垮了巴林銀行。我們沒有止損而丟失金錢一定能堆起一條哈大公路。必須要學會如何輸錢,這比學會如何贏錢更為重要。事實上,即使你只對一半,你也應該贏。關鍵在於每次搞錯的時候要反損失減到最少。
我們不是神,我們是人。我們應記住一位金融巨頭的忠告“警悚者生存”。既然高利潤對面是高風險,我們就應在可怕的風險前面划上一條止損線,它可能就是我們的生命線。
國外頂級的資金管理
下跌市場補倉三大訣竅
Playing the Market's Money 玩市場的錢
假設你的目標是在未來一段時期內實現收益的最大化。只要不虧及你的初始資本,你願意為達到收益的快速增長而做任何事情。在此前提下,你可以設計一個特殊的系統,這個系統在初始資本上只冒很小的風險;但在“市場的錢”上冒最佳比例的風險。
舉個例子,假設你在1月1號以$100,000作為初始資本。你的目標是到12月31日時,以初始資本的最小損失風險換取盡可能多的收益。那麼你應該這樣做:
你應該以初始資本的1%來冒險,而在市場的錢上以最佳比例冒險(或近似地)。讓我們假設你的系統的最佳風險比例為10%。
這個系統真正的優勢是什麼呢?一旦你有了利潤儲備,照此方法去做,你的利潤增長潛力會奇跡般地被放大。讓我假設你的第一筆頭寸是原油期貨。你的初始投入是$100,000 * 1% = $1,000。此時你的第二筆交易來臨,你已經在原油上有了$3,000的賬面利潤。現在,你可以投入的資金為,來自初始資本的$1,000,再加上賬面利潤的10%,即$300。因此,在同樣的系統下,你的第二筆交易可以投入的資金為$1,300。
假設你對這種模式已經玩得很溜了。到三月份,你已經積累起$25,000的利潤。此時,你的潛在投入量為$1,000(初始資本仍然的1%),加上利潤$25,000的10%,也就是另外的$25,00。共計$35,00,占總資本($100,000 + $25,000)的2.8%,冒的風險比原來的1%翻了一倍多,但這並沒有危及你的老本(還是1%)。
“薄利多銷”和“暴利”原則的分野
注碼法的價值
從倉位大小看收益期望值的穩定性
請你拿出計算器機算一下,這3種情況的總贏率和總賠率的大小。
10次 100次 1000次
總贏率 50.5% 82% 99.92%
總賠率 49.5% 18% 0.08%
總贏率=所有贏利概率之和
總賠率=所有失敗概率之和
我們可以看到,當成功率為55%時,在10次贏1元的情況下,成功的總概率為50.5%,風險是相當大的。在100次贏1元時,成功的總概率為82%,這種情況還可以投機。在1000次贏1元時,成功概率高達99.92%,可以完全保證長期贏利。
如果把1 元換成1萬元,在同樣的成功率為55%的贏利水平的市場裡,大倉位可以在較短的時間內賺到1萬元,而小倉位要在更長的時間裡才能賺到1萬元。沒有耐心的人,平時沒有大機會的條件下習慣滿倉殺進滿倉殺出,風險還會大幅增高。如果在機會來了時,分倉慢慢買入,心平氣和,做對了行情時逐步正金字塔加倉,做錯了行情時及時斬倉或者減倉出來,既減小了風險,又能保證穩定適當的利潤,這是比較好的最普通的交易策略。
本實驗的好處在於,它把其它一些復雜的因素隔離開來,只考慮了倉位因素對收益穩定性的影響。事實上,做對了盡量持有和正金字塔漸步加倉,做錯了盡快斬倉或者減倉,贏利概率還會大幅提高。除此之外,對資金管理的其它策略也能大幅增加收益,減少風險。
以時間為代價換取收益的穩定性的辦法應該作為職業炒手的一種意識。根據行情客觀情況來決定輕倉駕御還是重拳出擊,這要根據自己的水準和經驗來決定,每次都採取輕倉的策略並不算合理,有沒有這種輕倉意識才是最重要的。當有大行情大機會來時,為什麼不重拳出擊呢?問題在於不能習慣性地天天重拳出擊!
