How many people saw the financial crisis coming and profited from it?

2007 Christmas party at a dear Friend’s apartment in Azabu, Tokyo. He was the head of MacQuarie securities in Tokyo. Champagne was flowing and i was passably drunk. There was a bunch of Lehman, Morgan Stanley, Citi and other dudes, all smart articulate people.

They were talking about the current uncertainty saying it was an opportunity to buy on dips. I rolled in saying this would be the first time sh!t was about to flow uphill. All those delinquent loans would not end well. The two things that people buy on credit are homes and cars. If home loans turn delinquent, then the rest would follow, contagion. I happened to be tracking Japanese car manufacturers US car sales and watched them plummeting. They laughed at me and concluded that i was a drunken lunatic. Point taken…

1 year down the road, December 2008 same party, half the people. The Lehman dudes were gone or at Nomura. One of the guys ran up to me. I was about to duck and hit the floor as if chased by the police: “no

-“No i don’t do drugs anymore, no i did not do your wife, no i didn’t steal anything, you got the wrong guy” routine. He shook my hand and said:

-“You are the only one who saw it coming and had the courage to speak up. Whatever You say i do. What should i do now?”

-“Really?”, i said. “I don’t remember. I might have been seriously drunk”

-“You were a little more than tipsy, but you were funny and more importantly you were right”, he said

I had correctly predicted the crash, and was too drunk to even remember it. Sounds about right:

-“Hmm, Plausible”, i answered. “Alright, so buy when monetary authorities roll out the big guns”,  i said.

-“OK, but Buy what then?”

-“Come on, do i really look like i know? Seriously? Monetary authorities are panicking. They are about to roll out the mother of all monetary bazookas. So, it does not matter what you buy, everything will rally”

-“Yeah, right. You are drunk again” and he walked away

He did not buy and I should have followed my own advice. I gave back a lot of performance in the 2009 rally. I insisted on shorting cyclical stuff when i should have shorted defensives. I was stubborn. I stopped trying to make predictions shortly thereafter.

Now, the implicit question is probably: are we in the same situation now? I don’t know and frankly, it really does not matter. I looked at my forecasting accuracy stats and concluded i should not be in the forecasting business anymore. Not encouraging for the rest of the industry considering i foresaw the crash and the ensuing recovery.

These are the main lessons:


  1. predict the next crash: you will sell too early. It is as useful as forecasting when you are going to get sick. Only hypochondriacs check themselves in hospitals before they get sick. Don’t listen to those market hypochondriacs telling you the mother of all bear markets is around the corner.
  2. Focus on why: what matters more? Why you got cancer or how to cure it?
  3. Duration and depth: when it happened, sell-side “quants” rolled out average duration tables. On average bear markets last for xx months. If You are sick, what would you think of a doctor who would say You will have 39,8 C fever and you will heal in 3 weeks 2 days, 17 hours?


  1. Recognise when it is there: build a system that tells you now it is time to sell and go short. To identify the top, use my floor ceiling method. It works objectively after the fact. Laurent Bernut’s answer to What is the most precise way to draw support and resistance lines for forex trading?
  2. Recognise when it is gone: Everyone got terminally beared up at the end of the bear market. They all, me included, missed the rally. Those who said they bought in March 2009 are like descendants of the Mayflower: deluded liars. To identify the bottom, use my floor ceiling method. It works objectively after the fact
  3. Have a bear market plan: bear markets are notoriously stressful. Not the brightest idea to devise an emergency exit plan when the building is on fire

#Quora: How can Renaissance Technologies make so much money from financial markets by hiring scientists/mathematicians with no domain knowledge o…

How can Renaissance Technologies make so much money from financial markets by hiring scientists/m… by Laurent Bernut

Answer by Laurent Bernut:

I have never worked at Renaissance, so please take my answer with a grain of salt, but here is a first hand story that could shed some light.

Red OctoberOn June 22nd in NYC, my colleague, who is also ex-US Department of Defense consultant and myself, met with one of the foremost US experts on sonar detection (good luck finding him on Facebook, LinkedIn). He is a physicist with multiple PHDs, geeky funny. His expertise is signal processing. He is the real “Hunt for Red October”.

It was one of the most refreshing experiences ever. He explained his world. I explained mine. Cotes de Provence Rose, beer and wild berry Zinfandel helping, we tumbled down the rabbit hole talking even about epistemology, the philosophy behind math.

