NQ (1) date, trades, gross points 4-Sep 4.00 51.50 5-Sep 1.00 -4.00 6-Sep 3.00 -11.50 7-Sep 2.00 -7.00 10-Sep 2.00 -7.50 17-Sep 2.00 22.50 18-Sep 4.00 30.50 19-Sep 1.00 -4.00 20-Sep 0.00 21-Sep 0.00 24-Sep 5.00 -18.00 25-Sep 0.00 26-Sep 1.00 11.50 27-Sep 4.00 27.50 28-Sep 7.00 -24.00 1-Oct 4.00 -13.50 2-Oct 0.00 0.00 3-Oct 3.00 67.50 4-Oct 2.00 -10.00 5-Oct 2.00 -8.00 8-Oct 2.00 -8.00 9-Oct 5.00 -10.00 10-Oct 3.00 9.00 11-Oct 3.00 -11.50 12-Oct 1.00 -4.00 15-Oct 2.00 -6.00 16-Oct 4.00 -20.50 17-Oct 2.00 88.00 18-Oct 5.00 -18.50 19-Oct 4.00 20.00 22-Oct 3.00 9.50 23-Oct 5.00 0.50 24-Oct 5.00 -22.00 25-Oct 2.00 56.00 26-Oct 4.00 -21.00 29-Oct 4.00 24.50 30-Oct 5.00 -20.50 31-Oct 7.00 -12.00 1-Nov 2.00 26.50 2-Nov 0.00 5-Nov 3.00 -5.50 6-Nov 3.00 -2.50 7-Nov 4.00 9.00 8-Nov 7.00 25.00 12-Nov 5.00 -16.00 13-Nov 4.00 -15.50 14-Nov 1.00 0.00 15-Nov 4.00 1.00 19-Nov 7.00 -4.50 20-Nov 2.00 49.50 26-Nov 5.00 7.00 27-Nov 3.00 -5.00 28-Nov 6.00 2.50 29-Nov 6.00 -13.00 3-Dec 8.00 -19.50 4-Dec 7.00 16.00 5-Dec 1.00 45.50 6-Dec 5.00 -11.50 7-Dec 10-Dec 11-Dec 12-Dec 3.00 -12.00 13-Dec 5.00 18.50 14-Dec 17-Dec 2.00 0.00 18-Dec 8.00 -19.50 19-Dec 10.00 -16.50 20-Dec 2.00 44.50 total 226.00 261.00 this comes out to about $4000 in net profit trading 1 contract, some days are pre-determined to be black-out days, such as Fridays in the past 2 months or so, Fed days and pre-Fed days... will black out all next week dead zone. Therefore this wraps up the 4-month simulation. I have to say that due to my full-time job and lack of automated order entry, only a few days of the above is real trading... but according to my past experience, slippage is well-guaged with NQ, especially with just 1 contract. I throw this out here to get an idea of performance comparing to any system or discrectionary traders out there.... what is your results trading the NQ in the past 4 months, on per contract basis. I believe the system has a lot of aspects to be improved. I have to say the market has become much more difficult than last year... back testing over 2000 data yielded much more sexy results.
In the Smart Trading 2001 issue of Futures magazine Larry Williams gave a trading methodology that I believe he said he uses for the S&P, Nasdaq and Bond futures. It would be interesting to see how this simulates? Richard
js1257: please explain the methodology that Larry gave in the Futures magazine issue. I suggest you refer to the article to get a good understanding of the method. However, I'll try and summarize it below: He uses 2 indicators: 1) Volatility Stop Indicator - the volatility stop sets values based on ranges of earlier bars. First, the indicator averages the ranges of n number of previous bars. That average is then multiplied by some factor. Next the product is simply added to the highest close over the range to get the next bar's sell stop, and it is subtracted from the lowest close to get the buy stop. The 2 main parameters are the lookback for the average calculation (n) and the adjustment factor. In the article the lookback was 7 and the adjustment factor was 2.8. (This description was provided in the Nov. 2001 issue on page 10). 2) %R indicator. Buy rules: If %R falls below 25% and that bar's close is above the volatility stop indicator, then either, A) buy at the market, or B) buy if price surpasses the high of that bar and continue buying at the successive lower highs of each bar thereafter, unless there is a close below the volatility stop. Sell rules: If %R rises abve 75% and that bar's close is below the volatility stop indicator, then either, A) sell at the market, or B) sell if price drops the low of that bar and continue selling at the successive higher lows of each bar thereafter, unless there is a close above the volatility stop. Williams uses 15 minute bars and a 5 period %R. In the article he presents an example of S&P trading.
