I compile statistics on my traders. My best trader makes money only 63 per cent of the time. Most traders make money only in the 50 to 55 per cent range. That means you’re going to be wrong a lot. If that’s the case, you better be sure your losses are as small as they can be, and that your winners are bigger.
– Steve Cohen of SAC Capital, via Stock Market Wizards
Winning percentage, or ratio of winning trades to losing trades, is one of the biggest stumbling blocks for new traders.
This problem stems from inexperience and human psychology. Namely, a lack of experience attunes the new trader’s ear to what sounds good at first blush, rather than what actually works or makes sense
Many “too good to be true” type trading systems are thus sold on the basis of incredibly high winning percentages, or “hit rates” – often 80% or better, backtested over some carefully selected time period. Because the human psyche is naturally tilted towards loss aversion, this is the kind of thing that appeals.
Neither the ego nor the pocketbook like being wrong. So, the beginner’s logic goes, the best way to get rich as a trader is to be right (i.e. to book winners) as often as humanly possible.
Unfortunately, this way of thinking is a trick and a trap – an expensive indulgence of our primal wiring.
Side note: If you are running a high frequency trading (HFT) operation executed by a row of liquid-cooled supercomputers six blocks from the exchange, the above may not apply to you. But HFT is incredibly hardware and R&D-intensive — almost more of an arms race than a trading methodology – and thus a separate kettle of fish.
The reality is that win percentage, good or bad, tells very little about the year a discretionary trader may have had – especially in relation to varying levels of conviction and position size.
Here are two opposing examples:
- A cautious discretionary trader fights his way through choppy, low conviction conditions for nine months, trading small and light the whole time – then loads the boat with large, high-conviction positions in the final twelve weeks of the year. He finishes with a subpar total win percentage (compared to his long-term average), but an excellent net annual return.
- A naked options seller employs his aggressive premium collection strategy to great success, making money every week for months on end. After batting a thousand for 40 weeks in a row, one of his bread and butter stocks – that of a large, liquid corporation – is blown to smithereens on news of an earnings restatement. The unhedged put options the trader sold short, a good-sized position, suddenly skyrocket in value. With a hit rate (win percentage) still well above 90%, the trader’s year has gone down to the toilet.
To further reflect on this topic, the below Q&A is adapted from a recent Elite Trader message board discussion, held between yours truly and some other traders, as to the benefits of high winning percentage versus lower winning percentage strategies.
The starting point was an assertion made in Peter Brandt’s book that, other things being equal, knowledgeable traders would prefer a methodology with a win ratio of 30%, over a similarly performing strategy with a win ratio of 70%. Novice traders go the opposite way, focusing on the appeal of being right.
The questions in bold are condensed versions of message board inquiries. My edited answers are in plain text.
Q. Assuming comparable returns, why would professional traders prefer a methodology with a 30% success rate over a 70% one?
I would characterise the argument a little differently. It is more that the methodology with the 30% hit ratio will be more robust. It will be more armoured against losses, since routine losses and aggressive risk control in respect to extended loss periods are already part of the program.
This 30% preference is also based on a rule of thumb awareness of how profits and hit ratios are typically distributed. In Stock Market Wizards, Steve Cohen is on record saying his best trader was only right something like 63% of the time. In other words, 63% is given as the high-end outlier among the sharpest group of short-term traders in the world.
Given the above, and given that markets have a historical tendency to distribute profits in a very “lumpy” fashion — often rewarding a trader in a handful of great periods a year — the lower win ratio also has a higher likelihood of robusticity in terms of conceptually fitting the profile of how gains are distributed.
To further back up the distribution argument, Kenneth Grant, author of “Trading Risk,” was a risk manager for many of the top dog managers at one point — Jones, Cohen, Bacon and so on. In his book Grant remarks offhandedly that the vast majority of great traders he worked with had a 90/10 profit distribution, meaning 90% of their bottom line profits come from 10% of their trades.
This in turn suggests a profile of boxing and counterpunching for very small amounts in play, then “pouring it on” in those isolated instances when all the stars align. And once again this is a 30% profile, not a 70%…
But to some degree the even more straightforward observation, which I agree with, is that newbies are routinely bamboozled by the notion of a high winning percentage statistic. It’s not relevant to long run success and may even be detrimental if undue emphasis or leverage is placed upon it.
Q. Doesn’t a lower success rate leave less room for error? Wouldn’t the greater accuracy of a higher winning percentage confer valuable long-term benefits?
I would argue the cost of greater accuracy is a diminished opportunity set (relative to catching meaningful moves in deep and liquid markets with a meaningful amount of capital in play). This is because the biggest and most powerful trends are most often born in periods of high uncertainty.
How do you catch a monster move born in uncertainty? You take multiple shots with small controlled risk and lever up opportunistically as conditions dictate. The real world manifestation of this is a series of small, low-risk probe type opportunities coupled with the profits from major outlier moves.
I think one challenge for many new traders (not speaking to anyone in particular) is that they don’t give a lot of thought to how profits are distributed, or what a “loss” really means. By many professional traders’ estimation, a small controlled loss is not necessarily a failure or a setback so much as an investment.