His world, signal processing, bears uncanny resemblances with ours. We explored Bayesian probabilistic determinism, which models (Gauss, Poisson etc) to apply to distributions, the cost of false positives (think trading edge), arbitrage between time and action with sparse data (confirmation). We spoke the same language. We were talking real problems: how do distinguish signal from the noise ? How fast ? What is the cost of being wrong ? What is the cost of being right ? Which statistical law applies to randomness ?

We entered a massive time distortion. We started around 2 pm and a couple of bottles down the road, but then after what seemed like 5 minutes, we were hungry. It was 10 pm. We could have gone on forever (*)

Compare this with glorified journalists, otherwise referred to as fundamental analysts.

  • “This is fairly valued”… life is unfair darling, so do you really think markets are fair ?
  • “On a sum of the parts valuation”… Frank N. Stein zombie valuation
  • “Fundamentals are strong”… Make fundamentals great again…
  • “Long term story is still intact”… Some HF reality TV celeb says that about Valeant by the way…
  • “On a DCF basis, our target price is +10% above current market valuation” … stop tinkering the terminal value to rationalise your subjective views
  • “i think there is 80% chance that” … bad arithmetic meets emotional roller coaster
  • “top quality management” … was also said about Enron, Bear Sterns, Kodak, GM, Chrysler, Valeant

Too much B/S bingo, too much theory,

Bottom line: “In theory, theory and practice are the same. In practice, they are not”. Yogi Berra, Yankee philosopher

Physicists approach the markets as a statistical problem. This is practical.

MBAs have too much untested theories in their head. It is costly and time consuming to unlearn all that junk.


(*) There is no way i could ever afford someone of that caliber; he charges something the size of Liberia’s national deficit per hour. But, he wants to send his granddaughter to Mars and he thinks our algo could be the right fuel, so we invited him to have fun with us. Maybe good guys do not always finish last…

How can Renaissance Technologies make so much money from financial markets by hiring scientists/mathematicians with no domain knowledge o…

Sharpe ratio: the right mathematical answer to the wrong question

Andrew Swanscott runs the podcast Better System Trader. This is great resource: fantastic interviews with real-life practitioners. I made a statement that attracted a few comments from listeners: Sharpe ratio is the right mathematical answer to the wrong question. Here is the answer as Andrew kindly re-posted on his website.

Sharpe was the right answer

First, let’s start with what Sharpe does well. There are two things it does well:

  1. Cross-asset unified measure: we all know that the most important component in alpha generation is asset allocation. Now, the difficulty is to have a single measure of risk adjusted measure of alpha. This is where Sharpe did the job. It could give a single number across many asset classes: be it fixed income, equities, commodities etc.
  2. Uncertainty: the human brain is hard wired to associate uncertainty with risk. It triggers the amygdala and activates the fight, flight or freeze reflex (see one of my posts about fear and greed). So, Sharpe is a good measure of uncertainty: it quantifies units of uncertainty adjusted performance.

Now, Sharpe ratio, as part of the modern finance package, was invented the same year of the coronation of the Queen of England. It was good, almost revolutionary for its time, since Batista in Cuba was fighting El Che and Fidel. But, like the UN building designed by Brasilian Oscar Niemeyer, it did not age well, and here is why:

Sharpe is not a measure of risk, it is a measure of volatility adjusted performance

Sharpe equates volatility with risk. Risk does not equate volatility and here are a few examples:

  1. Low vol may be extremely risky: LTCM had low vol. In fact, their strategy was to be short gamma. It worked until it did not. Fast forward 2008, vol funds collapsed one after the other. Low vol does not equate risk.  As the great American philosopher Yogi Berra reminded us: “in theory, theory and practice are the same. In practice, they aren’t”. In theory, CDOs and CDs were AAA, low vol high yield products.In practice, they weren’t
  2. CTAs like Ed Seykota, Tom Basso, Bill Dunn, William Eckhardt etc: they have supposedly hopelessly low Sharpe but have clocked >+25% year in year out.

Does it mean that the CTAs have risky strategies ? No, it means they have low semi-volatility adjusted strategies. Semi-vol is just downside volatility.

Volatility means uncertainty, just learn to get comfortable with it

As much as uncertainty is not pleasant and may trigger some reptilian alarms in our brain, we must learn to live with it. It involves mindfulness meditation, strict formalisation of strategies etc. Do not pray for an easy life, pray for the strength to endure a tough one.

Now, what is risk ?

Risk is not a small paragraph at the end of a dissertation called investment thesis. Risk is a number. The only difficulty is to find the adequate formula that goes along. There are two types of strategies: mean reversion or trend following. Please read my posts on the subject.

