Reverse Strategy Ideas?

[vesSon6]

Sample sizes should be quantified in time, not necessarily transactions. Market conditions (and volatility) proceed beyond 22 times

The sample dimensions and time is really irrelevant here. I am only concerned about shifting TP, SL and reproducing the precise reverse of a given set of results keeping all about a system constant except entrance direction. One ought to be able to do this given set and a timeframe of information whether it 1000 or be.

I'm in class right now so I will get on in about an hour to react to the other articles and place the egy that I'm using. Yet again, the egy is irrelevant. We are just concerned about creating the opposite, as long as the entrance rules are the same it doesn't matter.

[BSesckViStiro]

It's possible to get up to modeling caliber that is 99%.

You need an older build of MT4 (208 and earlier I think ), and tick information. I've got both, but nowhere near tick information to make backtests. And tack on the issues that I have in keeping information, and openings in the information, it's very painful to work with.

[TesSaaxcs00]

This does not make any sense at all to me.

Makes perfect sense.

Trades: 679. Wins: 254 (37 percent ) Losses: 425 ( 63 percent )

Let's assume a neutral distribution.

Now if you double the TP to match the SL you would also cut your wins in half, ie:

679 trades. 552 losses, 127 wins.

Now if you further half the SL, that means the SL is double is likely to acquire
hit, cutting your wins in half an hour.

679 trades. 63 wins. 616 losses.

At that point the number of wins is no longer statistically significant, and variance over the years may be /- 20 percent of the entire.

What I'm getting at is this: Your rules are more or less random and have no advantage. All you're doing is messing with SL and the TP and getting results that less or more match a random distribution.

Any wins based on these rules are the result of luck. But over enough time that your account will trend to 0 if you trade it.

But logically by doing the reverse should not the preceding losses currently be wins?

If you'd retained the TP and SL the same, it probably would have been fairly close (plus/minus sample variance if you alter the dates). But you did not. Even when you had, this method wouldn't be profitable. Even when 254 wins to 425 losses is the mean, and you turn them to receive 425 wins and 254 losses, that less than 2:1, and that means that you won't break on the .50 ROI.

This is not a disperse dilemma, altho spread would certainly come into play when factoring in profit. It's a simple problem. It's what happens when you don't have an advantage.

Edit:

Sample sizes must be quantified in time, not automatically trades. Market conditions (and volatility) proceed beyond 22 days

Should be based on both. You want a wide range of trades, but you also must undergo a lot of different market requirements.

[testo75]

Hello All, I just have some thoughts which I want to set forth and my logic supporting them and the results from analyzing them I would like to share which are currently vexing me to no end and want to hear you all of your thoughts on why this is or possibly how I could address this issue and determine if its worth pursuing. I'm a developer so this is not a matter of my ability or skill merely am not understanding this is not working when every object of logic says that it needs to. So heres the problem...

I'm sure...

it is because the majority of losses come from the tiny fake outs that occur during consolidation periods or whatever. You do get some wins because it is going to call the trades right sometimes. Now, however, since you are moving the opposite way the fakeouts stay, but as you are now moving against your winners, those are now lost by you . It is like losing the trades as well as the fakeouts you'd have one the first way. And by the way, those little fakeouts usually lose no matter how you trade them. It is like entering the market and exiting at the exact same moment. Here is a pic to explain what im referring to. This is having a simple MA cross egy:



you will find just five wins and ten losses (5/10) so basically, 33% wins, 66 percent declines. Now let us undo the trades:



now we've got just two wins and thirteen losses (2/13) or 13.33percent wins and 86.66percent losses. Those little fakes are. The only method to have a profitable system would be to filter those out.

I hope this helps to explain it somewhat


oh, and this is all based purely on the number of wins and losses, no SLs or TPs. Rabid's post talks more about those.

[esskestis]

The sample size and time is really irrelevant here. I am concerned about repeating the exact reverse of a set of results keeping everything about a system constant entry leadership and switching TP, SL. An individual ought to be able to do this given a time and set of data if it 1000 or be using a system that makes a 100 trades.

