Jeff Castille 

So.....let's define a reversal bar again........ #1 IT MUST COME AT THE END OF A MOVE (UP OR DOWN) #2 IT MUST HAVE HIGHER VOLUME #3 THE WICK MUST BE 2/3+- OF THE ENTIRE CANDLE

The magenta bar must have higher volume? Or the bar before it must have higher volume? I'm assuming the 2/3 wick/magenta bar must have higher volume than the previous bar, correct?

Also, I'm still not clear on your stop loss strategy. I know you do 2 cars, +8 when go b/e or +1 on the other targeting +16, but where are your initial stops? Sorry if I missed it...

I was looking at the reversal indicator code, some would argue the definition of an inside bar is where Low[0] > Low[1] and High[0] < High[1] but you are using <= and >= so I'm not sure if what you are marking are truly inside bars. Are you sure what you read defines inside/outside bars to include the highs and lows equal to the previous bar?

So, below is a chart using your template. I just want to make sure I'm reading these correctly because I'm prone to automate and backtest setups once I understand every detail. Tell me if my assumptions are correct.

shodson 

I was looking at the reversal indicator code, some would argue the definition of an inside bar is where Low[0] > Low[1] and High[0] < High[1] but you are using <= and >= so I'm not sure if what you are marking are truly inside bars. Are you sure what you read defines inside/outside bars to include the highs and lows equal to the previous bar? So, below is a chart using your template. I just want to make sure I'm reading these correctly because I'm prone to automate and backtest setups once I understand every detail. Tell me if my assumptions are correct.

Hey Shodson,

Yes.......I purposely have the code set to identify an inside bar as equal to or less.....

You are correct ....#1 is NOT a reversal bar as it did not meet the criteria.

#2 did meet the criteria.....however Geithner started talking at 7 am PST and you can see when he finished when the market stopped jerking around and took off to the upside.

#4 unfortunately was a stop out......as it went down first and then exceeded the previous bar by +6......not enough to get my first target.

shodson 

The magenta bar must have higher volume? Or the bar before it must have higher volume? I'm assuming the 2/3 wick/magenta bar must have higher volume than the previous bar, correct? Also, I'm still not clear on your stop loss strategy. I know you do 2 cars, +8 when go b/e or +1 on the other targeting +16, but where are your initial stops? Sorry if I missed it...

See post #1 for stops.....the magenta bar itself must have higher volume (see example a post or two back)

I'm backtesting a little when I have time. Here is an equity curve for just inside bar (easiest to test). The profit target I've tested everything from 8 to 16 ticks. This is 8.

My conclusion is that this has worked unusually well during the past 2 months with a PF of 1.5. However, testing over the past year produces results that are not as good (PF 1.14). 1.14 leaves no room for margin of error, mistakes, or anything else. It's walking a very thin line. testing longer term has given the same results around 1.14.

The win % for past 2 months has been 85-90%. Even for the past year it's 80%. The problem isn't the winners, it's the losers. The stops really hurt.

Now I don't take into account news or lunch. I don't think it's really important because the perf is so good for the past 2 months even not taking into account news & lunch. So even if I add support for news & lunch, the last 2 months are still going to be very good compared to the past year.

Also using two targets and going to breakeven would improve results. However I don't know how to test such things with tradestation. I can do that with ninja but it's a lot of work, compared to tradestation where it's relatively easy to do simple tests.

so I'm not sure what to make of it. In the past this has had major losing streaks. That could happen again at any time. or it could continue working for the next year.

I guess the best would be to milk it while you can and if it stops working then be very careful.

equity curve for 8 tick chart for past 12 months using commissions but NO SLIPPAGE (any slippage whatsoever will kill it since the avg trade is only $9.

shodson 

I was looking at the reversal indicator code, some would argue the definition of an inside bar is where Low[0] > Low[1] and High[0] < High[1] but you are using <= and >= so I'm not sure if what you are marking are truly inside bars. Are you sure what you read defines inside/outside bars to include the highs and lows equal to the previous bar? So, below is a chart using your template. I just want to make sure I'm reading these correctly because I'm prone to automate and backtest setups once I understand every detail. Tell me if my assumptions are correct.

hope you don't mind me marking on your chart. here are the trade setups I see.

