Brain teaser 1: Is it possible to turn a losing system into profitable one? | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Very Profitable System - Roi system 355 replies

MA, Gann and Ichimoku system (profitable system!) 60 replies

Can someone code this profitable system into an EA? 52 replies

Change a losing system into a Winning system 68 replies

Brain teaser 2: Who made the most? 9 replies

Attachments: Brain teaser 1: Is it possible to turn a losing system into profitable one?
Exit Attachments

Brain teaser 1: Is it possible to turn a losing system into profitable one?

  • #1
  • First Post: Edited 2:09pm Dec 20, 2010 1:56am | Edited 2:09pm
Note: Please read the question again because I have added some clarification/modification to minimize the confusion.

Imagine that you have built a 100% mechanical system where you place your entry when a certain event happen. This event could be moving average crossover, price action at S/R, or even surge in sunspot activity...you name it, it just does not matter.

The exit strategy is simple. You just wait until the price hit your TP/SL. That means once your place your entry with predetermined TP/SL, you do nothing other than waiting.

You have tested 3 combination of TP/SL for a thousand trades and you have came up with the following numbers after spending years for live testing:

Set 1: TP = 10, SL = 20, Win% = 59%, Loss% = 41%
Set 2: TP = 20, SL = 20, Win% = 52%, Loss% = 48%
Set 3: TP = 30, SL = 20, Win% = 44%, Loss% = 56%

p/s:
1. The spread is 2 pips and it is not included in TP/SL. So your profit = 8 pips and loss = 22 pips for set 1...
2. Fixed lots for MM.
3. There is no breakeven.

You are now looking at the performance number and obviously your system is not profitable with these TP/SL. So the question arise before you dump the system or test for another set of TP/SL or even modify your entry/exit criteria: Is it possible to turn this system into a profitable one based on the above set of results? If it is not possible, why? If it is possible, how?

Please prove it before you simply say yes or no.
  • #2
  • Dec 20, 2010 2:51am Dec 20, 2010 2:51am
Quoting ForexQuant
Imagine that you have built a system where you place your entry when a certain event happen. This event could be moving average crossover, price action at S/R, or even surge in sunspot activity...you name it, it just does not matter.

The exit strategy is simple. You just wait until the price hit your TP/SL. That means once your place your entry with predetermined TP/SL, you do nothing other than waiting.

You have tested 3 combination of TP/SL for a thousand trades and you have came up with the following numbers after spending years for live...


I can actually go with the [Set 1: TP = 8, SL = 22, Win% = 59%] with reversing my trades.

Consider this.. Lets say you take 100 Trades and you win 59% with the above set of TP and SL, you will still be in loss because the Stoploss is too high compare to TP.

For 100 trades, if you win 59 then you make profit of 472 pips and lose 902 pips.

Instead if you reverse the trade (buy when there is a sell signal and vice versa)

You will win 902 pips while losing 472 pips and at the end you would get 430 pips... which is a clear win :-)

Let me know what you think about it.. does it work? if it doesnt then give me reasons why it doesnt
  • #3
  • Edited 3:21am Dec 20, 2010 3:05am | Edited 3:21am
Quoting reachjj
I can actually go with the [Set 1: TP = 8, SL = 22, Win% = 59%] with reversing my trades.

Consider this.....
Good try but please be reminded that 2 pips of spread already been deducted. So Set 1 actually has TP = 10 and SL = 20 before spread. If you reverse the trade direction then the new TP/SL should be TP = 18 and SL = 12 after spread.

Please try again.
  • #4
  • Dec 20, 2010 3:21am Dec 20, 2010 3:21am
I'm guessing...

http://en.wikipedia.org/wiki/Parrondo's_paradox

Will figure it out in a mo.

EDIT:
Or is it simply waiting for it to go 8 pts past entry, then entering with a 30 pt stop and a 10 pt target?

EDIT2: explaination:

(assuming these are all the same entries).

