The four horsemen of apocalyptic position sizing used by professional investors
Despite picking a fair share of good stocks, it is still tough to generate some consistent serious alpha. Picking the right stocks and exiting them well tells You how often You win. How much You win, however, is a function of how much You bet. Some professional investors pay surprisingly little attention to their bet sizes. Below are four algorithm often practiced by professionals that can
- Four popular bet sizing algorithms used by professionals that have negative gain expectancy
- Size does matter in the markets: 1$ or 100 will have a different outcome
When it comes to bet sizing, there are only two sizes: either too much, either too little. As a professional short-seller, position sizing is mission critical. Successful positions shrink. Not only do they contribute less and less, but they also tilt exposures (net & net Beta). To add insult to injury, they become less noticeable. On the other hand, unsuccessful positions balloon. They immediately hurt. So, I have spent years studying the science of bet sizing. I sought to learn from other investment professionals. It eventually dawned upon me that Long biased people rarely ask themselves the same questions. For them, bet sizing does not have the same degree of urgency. Worse even, it became apparent that some position sizing algorithms had outright negative expectancy, or nasty side effects that they were never even aware of.
Horseman 1: Liquidity. If You can’t get out, You don’t own stuff. Stuff owns You
Getting into a position is like buying a boat, or a second house. You can do that any day of the week. Now, selling a boat is tough (been there, done that). It may take time to build a position in a stock. Time is an expensive luxury few market participants can afford when they want to liquidate.So, no matter how attractive a story may be, if you can’t exit easily, just don’t enter.
Rule 1: don’t size your positions so that they may go Hotel California on You:
“You can check-out anytime You like, but You can never leave!”, Don Henley, Hotel California
Horseman 2: High conviction: feel-good position sizing
Disclaimer: this position sizing is used by the greatest and the worst investors. The classic rationale is: “if You believe in something, then you should go big or go home”. What else is it but a feel good position sizing algorithm ? Risk is not quantified but subjectively assessed. The problem is mental accounting, or the constant emotional revisionism of the situation. Jack Welch said: “what can be measured can be improved”. If You can’t quantify your risk, then don’t expect improvement in consistent alpha generation capability.
The greatest investors also use conviction as a position sizing algorithm. The only difference is that they express conviction in units of risk. They quantify risk first and then put chips on the table according to their perception of the reward. If an idea does not pan out, risk can be parred down.
Horseman 3: Equal size: one-size-fits-all and the volatility roller coaster
This position sizing algorithm will not bring ruin, but it has negative side-effects that may prevent You from achieving your obejctives in terms of performance, attractiveness to investors and quality of life…
Equal weight is a form of laziness:
First, let’s look at the math behind equal weight. All trading systems boil down to their trading edge (Avg Win% * Win% – Avg Loss% *|Loss%|). Since all bets are equal, equal weight implicitly puts emphasis on the signal, and excludes the value of money management. In other words, stock picking has to be consistently above 50% to absorb losses and generate a profit. Unfortunately, no system works all the time. So, equal weight carries cyclicality in performance.
Ignoring volatility at the position sizing level invites volatility in the portfolio
Not all stocks have the same personality. Some are more volatile than others. For example, internet stocks tend to be much more turbulent than utilities. If all positions are sized equally, then the most volatile stocks will drive the volatility of the overall portfolio. Morality, ignoring volatility at the position sizing level will in turn invite volatility in the portfolio.
Horseman 4: Average down, martingale and the certainty of ruin
Rookie gamblers always come up with some elaborate scheme to break the casino. It is usually a variation on the theme of doubling down after each loss. They believe that the losing streak will end and they will recoup their losses. This position sizing algorithm is known as martingale. Let’s look at the math behind this algorithm
1. Adding to a losing position reduces the hit ratio
2. even if You had infinite capital, the most favourable outcome would be break-even. First, do You have infinite capital ? Second, any other outcome before the most favourable one carries an interesting probabilistic property called certainty of ruin
3. Doubling down means adding to losers. Resources have to come from somewhere, probably winning trades. Books written by successful market participants always emphasize “cut losers, ride winners”. Do You know any successful market guru who says “cut your winners, ride your losers ?”
