I assume others here are using some form of VAR calculations in their automated trading strategies? Prior to releasing an order to an exchange, I ensure that it does not put my VAR above my risk tolerance. For "Naked" orders, this is straight forward... Its a function of the probability distribution of returns on the underlying asset and position size. I tend to use stops (and targets, via Rithmic Bracket Orders) in my algo trades. My current methodology for these Open-Stop(-Target) order pairs(triplets) is to consider the VAR effect of the bracketed order to simply be VAR += LossLimit . This is fine most of the time, except in fast markets where an open can be filled, and the the market moves through the stop before the stop hits the exg, resulting in a rejected stop (I opened a thread about this during the VIX blow up awhile back...) and a naked position. Obviously, the position should be liquidated immediately by the algo when it receives the reject message from the exchange, but there is a non-zero period of time between reject and response where the VAR SHOULD increase substantially because there is no loss hedge in place... my question is: Do you incorporate micro-structural risk into your VAR calculations? Currently, I do not..I only historical price movements, and I trust my stops will be filled. I mitigate this by using volatility based stops (I didn't before....until tight-stops on NQ almost got me), but micro-strutural risks are real....and I'd be curious to know how the experts deal with them?