Risk is not difficult to quantify. It is only difficult to identify. I have come up with the common sense ratio as it recaptures both mean reversion and trend following strategies.

CSR = tail ratio * gain to pain ratio

sense ratio as it recaptures both mean reversion and trend following strategies.

CSR = tail ratio * gain to pain ratio

Please use the trading edge visualiser to find out your personality:



Bottom line, we have associated risk with volatility. We have come up with a measure of volatility adjusted performance and deem it a risk measure. CSR, on the other hand, is a unified risk measure that can be used across asset classes. It measures risk according to strategy type.

Please subscribe and get some files, material and resources. This is all free so take advantage of it.

Regardless of the Asset Class, There Are Only Two Types of Strategies

Finance is one industry where there is no shortage of creativity. There is always a new strategy, investment vehicle, or asset class. This alphabet soup is confusing, particularly when it comes to assessing risk and reward across asset classes. Yet, there is a simple universal way to classify strategies. They fall into two buckets: either mean reversion or trend following. Simply said, the exit policy determines the win rate, which then shapes the return distribution.


  • A powerful visual representation of style/gain expectancy: Call to Action: our commitment is to help people become better traders. if YOU want to visualise your style, opt-in and we will send YOU a portfolio diagnostic tool for free
  • Regardless of the asset class, there are only two types of strategies: mean reversion or trend following
  • Each strategy type has a specific risk profile, which require different risk metrics. Common Sense Ratio recaptures risk for all strategy types (Read this, it is important)
  • How to increase the win rate, gain expectancy and overall profitability depending on strategy type ?

I. The only two types of strategies: mean reversion or trend following

Over the years of patiently testing multiple algorithmic strategies, patterns in the return distribution repeated over and over. It eventually became apparent that strategies fall into two buckets: mean reversion or trend following. Attached are graphical representations of the gain expectancy of mean reversion and trend following strategies. The reason why the same patterns repeat themselves is simple: exit policy.

Market participants tend to treat exit as a single final event. Each trade is a binary event: either it is profitable or not. The accumulation shapes the return distribution. Hit ratio is then determined not by what we enter, but how we exit.

Charts below are return distributions for each strategy type. They are also visual representations of gain expectancy. One image speaks more than a thousand words. This representation changed my life. It permanently altered the way I perceive the markets. The game is about tilting gain expectancy: contain the left tail, moving the peak hit ratio to the right and elongating the right tail. This visual representation is a powerful tool. This is why we want to share it. We are committed to helping people build smarter trading habits. Sign-in to our newsletter (it’s free) and we will send you a portfolio diagnostic tool.

Mean reversion strategies compound small profits

Gain Expectancy - Classic Mean reversion

Death by knock-out: many small profits. a few knock-out blows

Mean reversion strategies compound multiple small profits. They rely on the premise that extremes eventually revert to the mean. They aim ato arbitraging small market inefficiencies. They often have low volatility  consistent performance. They perform well during established markets: bull, bear or sideways. They unfortunately perform poorly during regime changes. They also perform poorly during tail events. The key issue is to contain rare but devastating blow-ups.

 Mean reversion strategies characteristics are (see graphical representation):
  • Moderate to high turnover
  • High win rate: often above 50%. The shorter the duration, the higher the probability of success
  • Consistent small average profits: trades are closed around the mean
  • Low volatility consistent performance
  • Potentially devastating left tail losses: make a little bit of money every day and lose a fortune in one day
  • Long period of recovery after losses:
Examples of mean reversion strategies are
  • Short Gamma: sell OTM options so as to collect pennies in front of a steam roller
  • Pairs trading (non FX): bet on the convergence between two historically correlated securities
  • Value investing: Buy low PBR stocks and “undervalued” assets
  • Counter trend: sell short shooting stars and catch falling knives

Mean reversion strategies post modest but consistent profits. They cater to investors who would look for low volatility returns. Their challenge is the left tail, those infrequent big losses that will take a long time to recoup.

Trend following strategies have a few home-runs 
Gain Expectancy - Classic Trend Following

Death by a thousand cuts: many frogs, a few princes

Trend following strategies rely on the capital appreciation of a few big winners. Whether they follow stories, fundamentals, earnings or price momentum, stock pickers are trend followers. They may fail to appreciate being called trend followers, but their P&L distribution tells a different story.

