Mar

22

Leverage, from George Coyle

March 22, 2012 |

 What is the best recipe for leverage? Is it safest to backtest models and find the worst points historically and measure leverage accordingly? So if a model would have gone down 20% at the worst point in its history unlevered I know 5x leverage would have equaled a wipe out and thus use less always mindful the system could see that point (or worse) again? Or best to factor historical probabilities using counts or scans to do so? For example if 12/14 days my model has lost and based on history there is a 2% chance of tomorrow being a loser should I then up the leverage? Conversely if I have won 12/14 in a row should I decrease leverage or is it best to let it ride when in the black upping the leverage?

On a related note in the options market, I find most models eventually reach ruin. If you buy options you run into prolonged periods of no vol and are so depleted by the time the high vol periods return your are in trouble. If you sell options unhedged you run into crash problems. Spreads (put/call) seem interesting but the most optimal points to employ them generally result in taking in 5% or less of the capital at risk and if you up the leverage you risk ruin here too. If you don't lever up you don't make much.

Employing no leverage makes for stability and lesser concern but then a reasonable AUM is necessary to generate $$ amounts that make the game worth the effort.

Leo Jia writes: 

Hi George,

IMHO, I don't use the historically worst point to determine my leverage level. I consider that event not statistically robust. What I do is that, assuming I have a profitable model, on each trade, I determine the leverage level based on the risk level according to the model at the entry price and how much I am willing to lose for any single trade. To minimize the impacts of large and prolonged drawdowns, I sharply reduce my position sizes during those periods.

 


Comments

Name

Email

Website

Speak your mind

Archives

Resources & Links

Search