Hello,
I have some feedback about statistics. First I’d like to say it’s a very good thing that so many stats are available (some of them behind the Advanced link).
But there is always room for improvement and here are my suggestions:
1. Add biggest winning and losing trade. Or even better, if possible, largest drawdown in a single trade and largest up move in a single trade in dollars. Avg win/lose already tells a lot but especially in young systems risky behaviour may not be reflected in these stats.
2. Replace Sharpe with Sortino in front page (or add Sortino in front page). Sharpe punishes for quick up moves which does not make sense.
3. Remove or move to adv side “Risk of x% account loss”. In my experience this tells nothing and can lead some subscribers to really believe that risk of losing 50% of account is zero which seems to be suggested for almost all systems (including those which have had a 50% drawdown!).
thanks,
- Panu
I could not agree more with point #2. There are countless publications refuting Sharpe as a valid risk measurement tool due to its upside standard deviation inclusion.
And let me tell you how unqualified anyone refuting the Sharpe Ratio is…
I believe sortino is in the advanced stats on each system.
Per 1) You can do that yourself.
Per 2) Not going to happen, and also not needed, since any serious subscriber probably uses Sharpe.
Per 3) Better to think of it as an expected value.
Sharpe is mysterious enough without expecting people to understand the even lesser known Sortino
If it’s that important, it shouldn’t be too difficult to present us with some reasonably well-performing pairs of systems where system A would be preferred over B based on the Sharpe ratio, but system B would be preferred over A based on the Sortino ratio.
Good luck!
My comment was intended not merely for the narrow scope of this website but for the industry in general.
The point is LTCM and a host of others had a wonderful Sharpe, but it masked the true VaR. I do not discredit Sharpe, but I also recognize its inherent deficiencies. Statistically one cannot argue these limitations and to compensate, Sortino is a capable addition.
http://www.premiacap.com/publications/RR_0901.pdf
http://www.prmia.org/pdf/Till_Satellite_Risk.pdf
Best,
Fred
You cannot judge by Sharpe OR Sortino. That is why other statistics also need to be used, such as profit factor, max DD, #trades, equity curve, history length,the vendor’s other systems, APD, and others. Many people learn to game a stat, but it is much harder to game multiple stats.
Trust me. You’ll see it in the curve. And if you don’t, I guarantee you’ll see this in drawdown.
Pity we don’t have intraday DD stats for LTCM. I know they knew their accounts were fluctuating wildly, but since they weren’t tracked by a third party, we didn’t know until they crashed and burned and they only reported on monthly and quarterly bases.
Not only that, but they were dealing in huge amounts of leverage. DUH! But that wasn’t the problem.
Have you ever wondered what would happen if you were long a worthless futures contract?
Say, ES went to 0, how much would you have lost if it happened quickly enough that you never had time to react. Or everyday was limit down so that you couldn’t exit until the contract was worthless? This is what I believe occurred with LTCM.
It’s a very extreme case, and the main risk was to the counterparties that smartly took the other side of these trades.
Just think about that one. I’m going to have to hash over and think a bit, but it’s the only explanation for what occurred.
In any case, you would have noticed the volatility far ahead of this happening.
