July 23, 2007
When Traders Blow Up
I was inspired by Taleb’s simulations on the hidden risks of “blowing up” as a Wall Street trader, so I built a
Here’s how I did it, please follow along and provide some critique, I’m not 100% sure if I did this correctly.
First I generated a random % return per year over a 5 year period, I did this for 1000 hypothetical traders and assigned a point value system to these random returns; 0 meant that you blew up before the first year was over (drawdown > 20%), 3 meant you were profitable for 3 years before blowing up, and 5 meant you made it to 5 years of profitability.
Then I calculated the mean and standard deviation for the point totals across the 1000 traders, modeled them over 10,000 observations, and produced a normal distribution curve.
The results were astounding!
You have roughly 1.42% chance, or 70:1 odds, of surviving 5 years as a Wall Street trader! Wow!
July 23rd, 2007 at 11:02 am
Seems like this kind of thing is very sensitive to how you model the hypothetical returns. All you said in the post was that you generated random per-year returns… but did you just use a normal distribution, and if so, with what mean and variance (and why)?
Doing it per year rather than per day hides traders that had a 20% drawdown during the year and then recovered. Did you mean to ignore those cases?
July 23rd, 2007 at 11:54 am
Richard: For this simulation the Mean was 1.382 Years (out of 5) with a Stdev of 1.564, and yes it was a normal distribution. Ideally there shouldn’t be anything less that zero as time “t” must be >= to 0.
Your right that an intra day or month draw down of 20% would cause the trader to blow up. I just used random yearly returns to make it easier for me to model right now and to create discussion.
July 24th, 2007 at 2:10 am
I think you are responding with means and stdev of your results (the trader’s viability in years). What I was trying to ask was, what did you use for the mean and std deviation of the returns you were modelling? It sounds like you generated a bunch of random yearly returns, but how did you decide what the correct (realistic) distribution of hypothetical returns should be? I think that would be the hardest part of getting to a realistic answer, by far.
July 24th, 2007 at 4:16 am
Ah yes, you are right re: the mean/stdev of the results.
I generated the returns using Excel’s RAND() function for a value between -1 and 1 (-100% and 100%).
You do raise a good question, how do we know if the random returns are realistic? We can solve this! I can rerun the simulation using a random yearly return using a normal distribution. As a full-time trader, what do you suggest would be realistic distribution for a trader per year? What mean and stdev should I use?
July 24th, 2007 at 5:29 pm
I have no idea what the right distribution should be, which is why I was asking how you decided what to use. -1 to 1 doesn’t feel right, as the extremes. I mean, plenty of traders make more than 100% in a year, so going from -1 to 1 can’t be quite right. (Sadly, some traders manage to lose more than 100% of their equity, as well!)
I would imagine the mean would be less than 0, as most traders do lose money, after all. I doubt the real distribution is all that bell-shaped… but would rather have a long flat positive tail and a shorter, fatter negative tail. Maybe there’s some sort of data available on the net somewhere?
July 24th, 2007 at 8:23 pm
Well on the extreme low end, the loss of 100% of your capital is a complete blowup. So that’s a good limit to use I think. For the sake of the model, I assumed that you could make 100% max. I could bump that up to 200 and see what happens.
For another randomly generated scenario, the mean and stdev for the returns was -0.0009 and 0.57709 respectively, so your suspicion is confirmed. The mean return for a trader will be negative.
January 19th, 2008 at 7:57 am
[...] also used RiskAMP to model the outcomes in my controversial “When Traders Blow Up” post and I continuously use it to model volatility outcomes for various stocks. Combined with [...]
January 19th, 2008 at 4:19 pm
I understand it would require a lot of work to make this more granular, but getting estimated return distributions by trading style or asset class would be a very interesting analysis. One could then shift their portfolio/system accordingly as distributions change – it would be very interesting to use this as a model for market dynamics and regime changes.
The problem, obviously, would be getting reliable return distributions, and other inherent data problems such as for those traders who trade multiple asset classes. Just a thought, though.
January 19th, 2008 at 5:48 pm
foquant: The entire idea behind this exercise is to get one thinking that blow ups are quite more common that people think.
January 19th, 2008 at 8:39 pm
sure… I’ve always heard the heuristic that 95% of traders fail in the first year. I believe that stat applies to individual/prop traders.
but, if you can identify, with any reliability, that a trader is more likely to fail in equities than in spot currencies, it may make sense to focus a heavier percentage of one’s portfolio in currencies than otherwise, as the return distribution would be more attractive. of course, just observing the distributions may show the same thing. if you could get reliable data, you could essentially tell which markets, at any given time, are relatively more or less efficient. could be just rambling on my part, though…
January 20th, 2008 at 1:19 pm
You know, you bring up a good point. There might be a different success rates for traders in different markets.
Your rambling is appreciated! :)