Sorry for not replying earlier! I always buy at the ask and sell at the bid. I believe that is the proper way of simulating market orders. If I was trading large size, I'd also look at bid/ask sizes. Here is my current backtesting results. This system trades a single NQM3 contract, using market orders only. It scores a mean(daily pl)/std(daily pl) of $82/$160 = 0.51 over a 19 day period, using an average of 3.3 positions *2 = 6.6 trades per day. The average daily profit/loss ratio is $147/$60 = 2.45 Code: Date # positions pts/pos $ p/l 3/13/2003 2 9.50 370.40 3/14/2003 3 0.50 15.60 3/17/2003 4 5.25 400.80 3/19/2003 7 -0.07 -43.60 3/20/2003 4 1.00 60.80 3/21/2003 2 5.25 200.40 3/24/2003 2 3.75 140.40 3/25/2003 2 10.50 410.40 3/26/2003 5 0.50 26.00 3/27/2003 4 0.25 0.80 3/28/2003 3 -1.17 -84.40 3/31/2003 3 2.67 145.60 4/1/2003 6 0.42 21.20 4/2/2003 1 -1.00 -24.80 4/3/2003 3 -0.17 -24.40 4/4/2003 1 4.50 85.20 4/7/2003 4 -0.63 -69.20 4/8/2003 5 0.70 46.00 4/9/2003 3 -1.67 -114.40 mean 3.37 2.11 82.25 on up days 3.31 3.44 147.97 on dn days 3.50 -0.78 -60.13 std 1.61 3.50 160.09 ($ p/l includes IB commissions) This system does not train inductively over past days, or retain any information day-to-day. It uses a few a-priori settings, two indicators, a signal deglitching/ de-thrashing algorithm and a performance-merit rule which effectively avoids trading in chop or when the indicators are not performing well. This not fine-tuned. Most parameters are determined as the day progresses. These systems are prototyped in Matlab, then converted to C, compiled and linked in with the IB TWS API in order to receive ticks and issue orders. I am currently backtesting on previous data (NQ currently) and doing some papertrading of the real-time C code. The question I need to answer is: Do these 19 days represent a typical group of trading days for NQ? I realize we are in a war. Has current events made these days more or less choppy or hard to trade (using any short term patterns)?