Trend followers kiss a lot of frogs: they have a low hit ratio, often between 30% and 45%. Performance is cyclical. Styles come in and fall out of favor. Volatility is elevated. Performance can be underwhelming for long periods of time. The key issue is to contain losses during drawdowns.

 Trend following strategies share those common characteristics (see graphical representation):
  • Relatively low turnover
  • Low win rate: 30 to 40%: see chart
  • Big wins and lots of small losses: right tail on chart
  • Relatively higher volatility
  • Pronounced cyclicality: style comes in and goes out of favor
Example of trend following strategies are
  • CTA type systematic trend following,
  • Momentum: earnings momentum, news-flow, price momentum
  • GARP investing: growth at reasonable price
  • Buy & hope
 Trend following strategies post impressive but volatile performance. They can go through long periods of underwhelming performance, which take their toll on the emotional capital of managers. Their main challenge is to keep cumulative losses small. Profits only look big to the extent losses are kept small.
II. How to measure risk for each strategy type
Investors suffer from a “nice guy syndrome”: some young women genuinely say they want to marry a nice guy, but unconsciously react to so-called “bad boys”. Investors genuinely say they want returns, but in reality they do react to drawdowns. More specifically, they are susceptible to drawdowns in three ways:
  1. Magnitude: never lose more than what investors are willing to tolerate
  2. Frequency: lull investors to sleep. Clients will trade performance for low volatility: big money is fixed income, not stocks
  3. Period of recovery: never test the patience of investors.
There are two ways to lose a boxing match: either on points or by knock-out. Mean reversion strategies score until they get knocked out. Trend following strategies lose on points. Risk is not evenly distributed. Therefore, each strategy deserves its own set of risk metrics.
Risk metric for mean reversion strategies: knock-out


Mean reversion strategies have low volatility, consistent performance and high Sharpe ratio. On the surface, they are what investors look for. The problem is mean reverting strategies work well, until they don’t. Big losses are unpredictable. LTCM had a great Sharpe ratio, at least until October 15th, 1987… Risk is in the left tail. The best metric for mean reverting strategies is therefore: tail ratio. Tail ratio measures what happens at the ends of both tails:

Tail ratio  = percentile(returns, 95%) / percentile(returns, 5%)
For example, a ratio of 0.25 means that losses are four times as bad as profits. Turnover then becomes an important variable: the higher the turnover the shorter the period of recovery. The two ways a mean reversion strategy can survive is either by 1) containing the left tail or 2) increasing turnover.
Mean reversion strategies will test investors’ nerves on two things: magnitude of loss and period of recovery. For example,  some strategies such as fundamental pairs trading post constant, reassuring but modest profits like 0.5% a month. Then one day, they post losses of 3 to 5%. They lose in one month the gains of an entire year.
Investors often succumb to the sunk cost fallacy with mean reversion strategies. They believe that big losses are rare and that managers will eventually make them back. This bias ignores probabilities, particularly the theory of runs. It also ignores opportunity costs. The good news is that there is an optimal point below which it makes more sense to redeem than to stick with managers who experienced a severe loss. It is often referred to as optimal stopping. Whilst the formula can be complicated, a simple rule of thumb is to redeem if losses are below 0.4 of turnover.
On the other, trend following strategies have tail ratios ranging from 3 to 10. Winners are much bigger than losers. So, tail ratio is meaningless for trend followers.
Risk metric for trend following strategies: erosion
Trend following strategies have typically low win rates. Risk is therefore not in the tails, but in the aggregates: are a few winners big enough to compensate for the multitude of losers ? Measuring risk then boils down a simple ratio of profits over losses. The risk metric for trend following strategies is therefore:
Gain to Pain Ratio = Sum(profits) / Sum(losses)
Trend following strategies will test investors nerves on frequency of losses and period of recovery. Frequency of losses is another word for volatility. Trend following strategies are volatile, but semi-volatility (downside volatility) is low. They can also post lackluster performance for long periods. For example, mutual funds have built-in cyclicality. Even if they claim to beat the index, mutual funds still lose money during bear markets.
GPR does not apply to mean reversion strategies because blow-ups are unpredictable. GPR can stay high until it is torpedoed by one or two bad losses.
Combined risk metric: Common Sense Ratio
 “Common sense is not so common these days”, Voltaire, French freedom fighter
Managers rarely define themselves as adherents of either mean reversion or trend following. Even so, it still would not be easy to assess robustness. Besides, the more risk metrics we use, the more confusing it becomes. For example, some managers have great performance despite a bad Sharpe ratio, so the question is “which matters more in which context?”
Since both metrics outlined above can be expressed in a simple binary ratio, combining them makes sense. When this ratio was first to colleagues and friends in the HF world, comments sounded like “yep, common sense”, hence its imaginative name: Common Sense Ratio
                                                                     Common Sense Ratio = Tail ratio * Gain to Pain Ratio
                                                               Common Sense Ratio = [percentile(returns, 95%) * Sum(profits) ] / [percentile(returns, 5%) * Sum(losses)]
Above 1: make money, below 1: lose moneyCSR is much more powerful than either metric taken individually.
Example 1: GPR = 1.12, TR = 0.25, turnover = 2, CSR =  0.275
Let’s take a classic mean reversion strategy that generates 10% p.a. (GPR = 1.1). It has a moderate turnover of 1.5. Within a 24 months period, it will post a monthly drop of -4%, with a 95% probability. This is 5 times as big as the average profit, and roughly 4 times as bad as right tail profits (TR = 0.25). Common Sense Ratio is CSR = 0.25 *1.1 = 0.275.
On the surface, it may look like a modestly attractive strategy. In reality, the period of recovery combined with magnitude of loss imply that investors will have to be patient. In very simple terms, the CSR shows that returns are not attractive enough to justify investing in such strategies. Try this with a few low volatility strategies. Risk is not where You think it is.
Example 2: GPR = 0.98, TR = 3, turnover = 0.5, CSR = 2.94
Now, let’s take a strategy that loses 2% over a complete cycle (GPR = 0.98). Best winners are 3 times as big as worst losers (TR = 3). Turnover is low 0.5. You may think, why invest in a vehicle that loses money ? It does not make sense. Yet, You are invested in such vehicles: welcome to the average mutual fund. 80% of mutual funds lose roughly 2% to the benchmark. Every now and then, they outperform with a vengeance. The rest of the time they suck air. Morality: over time, mutual funds are poor investment vehicles if You stay invested through the cycle.
III. How to tilt the win rate, gain expectancy and overall profitability depending on your win rate
Gain Expectancy - Alpha SecureThe question boils down to: is it possible to combine the benefits of both strategy types without having the drawbacks of either one ? How can we generate a return distribution that would look like the one on the chart ? (*)
The whole game of investing is about generating a return distribution that would have the following characteristics:
  1. No left tail: small losses like a trend following strategy
  2. Long right tail: big wins like a trend following strategy
  3. High win rate: above 50% win rate like a mean reversion strategy