Youre not thinking this through.

The time period does matter because it drives volatility. Here is an example:

Take a full candle (close is lower than open) with adequate sized tails on both ends. You say to yourself, market stunk now, ended down'. Does that imply that day, that everyone lost money? Of course not. You know the result of the candle, but was it formed? Can the market have a rally and remain at its high for 80 percent of the day only to nosedive at the end? Can it begin down however a mid-day rally made the large to get an epic struggle by the bears bring it back down? Was it down daily and it just so happened to end down - handed?
Now, this is just 1 candle, 1 day. How many days of differing volatility to you want to expose your EA to?

The response is: A LOT OF THEM!

Therefore, in the brief time period you posted, its possible that you have inadvertantly done some curve-fitting of your own. Maybe you chose the series of candles that favorite rapid down movements and slow upward movements (quitting you out more). Or maybe you didn't. Only way would be to keep on testing.

[vesSon6]

Wow lots of great discussion going on in here! Thanks for all the great replies guys! Though I believe my goals continue to be misinterpretted.

I'm not worried about how this ea will do forward testing or below every single market state or over a decade worth of information or on every time frame or all of the things you've indicated. Is create the exact results that are opposite given a fixed time interval. That time interval doesn't matter if its one day or 2 weeks or 10 years. If my EA makes 10 transactions in 1 month, great. If 90% of them fail, I don't care. I just need to figure out a systematic way to reverse anything egy was used if it be entrance or a egy using a'edge' though it occurred to not be profitable. I mean exactly this, by reversing. .

Original Trade:
Long TP:5, SL:10
The commerce hit the SL and dropped.

So what exactly does this mean? This means that during the course of the transaction my TP was not struck but nevertheless my stop loss was. By reversing the direction of the transaction I'll enter in the exact same level therefor, price will move on as it did and it will hit on my TP that was the SL now that commerce is a winner. Nothing complex.

Many of you might say this is stupid and you want to account for many market conditions or anything but thats not the point, its the idea at hand here. What if we have a system backtested over 5 years with tens of thousands of transactions and conducted across all market conditions with what we believe to be a advantage but when we test it that the outcomes are 80% losers.

We can use the same theories demoned above in that single trade and use it to the entire system effectively so we now have 80% wins and 20% losers. Over settings and the same time interval.

Again many of you might say this doesn't matter cause its not testing and it didn't account for many market conditons. The point by being able to do this is to reveal a few things about our systems that neglect, and open us up to a new edge by reversing the transactions. The egy was a ranging egy playing selling tops and buying bottoms, the system neglected and was backtested over 2 years. But by reversing the transactions and getting profitable results you now realize that hey this EA is doing a excellent job of recognizing set ups, but they are meant to get a breakout egy instead etc..

Take a peek at the 2 charts I've posted and you will see I got this way of reversing to work now. My reduction is now the% triumph. And also you are able to observe the transactions happen in the same level of their previous counterparts and the outcomes have all been reversed precisely also.

Thanks guys, keeps the thoughts coming. The charts posted reveal that this can be carried out. And before any of you ridicule the R:R here with SL:20 and TP:4 I want to stress that thats not the issue here, it might not be smart, yes I concur. But we are searching for a systematic technique to use to any system that fails to reverse it. It just so happened with this particular EA I was testing it on thats how the levels proven to be.

[TesSaaxcs00]

So what exactly does this mean? This means that throughout the course of this transaction my TP was never struck but yet my stop loss was. So therefor by reversing the direction of this transaction I will enter in the specific same amount, price will move on exactly as it did and it will hit my TP that was formerly the SL so now that commerce is a winner. Nothing complex going on here.

Until it goes pips in one direction, then 11 back.

If you would like to reverse the machine rules you need to keep the MM, SL and profit goals. If you do, you can not cut them in half, the rules and the outcome change.