Price goes 8 pip in your favour 59% of the time, and 18 pips in your favour 52% of your time.
This means in 100 trials (because I know a lot of you hate working in %), 41 won't hit +8, 52 will hit +18 and 7 will hit +8 but not +18.
So you enter at +8, 52/59 will hit +18, 7 will revert to stop (which is 30 pips for 22 plus the eight you're up.

So you get 7*-30+52*10 = 410 > 0
  • #5
  • Dec 20, 2010 3:23am Dec 20, 2010 3:23am
Quoting TraderHotch
I'm guessing...

http://en.wikipedia.org/wiki/Parrondo's_paradox

Will figure it out in a mo
Nope this looks like a Parrondo's paradox but it is not any kind of paradox.
  • #6
  • Dec 20, 2010 3:40am Dec 20, 2010 3:40am
Yeah, figured, thought it was worth a guess though ;-)
  • #7
  • Dec 20, 2010 3:46am Dec 20, 2010 3:46am
Quoting TraderHotch
I'm guessing...

http://en.wikipedia.org/wiki/Parrondo's_paradox

Will figure it out in a mo.

EDIT:
Or is it simply waiting for it to go 8 pts past entry, then entering with a 30 pt stop and a 10 pt target?

EDIT2: explaination:

(assuming these are all the same entries).

Price goes 8 pip in your favour 59% of the time, and 18 pips in your favour 52% of your time.
This means in 100 trials (because I know a lot of you hate working in %), 41 won't hit +8, 52 will hit +18 and 7 will hit +8 but not +18.
So you enter at +8,...
Is 2 pips of spread included in your answer?
  • #8
  • Edited 7:19am Dec 20, 2010 6:47am | Edited 7:19am
I only have one cup of coffee in me so far this morning, but I'm going to wager that this is a dead system no matter which way you try to tweak the numbers. The problem is that the system can't correctly predict the direction often enough. Even if one were to switch the TP and SL thinking that you will cleverly reverse the roles of those numbers, you still don't know which trades in PARTICULAR will be winners and you will still lose on average.

I think the numbers have more to do with the probabilities involved in market volatility or a coin toss. Obviously, the closer target has a higher probability of being hit first, as seen in your first option with a 59% win rate. I think your discussion will boil down to the folly of MM, or maybe even the fallacy of it in the face of a losing system.

Beat me up Scotty
  • #9
  • Edited 11:13am Dec 20, 2010 10:41am | Edited 11:13am
Quoting ForexQuant
Nope this looks like a Parrondo's paradox but it is not any kind of paradox.
This is not a Parrondo, because one must have a winning game in the set to create an advancing ratchet and this has two losers and a break even game, so no ratchet effect.

set 1 1000 trades
590 * 8 = 4720
410 * -22 = -9020
is a loser
set2 1000 trades
520 * 18 = 9360
480 8 -22 = -10560
is a loser
set 3 1000 trades
440 * 28 = 12320
560 * -22 = -12320
break even

So in none of the sets is there a profitable trade set. Now I am not sure what you are asking next. Are you asking is there a way to use combinations of the 3 sets (games) to come up with a winning overall system in 1000 trades? I do not see how that would be done. Would be very interested if you had a different view.

In your question you are not allowing any other combination of TP/SL only the three sets you listed and predetermined before each trade is place.
Quote
That means once your place your entry with predetermined TP/SL, you do nothing other than waiting.

EDIT:

It just occurred to me that just because you say it has a winning % of 59%, does not mean that it has a losing % of 41% because you could have withheld the % of Break even trades. If that is the case the variables are open, because parameters are missing.
  • #10
  • Dec 20, 2010 10:47am Dec 20, 2010 10:47am
Quoting ForexQuant
Is 2 pips of spread included in your answer?
I thought you said it was included, must of read it wrong, it's something along those lines, thought it was a bit odd that I didn't need the third stat. Simples.
  • #11
  • Edited 1:05pm Dec 20, 2010 12:37pm | Edited 1:05pm
Quoting smjones
set 1 1000 trades
590 * 8 = 4720
410 * -22 = -9020
is a loser
set2 1000 trades
520 * 18 = 9360
480 8 -22 = -10560
is a loser
set 3 1000 trades
440 * 28 = 12320
560 * -22 = -12320
break even
I believe all of us will came out with the same conclusion the moment we look at the performance number without thinking deeper.

