Time of day is a good one....but there are exceptions to this rule.....like Fed announcement days, etc. You should use an override parameter for this one: Trade_Dead_Zone=N Dead zone defined as 11:30 am -> 1:30 pm (or define your own). So on most days you would not be trading in the dead zone.
Looking at your chart there were entries made at extremes of a bar. Question: How many of your sell trades happened on top of bar and how many buys happened on lows of bars? What does system do if it can't entry at desired price? I spent years with texting S&R methods, more trades means overall losses cause you need to use at least one to one and half ticks slippage then include commission, often leaves less than a tick of profit left and should a report come out or special news would hurt performance badly, but this is what I have tested, more is not better.
Quite a lot actually. But then, it catches not only the absolute top and bottom but all the intermittent dips in the middle of the trend. I've yet to find a good cure to iron out the kink. As far as slippage is concerned, when I ran the backtest I threw in $12.50 (1 tick) for slippage on every order filled. Since all my orders would be market order, I figure it would cost me 1-tick on each trade. The commission was also baked in. Even with the slippage and commission, the result was net positive. I ain't sure that answered your question. Let me know if I missed anything. The beauty of automation is that every trade is executed without my presence. So the logic that dictated my thinking when I set out to devise my strategy is, if that is the case, then why not go for the strategy that would yield the largest net positive return. So I ran a bunch of backtests of the same strategy over many different timeframes. It turned out shorter timeframes gave higher net return (profit/loss plus commission and slippage). If, for example, 20 trades were executed using the 15-minute chart , 233-tick chart had 500 trades. But whereas the 15-minute only made $1,000 during x number of days, 233-tick made in excess of $25,000. The point I want to make is that the difference was rather huge between two timeframes. If the strategy is profitable, why not let it run on autopilot and allow it to generate the largest possible amount of money? So my thinking goes.
What is the result if you do the same test on other periods of data? The result should be more or less the same, if not your result is based on luck. You should test this on years of data. Be careful with backtesting. I saw already backtests that made consistent and good profits by taking a few ticks each time. When they started to trade it in realtime they lost in a very short time (less than 1 hour) $900 per contract. These few ticks were supposed to be taken theoretically, but were not there in reality. $25.000 in 500 trades means average 1 point profit per trade. That is dangerously close to zero. If you would make 2 or 3 points average, slippage or a small miscalculation would still keep you in profit. But profits of 1 point or less can quickly become losses in realtime trading. Backtesting can sometimes create the illusion that you have a good system.
Exactly and all he needs to do is mine his backtested trades and determine what kind of filter will keep those trades from occurring. So his basic method is OK, all he needs now is a filter.
I have 12 automated systems running, but I have opposite results where some trades left that "iffy" territories of when volatility comes into play causes of HFT, slippage increases during these times where it makes segments of trading plain stupid for trades to occur cause I understand in day trading I will always have inverse of more risk to reward on most methods, but for me to maintain very low losing percentages, price has to be stable. I stopped looking at best net return five years ago as NOW I seek lowest losing percentages cause ES does not trend like most markets as it has high degree of chop, and chop for me is my friend so having low losing percentages allows averaging down and has increased bottom line 140% but do have times where if nailed with three losses in short amount time takes 4-7 weeks to recover but even so bottom line goes up for all the breakeven plus one tick breakevens. My concern with your method is actually getting filled at shorts on highs and longs at lows as big scalpers and brokerage puts in orders as soon as price moves away from those areas by 1-2 ticks, so your order will never be first to the party and if you doing market orders, you will always lose a tick or even two for lack of volume. Also at some point if it does well, volume gets to be difficult, putting in 300 lot order when at night, can run through 1-3 price levels, one lots usually just one tick. It is like I have program for Crude oil, it does ten lots and hardly ever happens at all same price level in middle of the night. Hey, much luck with it, automation the way to go, some days I forget they are cranking it out. But until I stop day trading since I am losing memory, I still enjoy taking money by hand.
Some of the filters I use is blocks of trades, program sees what "normal" is of last ten minutes, then if blocks depending on overall volume, if it senses blocks coming in like HFT, it stops looking for signals. But when you rely on S&R, you are screwed. I would test on more of a longer term filter and trade with the trend, throw in swing average to start sensing when reversal point should come in to hunt for extremes and fail safe when price gone beyond extremes for hard trending. Most of my back testing for me has show reduction of trades and bottom line doesn't change much.