國外大基金投資非常穩健,年收益一般在20%~30%已經很了不起了,它們更強調穩定的收益策略,不在一時的大贏。短時的大贏策略容易帶來短時的大輸,必然造成資金帳戶的極大風險,對於任何基金都是十分危險的舉動。分散小規模投資也許是最好的辦法,剛好體現了本實驗的實質所在。在同一個市場或者相關性很大的市場裡分散投資效果不好,對於有較強相關性的品種之間搞分散投資將失去分散投資意義,這恰好是投資界不太明確的地方。
最後提醒,投機是一項複雜的活動過程,不是僅依靠倉位就能解決投機中所有的問題,倉位僅是一項不能忽視的重要問題。
The Foundations of Money Management
I've found such a system. With numerous tests it almost never had under 90% profitable trades. The results of one such a test are given in Table 1 in Omega Research TradeStation format. The code for the system is in Appendix 1; you may copy it to Omega TradeStation or SuperCharts and go along winning (in the sense they usually mean winning, that is, having a profit on most trades). The system's main secret is a pseudo-random number generator (too "pseudo" in TradeStation, but doesn't matter much). Then it all goes as usual: if the position is profitable, close it. If the market goes against us, turn investors. Having enjoyed working and socializing with customers of two brokerages over a couple of years, I can insist that is just what most traders do - except the fact they formally replace the random number generator with analytic forecasts, indicator signals, the neighbor's opinion in the pit or just a momentary impulse. The problem is that winning at an exchange and earning money at an exchange are far from being the same.
Surely, the profit seen in the Table 1 example is casual, a result of a lucky dice roll, whereas it would not be profitable in most cases. But if one changes the system entry parameters to more reasonable levels, i.e. sets mmstp=1, pftlim =4, maxhold =10, this will make the system profitable in most tests.
So exploiting the principal idea of speculation - close losing trades fast and let profits grow - combined with money management allows to earn money even from random trades. Most people act just opposite to this principle; they let losses grow, hoping the market turns and proves how right have they been, and quickly close their profitable positions to prove how right they're at the moment. Most beginners and many self-styled pros, as our experience shows, are sure that the skill of market forecasting equals the ability to earn money at the market. Getting a profit on a given trade for them means proving their prognostic abilities and, consequently, their skill in making money.
A person unfamiliar with trading as a business could be puzzled by the fact that "successful investing and trading have nothing in common with forecasting"*. There is bad news and good news. The bad news is: markets cannot be prognosed. The good news is: one doesn't need to do that to have profit. We are concerned not with getting a profit on every trade, but on making large sums when we're right. The number of profitable trades may in this case be less than losing, that is, it is possible to use worse-than-random forecasting!
As a famous trader Paul Tudor Jones said: "I may be stopped four or five times per trade until it really start moving". That is, Paul may win only on a measly 20-25% times! Yet he'd had three-figure (percents) of income in five consecutive years with very low capital corrections1. Almost 100% of Steve Cohen's very large profits are taken off 5% of trades, and only 55% of his trades are profitable at all. Despite that in the last seven years he'd made 90% per year on the average, and had only three losing months (the worst losses were -2%)2.
The widely used by professional methods of trend following, as a rule, bring about 30-40% of profit. Profits or losses in any given trade do not matter - as long as the amount of money earned per average trade is positive. This value is called mathematical expectancy. The mathematical expectancy equals the sum of products of profit probabilities minus the sum of products of losses probabilities, multiplied by the losses' size
Simplified, the expectancy may be estimated as the probability of profits multiplied by the average profit minus probability of losses multiplied by the average loss. In terms of the Omega Research TradeStation this looks like:
Table1.