In conclusion, there is a reason casinos have gold, marble, paintings from masters and gamblers declare bankruptcy. Double down on losers and You will go broke. One more thing about probabilities, it’s not about if, it’s about when.
Size does matter in the markets. Not paying enough attention to position sizing has consequences that range from unpleasant volatility to certainty of ruin. Position sizing is not a glamorous topic, but in highly competitive sport, every little bit of edge counts
Hi Laurent – Just curious about how you calculate your pos size?
I do use a lazy method, fixed % but then adjusted for volatility (ATR Precent)
Open to and ideas you are willing to share on how to work the funds better/harder
Before we move on to my small story, would You like a position sizing algo on Excel that would be mindful of volatility, risk ?
I am a short seller. The problem with shorts is that they balloon when unsuccessful and shrink when they work. So, position sizing is always a thorny issue.
There are two types of personalities when it comes to sizing: conservative or aggressive.
I have always wanted an algo that could at times be conservative and at times more pro-active.
Then, the math dawned upon me. I was using one side of the equation (risk per trade) while I could use the other side (capital allocation).
Weight = risk per trade * Capital equation
Using a car analogy, the idea is to have one side be the accelerator (risk per trade) while the other one would be the brakes (capital allocation).
Now, like a car, the trick to reduce gas consumption is to use the accelerator judiciously enough not to have to use the brakes.
So, risk per trade would be sensitive to NAV fluctuations: increase when NAV rise and vice versa.
Meanwhile, capital allocation would have to be unresponsive enough to weather small fluctuations but really reduce size if something material happened.
I wish I could post pictures on this comment, but they both are convex. One is upward slopping, while the other is downward slopping.
The combination of this is the answer to the age-old problem: be conservative when things aren’t working and aggressive when it works.
For example, at the moment, my system does not do very well, so it trades at minimum risk over a reduced capital. The trading risk is somewhere around -0.10%.
Should performance increase by +1%, risk per trade would still be depressed, but capital allocation would not be as strict and trading risk would revert back to -0.15% to -0.2%.
This elasticty is useful during drawdowns and allows quick recovery. It materially increased my trading edge by over 1/3.
I hope I answered your question. If You need more information, I can dig out the Quora post or write a new one on the Blog
Thanks for the offer of the spread sheet, that would be great if you can email it over.
I use trend following long, mean reversion long and two shorting systems so I have ran into much the same problems as you.
When the system is working you want to push it hard but when it is not you want the accelerator taken off.
Good interview on Better System trader as well.
Hello James, This was the post on Quora
I’ve found your writings fascinating. Can I also get hold of the spreadsheet, that you offered to James? Would this type of risk management work on a portfolio of assets?
I’m interested in hearing more about the habits of successful traders, if you plan to write on that subject.
In another note you have presented a calculation to highlight the edge that a system may contain. Could that equation be made more accurate by including the variance of losses and profits? Or, am I overcomplicating?
I also noted your
Many thanks for the kind words.
As for the spreadsheet, 1. please follow the link 2. Register 3. Download the file
If You experience any difficulty, please send me a mail at email@example.com and I would be delighted to help You
About habits, thank You for the suggestion. A few posts are coming. I am also quite active on Quora
On this website, please read the posts about the game of two halves and the game of two thirds. They are designed to mechanically improve your trading edge. By the way, they tie in with the file You are about to download
As for variance of Loss and Profits, the above two games synthetically achieve the same results. One word about variance though, it fails to capture the inherent unpredictability of the markets. For example, the massive move in August took a lot of funds by surprise. It is rare but it can happen, and if it can, it will.
Once again, thank You very much
As for variance of losses, the objective was to build a simple distribution of profits and losses
Thanks for taking the time to reply to my queries.