total 226.00 261.00 this comes out to about $4000 in net profit trading 1 contract In my personal experience, I've noticed that systems that backtesting well but with an average profit of less than 4 ticks on NQ or ES tend to do just slightly above break-even in real-time trading because of the usual issues related slippage & other execution issues. You may have already factored this into your results above, but if not you might consider what I do to generate more "conservative" numbers. Here's what I do: 1. I charge 1 tick commission per trade 2. If I am using a limit order I only count a fill if price trades 1 tick past my limit since the high/low on a bar was "last traded" and I don't know if it was the bid or ask. 3. For stop orders, I charge myself a tick slippage if the ATR of the bar was > the ATR(period) where I usually use period = 12 (for 5 min bars or period=6 for 30m bars - just personal preference, most period settings will generate same results). 4. For every multiple of ATR I charge myself another tick slippage. E.g. if the current bar was a long range bar and is 3x the ATR(period) I will charge myself 3 ticks slippage. Generally speaking, my system charges an average of 2 ticks for commissions + slippage per trade. This is for ES & NQ. Other futures the tick setting may not be appropriate (e.g. 1 tick not appropriate for big S&P). If you system already accounts for slippage appropriately, then 1.15 points per trade is decent if it fits your style of trading and execution is straight forward.
dottom, Excellent information! I've had the same experiences with scalping systems that look great on paper. Then, I'll sit there and watch the NQ or ES trade at my limit then move away leaving me unfilled. So, I've gone back and done just what you've described, and seen some sobering changes. I've done the sensitivity anaysis for those trades which limit on the highs or lows of a bar. If I have to get 80% of them for the system to be profitable, then forget it. On the other hand, if it takes only 15% or 20% of those trades to make the system profitable, then I'll look at it a bit further. I've found that entries at the market have worked both for and against me about 50/50. So, on a very large average (e.g. trades on the order of 500 or greater in the backtest), assessing a tick or two of slippage has not been representative of my personal results. (You comment about a straightforward and fast execution system ringing especially true for me in this case.) Your ideas for simulating higher ATR bar entries and exits are much appreciated! That will make a great addition to my backtesting approach. Thanks much!
My systems simply buy at the ask, sell at the bid. They check what the bid/ask was at the time of the signal, then what it was 3-5 seconds later, and give me whichever is the worse fill. Of course I test on tick data, not on bars. voodoo
Good points dottom, voodoo and Dave.. dottom, I have read your posting in another thread a while back about the components of your ES system, without going into details, would you mind telling us the average ticks/trade for that system? Just wonna get an idea where I am in the performance spectrum. FYI, the results posted are gross results with no commission/slip factored in... therefore it appears this system is not robust at the current market condition. The same system, while backtesting over 2000 and 2001 data, yielded 4445 NQ points over 1252 trades, about 7 ticks per trade, before commission and slippage. Since it is mostly a trend follower, it appears the market is difficult for trend followers in the past months. Another factor is the fact that market is at such a low level now, the signal / noise ratio gets much lower, since now matter how low the market go, the noise amplitude does not go down proportionally. Therefore I think for the same system on the same market, a level at 3000 is much easier to trade than at 1500.