Total Net Profit | $562.70 | Open position P/L | ($75.60) |
Gross Profit | $1,269.40 | Gross Loss | ($706.70) |
Total #of trades | 276 | Percent profitable | 92.75 % |
Number winning trades | 256 | Number losing trades | 20 |
Largest winning trade | $54.90 | Largest losing trade | ($126.50) |
Average winning trade | $4.96 | Average losing trade | ($35.33) |
Ratio avg win/avg loss | .14 | Avg trade (win &loss) | $2.04 |
Max consec.Winners | 39 | Max consec.losers | 2 |
Avg #bars in winners | 1 | Avg #bars in losers | 17 |
Account size required | $177.30 | Return on account | 317.37% |
In a newsgroup discussion one follower of Elliott's theory said: "Market is no gambling - we make no bets". Not being an Elliott adherent, for whom everything is pre-arranged, we do make bets. Since the result of any trade is unknown, any trade is a bet where we win or lose a certain sum. The principal difference between gambling (betting) and market trades (speculations) is first, that gambling creates its own risks and speculations re-distribute the risks already present on the market; second, the on a market a trader is able to provide himself with a statistical advantage, that is, a positive expectancy.
Appendix 1. A system giving over 90% profitable trades. {********************************************************* Random System №1. Copyright (c)2001 DT Parameter values by default: mmstp =1,pflim =4,maxhold =10 **********************************************************} Inputs: Bias(.025), {Random entry parameter} mmstp(100), {Stop loss parameter} pflim(.1), {Profit target limit} maxhold(50); {maximum holding period}; Var:Trigger(0),Signal(0),ATR(0),num(1); trigger =random(1); if trigger < bias then signal = -1; if trigger >1 - bias then signal =1; ATR =XAverage(TrueRange,50); { Random Entry} If signal =1 then Buy("Random_Mkt.LE")num contracts next bar at open; If signal =1 then Sell("Random_Mkt.SE")num contracts next bar at open; { Standartized Exits} if marketposition >0 then begin ExitLong ("MM.LX")Next Bar at EntryPrice -mmstp*ATR stop; ExitLong ("Pt.LX")Next Bar at EntryPrice +pflim*ATR limit; if barssinceentry >=maxhold then ExitLong ("Hold.LX")at close; end; if marketposition <0 then begin ExitShort ("MM.SX")Next Bar at EntryPrice +mmstp*ATR stop; ExitShort ("Pt.SX")Next Bar at EntryPrice -pflim*ATR limit; if barssinceentry >=maxhold then ExitShort ("Hold.SX")at close; end; |
Appendix 2. The simplest system number 2. {************************************************************* The Simplest System №2. Copyright (c)2001 DT **************************************************************} Input:Price((H+L)*.5),PtUp(4.),PtDn(4.); Vars:TrendLine(C),LL(99999),HH(0),num(1); if MarketPosition <=0 then begin if Price < LL then LL =Price; if Price cross above LL +PtUp *.001 then begin buy("Simpl.LE ")num contracts next bar at market; HH =Price; end; end; if MarketPosition >=0 then begin if Price >HH then HH =Price; if Price cross below HH -PtDn *.001 then begin Sell("Simpl.SE ")num contracts next bar at market; LL =Price; end; end; |
Appendix 3. Data output to a file to compute mathematical expectancy {************************************************************* Expectancy Output Copyright (c)2001 DT **************************************************************} Var:RMult(1),R1(1),Trades(0); Trades =TotalTrades; R1 =PctUp *.001 *BigPointValue; RMult =PositionProfit(1)/R1; If barnumber =1 then print(file("D:\TS_Export \M trading.csv"),"Qty",",","Profit",",","Initial Risk",",","R multiple"); If Trades <>Trades [1 ]then print(file("D:\TS_Export \M trading.csv"),Num:10:0,",",PositionProfit(1):10:4,",",R1:10:4,",",RMult:10:4); |
To be just we should mention that it is possible to create a "gambler's advantage" - so a mathematician Edward Thorp has developed strategies with a positive expectancy for playing blackjack, which he'd successfully used in Las Vegas gambling houses. When they stopped letting him in, he published his methods1, after which blackjack rules had to be altered to remove the gambler advantage. In late sixties Thorp took interest in shares market and became a manager for a private investing partnership: " Our significant rival then was a Harry Markowitz, a future Nobel prize winner. After 20 months we had +39,9% profit compared to Dow Jones' +4,2%. Markowitz went negative in a couple of years, and we're satisfied with our stable results… about 20% yearly (standard deviation around 6%0 and zero correlation with the market".