We can apply the same theories demoned above in that single exchange and apply it to the whole system effectively so we now have 80% wins and 20% losers. Over the specific same time interval and preferences.

As was said, you can not reverse something which's not symmetrical. If you flip atleast from the cases mentioned above, the expectancy, you end up cutting off your ROI proportionally.

[vesSon6]

Unless it goes pips in one direction, then 11 back.

If you would like to reverse the machine rules you have to keep the MM, SL and profit goals. If you do, you can't cut them in half, the principles and the result change.

My motives for cutting off my SL and TP in half switching them about was to keep the same MM and chances which price would reach that level.

In the event the first trade hits my SL of 5 and never hits my TP of 10 that means when I reverse it I need to create my previous SL my new TP. If I had to keep them exactly the same then it wouldn't really be the same trade cause now price would have to go an extra 5 pips at the exact same way where it previously hit the SL just to hit my new TP

Example
EURUSD price is 1.3000 I enter Long
SL at 1.2995
TP at 1.3010
Fundamentally because of 2pip spread price has to go 3pips against me to hit SL and 12 pips in my prefer to hit TP

Result. Reduction, price hit SL at 1.2995

Reverse the transaction
EURUSD price is 1.3000 I enter brief
SL at 1.3014
TP at 1.2999
Fundamentally because of 2pip spread price has to go 12pips against me to hit SL and 3pips in my prefer to hit TP.

Result: Win, price hit TP at 1.2999

Notice how by clipping my TP in half (actually just shifting my SL and TP) I maintain the same likelihood that price will hit that level when reversing the trade? Otherwise if I maintained a 10pip TP still then price would have to go an extra 5pips south in my brief trade for a win. It wouldn't be the same trade. Hopefully my case cleared that up a bit. The spread is actually just the only problem here when everything takes place in a continuous environment when backtesting.

I know the problem with symmetry as well, no way around that this won't ever be perfect. When talking about this closed environment of data that is backtest again that's largely a result of the spread. The Expectancy remains the same as I maintain the same probabilities for price moving a certain way.

[TesSaaxcs00]

My motives for cutting my SL and TP in half switching them around was to keep the same MM and probability that price would reach that amount.

Uhm, but you are not.

In the event the first trade hits my SL of 5 and never hits my TP of 10 that means when I reverse it I need to create my previous SL my new TP. If I were to keep them the same then it would not truly be the same trade cause today price would need to go an additional five pips in that same way where it formerly hit the SL only to strike with my new TP

Right, but price does not move directly in a lineup. It wiggles, which means price is able to move back 5 pips before going ahead 10 pips. If your SL reaches 5 pips, you are getting stuck in the wiggle.

EURUSD price is 1.3000 I input brief
SL at 1.3014
TP at 1.2999

I think your numbers are mistaken. If you are setting your TP 1 pip along with a 14 pip SL price won't even cover your own spread. Should you go short at bid 1.3000 your breakeven will be 1.3000 - spread*stage, spreads are tacked on to the ask, maybe not the bid, so you cover them when you close the trade. That is unlike buying at the ask, in which your ask includes the spread.

See how by clipping my TP in half (actually simply shifting my SL and TP) I maintain the identical likelihood that price will strike that amount when reversing the trade? If I maintained a 10pip TP still price would need to go an additional 5pips south in my trade to be a win. It would not be the same trade. Hopefully my case cleared that up a little. The spread is actually only the problem here when everything takes place in a continuous environment when backtesting.

No, it does not. If price moved in a straight line, maybe, but it moves back and forth. The nearest departure will be executed in this example. You can't get a perfect reverse of the trade the symmetry issue is reverse the ruleset.

[esskestis]

Right, but price does not move straight in a line. It wiggles, so price can move back 5 pips before going forward 10 pips. If your SL is at 5 pips, you're becoming stuck at the wiggle.

Thats exactly right and were the OP is becoming lost. The price is not linear!