Quote
So in none of the sets is there a profitable trade set. Now I am not sure what you are asking next. Are you asking is there a way to use combinations of the 3 sets (games) to come up with a winning overall system in 1000 trades? I do not see how that would be done. Would be very interested if you had a different view.

Well I did not specify any methods to solve the puzzle. It is up to you. My question is if it is not possible, why? If it is possible, how? This is the fun part of brain teaser right?

Quote
In your question you are not allowing any other combination of TP/SL only the three sets you listed and predetermined before each trade is place.

Of course I can give you more combination of TP/SL but believe me too much info will only blind your mind.

Predetermined TP/SL is just a part of the system rules. Just stick to it.

Quote
EDIT:

It just occurred to me that just because you say it has a winning % of 59%, does not mean that it has a losing % of 41% because you could have withheld the % of Break even trades. If that is the case the variables are open, because parameters are missing.

There is no breakeven because I have mentioned that you do nothing after placing the order. So your exit will be always at predetermined TP/SL.
  • #12
  • Dec 20, 2010 12:59pm Dec 20, 2010 12:59pm
Quoting TraderHotch
I thought you said it was included, must of read it wrong, it's something along those lines, thought it was a bit odd that I didn't need the third stat. Simples.
mmm...Sorry I must caused some confusion. I have edited my original question. Please read again.
  • #13
  • Dec 20, 2010 1:09pm Dec 20, 2010 1:09pm
See what happens when you raise the TP even further to 38.

You're keeping the SL constant while raising the TP by 10 on each new trial set, and the expectancy has moved from -4.3 to -1.2 to 0. At some point the rising expectancy will turn back downwards, but there may be zone in the range of TPs in which it turns positive. For instance, if the 38 TP trial yields a win rate of 37% or more, then the expectancy is positive at that combination of SL/TP.
  • #14
  • Dec 20, 2010 1:29pm Dec 20, 2010 1:29pm
Are my calculations right?

PipTP PipSL %W %L ActTP ActSL 100Tds Inv%W Inv%L InvTP InvSL 100Tds
10 20 59% 41% 8 22 -430 41% 59% 18 12 30
20 20 52% 48% 18 22 -120 48% 52% 18 22 -280
30 20 44% 56% 28 22 0 56% 44% 18 32 -400

So the first set could be a winner in an inverted approach... Am I Right?

Regards
  • #15
  • Dec 20, 2010 1:32pm Dec 20, 2010 1:32pm
10*59-20*41 = -230
20*52-20*48 = 80
30*44-20*56 = 200

The last two are already profitable. Of course, you have put up only a very limited amount of system stats. Much more is needed before actually committing to one of them (distribution of trades, ...).

What do you mean that the spread is 2? Why does it matter?

If I buy EUR/USD at 1.3100 with a 20 point TP, I will sell at 1.3120. The spread is already included. Why do you subtract it another time? Why does your TP not include the spread? Is this a MetaTrader thing (I'm using different software)?
  • #16
  • Dec 20, 2010 1:36pm Dec 20, 2010 1:36pm
Quoting capitalist88
See what happens when you raise the TP even further to 38.

You're keeping the SL constant while raising the TP by 10 on each new trial set, and the expectancy has moved from -4.3 to -1.2 to 0. At some point the rising expectancy will turn back downwards, but there may be zone in the range of TPs in which it turns positive. For instance, if the 38 TP trial yields a win rate of 37% or more, then the expectancy is positive at that combination of SL/TP.
Yes you are right. It is possible that expectancy is still increasing after TP = 30 as long as it has not gone beyond its max expectancy point.