Besides expectancy, most traders have problems understanding risk. For instance, a historian by education, (former) head of a regional investing company with assets over a million dollars by summer 1997 was sure that "risk doesn't exist so it cannot be measured" and also sure that "one shouldn't sell shares at a loss". What can one say about amateurs then… Risk does exist and it can be measured. It is considered that risk is a volatility measured as the standard deviation of the changes of actives traded. This holds true for investing risk, speculative risk is more adequately defined as standard deviation of capital changes. By both those definitions risk is heavily underestimated. According to Murphy's laws, the worst is yet to come; We shall employ the following definition: risk is the amount of money we are ready to lose before withdrawing from a losing trade.
Before opening a position it is necessary to define the point where we close the position wit a loss to save capital - the so-called stop loss1, or where we open an opposite position, having made sure of our mistake concerning the market direction - the so-called stop-and-reverse. The difference between the entry point and the stop loss point multiplied by the number of lots is the starting risk or 1 R2, independent of how and in which units we measure the stop level, be it dollars, percents, volatility units or six-packs. This definition of risk is not equal to the first definition - the risk may be many times the 1 R if the stops are not executed due to lack of discipline3, gaps against the position or unexpectedly high slippage. The profit, then, can be defined in units of risk per share or in multiples of R. In terms of multiples the basis rule of speculation will be formulated as: keep losses at the level of 1 R as long as possible and let profits reach many times R.
The expectancy in multiples of R will mean how much can we win or lose per unit of risk in an average trade. To calculate expectancy in terms of multiples of R we must place the results of our trades in a table with the following columns:
Number of lots | Profit or Loss | Starting risk | Multiple of R |
The Profit or Loss must take into account broker commissions and slippage. Multiple of R is calculated by dividing the second column by the third. Then to calculate expectance it is enough to add up the values of the fourth column and divide by the number of trades. This method is also works with "intuitive" trading.
So, we do have a winning strategy - what next?
We can open a brokerage account and bet all our capital with the maximal leverage.
Here the most important thing - the money management begins. To clear the situation here is a pair of facts. Ralph Vince invented a game, where bet size was the only moveable parameter. He chose forty doctors of sciences (i.e. not the dumbest people at least) as players, none of which were professional traders or studied statistics. The doctors played a game where 100 random trades were generated, one by one. Every one began at $1000, and before every trade one had to make a single decision - how much (up to 50% of the capital) to bet. 60% of the time the players won their bet, and 40% of the time they lost their bet. This game has an expectancy of 20 cents per dollar risked, i.e. in the long run the player can receive 1 dollar 20 cents per dollar. The academicals made their 100 bets, enough to resolve the expectancy. Making the same trades, they finished the game with different results. Guess how much of them increased their starting capital? Two of forty. 95% of doctors lost money playing a game with a positive expectation!1
Van Tharp made an even more striking example. In an Asian Tour for Dow Jones Telerate TAG (Technical Analysis Group) he gave lectures in 8 cities before 50-100 listeners each time, most of them professional traders for large companies or banks that traded shares, bonds or exchange rates on Forex. In an analogous game over a half of highly professional traders lost!2 Another personal example - a trader offered a similar game to a friend employed by Charles Schwab as a leading analyst. At the first level the distribution of multiples of R with an expectancy of 0,45 and 60% profitable trades. To get to the second level one had to make 50% profit in 100 trades. The result was "I cannot get to level 2 in a day!"3. In 1991 Brinson, Singer and Beebower published a research of the efficiency of 82 portfolio managers in a 10-year period, which showed that 91,5% of all profit was generated by asset distribution3. The asset distribution meant the division of capital between cash, shares and bonds. Only 8,5% of profit was due to buying and selling the right stocks and bonds at the right time.
