Traders' Roundtable

I tested subsequent performance from drawdowns in a system with a MAR ratio of 1.42. Trading commences on the first date of a 25% drawdown and ends a year from that date.

Well, the regression to the mean principle seems at work here..... and there's some other material on this forum (posted by Dean Hoffman, if I remind well) stating that the most appropriate time to start/increase position in a system is after a deep DD.

To complete the test, you could maybe try to evaluate performance of the system after a new equity high, and find if there's a statistical difference (unluckily, statistical significance arose - according to the books - with a sample of at least 30 data..... that is 30 DD periods... the more you have the better....).

This is interesting stuff, but when one tries to get to practical rules, it does arise more questions than answers (at least to me): for example: with standard TF systems, most of the time you are not living neither with the worst DD's (by definition the worst one's should be few and far apart) nor with a new equity High.

So, what do you do in "normal times"?
Stop trading? Lighten up the posizion size?...

In the end, this seems to me like a "buy the dip" approach..... people more clever than me, like Victor Niederhoffer, Amaranth and other..... had clearly shown that the one's betting - one way or the other - on the regression to the mean approach, need to be wrong just once to end belly up.

Just my two cents.

This is Tom Basso's "Buy the equity curve after a huge dip" strategy. You could evaluate the benefits of this strategy by comparing its 12 month returns, to the returns of all 12 month holding periods, as Tom did. Is DipBuying's 12 month return, better than the average (mean) 12 month return? Is DipBuying's 12 month return, within the top 50% of all 12 month returns (median)? Basso went one step further and tabulated the data as ordered pairs

• ( (%return in prior K months), (%return in next N months) )

where he chose K=3, N=12. This allowed him to make scatterplots with prior performance on the horizontal axis and subsequent performance on the vertical axis. See (this post) on the Forum. Basso's findings were,

• If you entered after a 3 month loss (drawdown), the following 12 months' return was positive (profitable), 68.5% of the time (318 instances out of 464).
• If you entered after a 3 month gain, the following 12 months' return was positive, 67.1% of the time (742/1105)
• If you entered regardless of whether the previous 3 months produced a gain or a loss, the following 12 months' return was positive, 67.6% of the time (1060/1569).

Basso was thrilled to find that entering after a 3 month drawdown gave higher odds of success in the next 12 months, compared to entering after a profitable 3 months. In my opinion, the advantage wasn't particularly large: 68.5% vs 67.1%.

By the way, Managed Account Reports magazine (the people who invented the MAR ratio) use "the largest drawdown since the inception of trading" in the denominator. Not "the drawdown of the last 12 months". So although it might be interesting to compute (12 month return)/(12 month drawdown), it's not actually a "MAR ratio".

I don't plan on trading the equity curve beyond the start date because it is an indirect way to explain what is happening.

The drawdown of a system represents something about the market. In trend following it represents a period of low volatility or non-trending conditions. Now, the question is are the markets more likely to start trending because they stopped trending?

I remember arguing with a math teacher in high school about roulette, and how the "odds" increase that the next spin will be black if there is an unlikely string of red spins. It took days for me to grasp that mathematically, the next spin has the exact same odds as the previous ones even if the string of reds is unlikely. That string of red does not influence the next spin. I think this is true in the world of finance.

P.S. I thought MAR ratio was short for Calmar ratio.... are we sure Managed Account Reports invented it?

Eric Winchell wrote:
P.S. I thought MAR ratio was short for Calmar ratio.... are we sure Managed Account Reports invented it?

We had already a thread about it:

viewtopic.php?t=2100&highlight=mar

Greetings.

P.S. I thought MAR ratio was short for Calmar ratio.... are we sure Managed Account Reports invented it?

We had already a thread about it:
viewtopic.php?t=2100&highlight=mar

I updated that thread to point out the mistake in the Wikipedia entry for Calmar Ratio, which started this misconception.

I believe in Basso's paper on his test, he admitted that his results were affected by survivorship bias. If I remember correctly, he only used CTAs that were active at the ending period of his test, rather than those which were active throughout the whole period.

Thus, in real life applications, one could find that the drawdowns used to compute the MAR are slightly understated.

In my backtesting, I found that adding capital during drawdowns can increase my total equity faster. The main reason for this is that once more capital is added to the fund, position sizes increase. Thus the system pulls itself out of drawdowns faster.

There is a downside to this however. Should the system suffer a string of losers after you have added capital, these losers tend to be bigger due to the aforementioned increase in position size.

Also, it tends to by much more difficult (for me at least) to add money during drawdowns than it is when new equity highs are reached.

I've had the lovely experience of beginning to trade a system near its historic drawdown levels. Only to then have the historic max doubled. ("Hows that Mr. smartypants", the markets said to me...) This was an outlier it seems.... But is has, can, and probably will continue to happen now and then. Ah risk.

Note that were that account a CTA, it probably would not have survived, thus not showing up on Bassos List.