Q. Doesn’t a lower success rate imply greater dependence on not missing a trade?
Yes and no. The question of dependence is debatable based on signal frequency.
For example, if you know from statistical records that you take 120 trades a year on average, you will certainly have a large enough opportunity set for the law of large numbers to work in your favour.
Where N equals number of trades, lower average hit rate increases the odds of a dry spell for shorter periods of N, but as N extends, probability closes the gap.
Q. Isn’t it logical to assume traders with higher win percentages will have consistently stronger profitability, just as a company with a more diversified client base is more likely to have stable financial performance?
The Bear Stearns High Grade Structured Credit Opportunities Fund had positive returns for 40 months in a row. Then it blew up. Vic Niederhoffer has had more than his fair share of long stretches of net positive return months — followed by repeated blowups. Merger arb guys are known for “eating like birds and shitting like elephants” or otherwise “taking their volatility all at once.”
In other words, the flip side of the argument is that artificially high stability tends to sacrifice robusticity. There is the danger of the relied-upon mechanism that produces that stability suddenly going poof. The real world instances of this are legion.
I would further argue the client diversity argument is misapplied. Simply put, lumpy is more lifelike. The trader with the “steady eddie” methodology that distributes gains with preternatural smoothness is often relying on one specialised approach, or one market, or one prevailing set of market conditions. He is the one more exposed to potentially disturbing change.
Strategies with lower hit rates and higher R multiples, in contrast, almost by definition tend to work across a wider number of markets and a far more diverse set of market conditions. Being able to exploit opportunities in, say, 30 different markets is thus akin to having 30 different clients. Very different than the guy who makes all his money in merger arb, or scalping the DAX, or some other very narrow strategy as high hit rate approaches tend to be.
Q. In addition to the “more dependent on winners” problem, won’t the lower winning percentage trader have a harder time determining when his strategy has gone off the rails? If extended periods of adversity are nothing out of the ordinary, how will he know?
The “single winning trade” argument is invalidated by a sufficiently large sample size of N. It may be true that a certain strategy has higher odds of a losing month, for example, but that does NOT automatically translate to higher odds of a losing year, because 12 months is a sufficiently expanded length of time for N to fully express historical averages.
The 70% winning strategy, in contrast, almost always requires significant sacrifices to maintain — usually in the form of lower R / higher leverage / reduced profits per trade. That is why high hit ratio strategies tend to be the province of day trading, merger arb, options selling, and so on, all of which present their own special challenges — not least capacity constraints and “flash crash” risk (as greater amounts of starting leverage typically have to be employed to make the light worth the candle).
Re, going off the rails, the short answer is this is a non-issue for discretionary traders who are not reliant on mechanical signals or heavily optimised system parameters. For a slightly longer answer, this also goes back to robusticity and diversity of opportunity — the more markets a strategy can trade, and the wider opportunity set of conditions it can handle, the lower the odds of permanent degradation (regardless of whether the strategy is discretionary or mechanical).
Again too, based on simple empirical observation of successful traders running meaningful amounts of capital ($10MM plus) across the market landscape, the strategies that seem to have the most diversity, longevity and ubiquity over time tend to trend towards a focus on higher R in concentrated profit periods, a minimized or ignored focus on winning % as a key statistic, and acceptable levels of daily and monthly volatility deliberately incorporated into the program.
Q. Are you saying lower winning percentage is a prerequisite for success?
No, I’m not arguing lower winning percentage is a necessary condition for success — more so that it’s a very regular and routine occurrence. One could certainly still have a high hit rate for an extended period of time, and one of the large global macro traders in “Invisible Hands” says his long term hit rate is 50%.
The abstract logic goes back to 1) near term uncertainty, 2) probing bets, and 3) the lumpy manner in which profits are distributed.
If you use a tighter risk point, you can establish a position with larger size. But doing that increases the likely number of times you will have to test and probe before catching a meaningful move. Outlier profits then come on moves that are extra large, or offer high quality pyramiding opportunities, or otherwise provide the perfect moment to go full throttle with tightly defined risk.
Q. What would you say is the maximum R, or reward-to-risk expectation on a trade, for a strategy that is profitable 63% of the time (the Cohen example) versus one that is profitable 30 or 35%?
Well, that’s the beauty of being a discretionary trader — there really is no “maximum” to speak of, any more than there are maximum earnings for an entrepreneur.
Let’s say a (very good) trader is able to maintain a 63% hit ratio through extreme precision and tight risk control, but is also very, very good at spotting those rare outlier moments when all the stars align for a huge trend.
Then say that same trader establishes a huge position in Ford Motor Co. (F:NYSE) at $2 per share in early 2009 and, thanks to a cushion of accumulated profits, decides to open up his risk envelope, riding it all the way to $13 per share. Given a gain of $11, what would his R be if he originally risked 20 or 30 cents on the trade?
There are also other ways to structure trades as such that, in certain circumstances, an outlier R can approach 50 to 1 or some such (or even a lot more). So it doesn’t really make sense to talk about a cap or a max, assuming the versatility is there.
Of course, I am talking about discretionary methods here rather than mechanical (or discretionary implementation with mechanical augmentation). But this also helps explain why the most profitable traders in the world are discretionary.