Let us play the game described by Vince. If there was no risk, i.e. we knew the result of each trade beforehand, it would make sense to bet all the capital each time. So every player would have gained $1000 ..(1.2 ^100)=$82,817,974,522.01 .
In reality, if we bet all $1000 on the first trade, we have a 40% risk to lose all at the first attempt. Even if we win and have $2000, betting all on the next trade would be exactly as insane.
Now suppose we bet $200 at a time. So if five first trades are losing, we again lose all. The probability of such an event is small, just over 1%. But are we ready for such a "small" risk, if we can lose all the money? Suppose we lose in the first two trades (16% probability), so we'd lose 40% of the capital. Beginning from the next trade we must gather 67% of profit just ot restore the starting capital. This effect is called "asymmetric leverage"5.
Table 2 shows that loses of over 50% need improbably large profits just to recover; so if we risk relatively large sums and lose our chances to end up wit a profit are negligible.
Suppose that we bet a certain percent of our capital and record the current capital after each trade. Repeat the 100-trades sequence again and again, and after a lot (1000 or so) series we'll be able to estimate the distribution of results. Evidently, we'll have different end profits, since the game is random-based. This is called Monte Carlo modeling.
Let us arrange the 1000 profit performances from 1000 series from smaller to larger. Then let us divide this range into 100 parts with equal number of variants in each - so every such a percentile will have 10 variants of performance. The first percentile will contain 10 worst results, and its top limit (number 10) will correspond to what they usually formulate as: "In 1% of cases the results will be inferior to… value". Statistically this percentile is called k-1. The border of the 50 percentile (k-50) would correspond to: "In 505 of the cases the result will be inferior to…"
Table 3 displays the outcomes of the 1000 series with different bet sizes in percents of the capital.
With 10% bet for each trade the minimal capital after 100 trades was 181,1$. In 1% of all trades our capital was under $405 (Profit k1). In 50% the trading yielded $4501 and less (Profit k-50). In 95% of cases the end capital was below $22411 (Profit k-95), and, corerespondingly, in 5% of cases the end capital was above $22411.
Let us review drawdowns (DD in the table). The drawdown is the difference between the maximal capital and its subsequent minimum before the new maximum is reached. With 10% bets in 50% of the cases the DD was over 48%, in 1% over 78% and the maximal DD was almost 90% of the capital. With bets over 30% of the capital we ape practically doomed to ruin. Once again we remind that this game has a positive expectancy - at win/loss probability 60% to 40% the win size relates to loss size as 1 to 1.
Steve Cohen says that: "the traders' general mistake is taking too large positions in relation to their portfolios. The, when the shares move against them, they are hurt too much to remain in control, they finally either panic or freeze in shock"1.
These examples described the importance of bet size in games with an undetermined outcome. So what is money management? An Internet search with those keywords yielded links to services for personal financial control, advices on handling others' money, how to control risk, on Turtle Trading, etc. According to Van Tharp, money management is NOT:
· a part of system that dictates how much you will lose in a given trade
· a way to exit a profitable trade
· is not diversification
· is not risk control
· is not avoiding risks
· is not a part of a system that maximizes performance
· is not a part of the system that tells where to invest
Money management is a part of a trading system that tells "how much". How many units of investitions should be held at a time? How much risk may be taken?