This type of strategy combines both short term compounding with long-term capital appreciation.

Investors following a mean reversion typically come in early and leave too early. The key is therefore to elongate the right tail. This is done through allowing a remainder to extend beyond the mean with a trailing stop loss.
  1. Set a stop loss (more on this in an upcoming article)
  2. When the trade means reverts, close half the position
  3. Set a trailing stop loss (not based on valuation) and close the trade once the stop loss is penetrated
Bottom line, markets can stay irrational longer than You think. When it stops making sense for You, it may start making sense for someone else. Ride their tail, but protect your downside.
If You are a trend follower, here is a simple game (game of 2 halves) that will mechanically improve your return distribution. We all know that making money is about cutting the losers and riding the winners. Here is an objective way to do it:
  1. Calculate your average contribution, divide it by 2
  2. Reduce by half every losing position below – half average contribution
  3. re-allocate the proceeds to winners
Bottom line, you have reduced losers and increased winners in a simple way (more in an upcoming article)
The  purpose of this article was to introduce a simple yet powerful way to reframe strategies independent of asset class. This enabled us to look at the merits and drawbacks of each type. We then looked at risk metrics that would best recapture their risk profile. We introduced a unified risk metric Common Sense Ratio that works across all asset classes. Finally, we looked at ways to tilt gain expectancy for each strategy type. Last but not least, if You want to know what your trading style, please subscribe and we will send You a diagnostic tool for free.
Preview of the next article:
The next article will deal with investor psychology. Short sellers have a unique perspective on investors psyche. We never sell short against buyers. We observe people who once held a position and are now processing grief. The next article will be about the psychology of grief adapted to the markets.
  1. Market regime: bear markets have several distinctive phases, with a measurable market signature
  2. You will never read an analyst report the same way again. You will learn to read emotions through language and back it up with numbers
  3. You will be better equipped to bottom fish for stocks
(*) : All above return distributions are derived from the same strategy. It has both mean reversion and trend following components. In order to draw a mean reversion strategy, the trend following component was switched off, and vice versa to build a TF distribution. Both components are normally switched on, third distribution