1. blue outside bar
2. magenta bullish reversal bar (but no trigger till green bar)
3. bearish reversal bar

# 4 doesn't have enough volume.

I use volume this way. if you have a spike in volume in a move up, that might indicate that sellers are stepping in and the move could be over. so I'm looking for increased volume either from the signal bar or the bar before. either one is good for me.

cunparis 

I'm backtesting a little when I have time. Here is an equity curve for just inside bar (easiest to test). The profit target I've tested everything from 8 to 16 ticks. This is 8. My conclusion is that this has worked unusually well during the past 2 months with a PF of 1.5. However, testing over the past year produces results that are not as good (PF 1.14). 1.14 leaves no room for margin of error, mistakes, or anything else. It's walking a very thin line. testing longer term has given the same results around 1.14. The win % for past 2 months has been 85-90%. Even for the past year it's 80%. The problem isn't the winners, it's the losers. The stops really hurt. Now I don't take into account news or lunch. I don't think it's really important because the perf is so good for the past 2 months even not taking into account news & lunch. So even if I add support for news & lunch, the last 2 months are still going to be very good compared to the past year. Also using two targets and going to breakeven would improve results. However I don't know how to test such things with tradestation. I can do that with ninja but it's a lot of work, compared to tradestation where it's relatively easy to do simple tests. so I'm not sure what to make of it. In the past this has had major losing streaks. That could happen again at any time. or it could continue working for the next year. I guess the best would be to milk it while you can and if it stops working then be very careful. equity curve for 8 tick chart for past 12 months using commissions but NO SLIPPAGE (any slippage whatsoever will kill it since the avg trade is only $9.

Cunparis,

We've been through this before........which is why "backtesting" without concern for news or lunch is worthless.........the last time you tested like this I looked through EVERY losing trade and the vast majority of them occured during news or lunch.

See post #41

Jeff

Jeff Castille 

Cunparis, We've been through this before........which is why "backtesting" without concern for news or lunch is worthless.........the last time you tested like this I looked through EVERY losing trade and the vast majority of them occured during news or lunch. See post #41 Jeff

It's impossible to take news into account for backtesting. Well not impossible but it'd involve writing code to query forex factory to get a list of all news items over the past few years. Would be a lot of work.

I understand your concern about news, but my reasoning is this:

Even when trading during news, the past 2 months have had extraordinary performance compared to the past several years. so adding "not trading during news" should increase the performance of both proportionately.

I don't have a way to automate backtesting with news but I can pick out a few weeks where the perf was bad and do it manually to see if avoiding the news would have helped.

Hopefully someone else with more time for backtesting will be able to investigate this further.

cunparis 

It's impossible to take news into account for backtesting. Well not impossible but it'd involve writing code to query forex factory to get a list of all news items over the past few years. Would be a lot of work. I understand your concern about news, but my reasoning is this: Even when trading during news, the past 2 months have had extraordinary performance compared to the past several years. so adding "not trading during news" should increase the performance of both proportionately. I don't have a way to automate backtesting with news but I can pick out a few weeks where the perf was bad and do it manually to see if avoiding the news would have helped. Hopefully someone else with more time for backtesting will be able to investigate this further.

The past two months is not in question since the results are stellar......
it's the rest of the time that is an issue.....lol

I know the only way that I've been able to get any sort of " good " backtesting has been with Barrons economic calendar up.....where I can bring up ANY day and see the news releases to check how I really would have traded on any given day.

Thanks for giving it a shot.

Jeff

Jeff Castille 

The past two months is not in question since the results are stellar...... it's the rest of the time that is an issue.....lol I know the only way that I've been able to get any sort of " good " backtesting has been with Barrons economic calendar up.....where I can bring up ANY day and see the news releases to check how I really would have traded on any given day. Thanks for giving it a shot. Jeff

I totally agree. And it's very time-consuming. And by doing it manually our brain can play tricks on us by convincing us why would wouldn't have taken a losing trade or why we would have bent the rules slightly to take a winning trade.

But I agree with you (except I use forex factory calendar instead of barrons). I'll pick a week or two and give it a try.

I'm still interested in trading it now, just being cautious.