So, money management is controlling the bet size. Te most radical definition known to us is given by Ryan Jones3: money management is limited to defining what sum from your account should be risked on the next trade. Pay attention that this definition does not list as money management controlling the size of an already open position, which Van tharp allows.
Table2.
% loss | 10 | 20 | 30 | 40 | 50 | 60 | 70 | 80 | 90 |
% profit required to recover | 11,1 | 25,0 | 42,9 | 66,7 | 100 | 150 | 223,3 | 400 | 900 |
Table3.
Bet size | k-50 DD, % | k-99 DD, % | Max DD, % | Worst profit case | k-1 profit | k-50 profit | k-95 profit |
1.00 | 5.87 | 13.25 | 18.30 | 900 | 956 | 1.215 | 21.426 |
5.00 | 26.86 | 52.32 | 68.17 | 484 | 654 | 2.401 | 5.346 |
10.00 | 48.43 | 78.36 | 89.49 | 181 | 405 | 4.501 | 22.411 |
15.00 | 64.77 | 92.81 | 97.48 | 71 | 237 | 6.586 | 73.936 |
40.00 | 98.81 | 100.00 | 100.00 | 0 | 0 | 783 | 687.933 |
So, we do have a winning strategy - what next?
We can open a brokerage account and bet all our capital with the maximal leverage.
Here the most important thing - the money management begins. To clear the situation here is a pair of facts. Ralph Vince invented a game, where bet size was the only moveable parameter. He chose forty doctors of sciences (i.e. not the dumbest people at least) as players, none of which were professional traders or studied statistics. The doctors played a game where 100 random trades were generated, one by one. Every one began at $1000, and before every trade one had to make a single decision - how much (up to 50% of the capital) to bet. 60% of the time the players won their bet, and 40% of the time they lost their bet. This game has an expectancy of 20 cents per dollar risked, i.e. in the long run the player can receive 1 dollar 20 cents per dollar. The academicals made their 100 bets, enough to resolve the expectancy. Making the same trades, they finished the game with different results. Guess how much of them increased their starting capital? Two of forty. 95% of doctors lost money playing a game with a positive expectation!1
Van Tharp made an even more striking example. In an Asian Tour for Dow Jones Telerate TAG (Technical Analysis Group) he gave lectures in 8 cities before 50-100 listeners each time, most of them professional traders for large companies or banks that traded shares, bonds or exchange rates on Forex. In an analogous game over a half of highly professional traders lost!2 Another personal example - a trader offered a similar game to a friend employed by Charles Schwab as a leading analyst. At the first level the distribution of multiples of R with an expectancy of 0,45 and 60% profitable trades. To get to the second level one had to make 50% profit in 100 trades. The result was "I cannot get to level 2 in a day!"3. In 1991 Brinson, Singer and Beebower published a research of the efficiency of 82 portfolio managers in a 10-year period, which showed that 91,5% of all profit was generated by asset distribution3. The asset distribution meant the division of capital between cash, shares and bonds. Only 8,5% of profit was due to buying and selling the right stocks and bonds at the right time.
Let us play the game described by Vince. If there was no risk, i.e. we knew the result of each trade beforehand, it would make sense to bet all the capital each time. So every player would have gained $1000 ..(1.2 ^100)=$82,817,974,522.01 .
In reality, if we bet all $1000 on the first trade, we have a 40% risk to lose all at the first attempt. Even if we win and have $2000, betting all on the next trade would be exactly as insane.
Now suppose we bet $200 at a time. So if five first trades are losing, we again lose all. The probability of such an event is small, just over 1%. But are we ready for such a "small" risk, if we can lose all the money? Suppose we lose in the first two trades (16% probability), so we'd lose 40% of the capital. Beginning from the next trade we must gather 67% of profit just ot restore the starting capital. This effect is called "asymmetric leverage"5.
Table 2 shows that loses of over 50% need improbably large profits just to recover; so if we risk relatively large sums and lose our chances to end up wit a profit are negligible.