However, in this case we assume that we only have 3 sets of test results. We do not consider any other possibility at this moment.

Just think this scenario as a scientific experiment. You want to make your experiment as simple (minimum variable) as possible so that it is easier to make observation and draw conclusion.
  • #17
  • Edited 3:55pm Dec 20, 2010 1:52pm | Edited 3:55pm
Quoting masilva
Are my calculations right?...
Ok you are right. The 1st set with reverse trade direction is barely profitable with net profit of 30pips after 100 trades, or you may say it has an expectancy of 0.3 pips.

Original direction: TP = 10, SL = 20, Win% = 59%, Loss% = 41%

Expectancy before spread = 0.59 * 10 - 0.41 * 20 = -2.3 pips
Expectancy after spread = Expectancy before spread - spread = -2.3 - 2 = -4.3 pips

Reverse direction: TP = 20, SL = 10, Win% = 41%, Loss% = 59%

Expectancy before spread = 0.41 * 20 - 0.59 * 10 = 2.3 pips
Expectancy after spread = 2.3 - 2 = 0.3 pips

Congratulations! You have won a consolation prize! Try work harder for the jackpot.

Well I did not say there is only 1 correct answer right?
  • #18
  • Dec 20, 2010 2:12pm Dec 20, 2010 2:12pm
Well, if you have trades which have returns which are anticorrelated, it is possible for the Sharpe ratio of some mixture to be better than the max Sharpe of each.

Is this getting closer?
  • #19
  • Dec 20, 2010 7:53pm Dec 20, 2010 7:53pm
Yes, it's definitely possible to turn this system into a profitable one.

The key concept to visualize / question to ask may be stated as: What is the conditional probability of achieving a TP further away, before retracing all the way back and hitting the original SL, given that price has achieved a closer TP?

There are 3 possible courses of actions along those lines: 1) TP2 | TP1 ("hold out" for TP2 given TP1), 2) TP3 | TP1, 3) TP3 | TP2. Each turns out to have a positive expectancy: +3.3, +5.3, +0.3 pips, respectively, if I crunched the numbers correctly on my trusty 20th century HP 12C. Case 2 is optimal.

So, the modified, now positive expectancy, still 100% mechanical, system looks like this...

When the entry trigger event occurs, do nothing; sit back, observe Mistress Price and wait patiently. (he said with a perfectly straight face, seemingly not for the first time)

If the price reaches SL = 20 first, before reaching TP1 = 10, wait for the next entry trigger event.

If the price reaches TP1 = 10 first, before reaching SL = 20, we'd better carpe diem now. Enter at TP1 in the direction of TP3. Set TP = 20, i.e., the absolute value of (TP3 - TP1). Set SL = 30, i.e., abs (TP1 - original SL).

Now do nothing; wait. (there he goes again)

Adjusting for spreads, Expectancy =

P (TP3 | TP1) x TP - P (SL | TP1) x SL =
44 / 59 x 18 - 15 / 59 x 32 =
13.42 - 8.14 =
+5.3 (pips).

  • #20
  • Dec 20, 2010 10:51pm Dec 20, 2010 10:51pm
Quoting ForexQuant
Note: Please read the question again because I have added some clarification/modification to minimize the confusion.

Imagine that you have built a 100% mechanical system where you place your entry when a certain event happen. This event could be moving average crossover, price action at S/R, or even surge in sunspot activity...you name it, it just does not matter.

The exit strategy is simple. You just wait until the price hit your TP/SL. That means once your place your entry with predetermined TP/SL, you do nothing other than waiting.

You have...
it is totally possible and done every day.

all you need to do is combine the winning percentages ( = 155) and sell it as a system to newbs. "i have a system that makes 155% profit". it's not a lie, so to speak, and some idiot will surely buy it without caring about things like drawdown and sample size etc.
Congratulations ScalaFX for coming out of the closet!
0 traders viewing now