Proof of Profitability | Page 2 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Breakeven first!!! Then i'll deal with profitability! (If all goes well) 1 reply

Prove profitability without revealing trade conditions? 36 replies

Profitability 0 replies

Vegas Profitability 3 replies

Risk Multiples and Profitability 1 reply

Attachments: Proof of Profitability
Exit Attachments

Proof of Profitability

  • #21
  • Mar 30, 2015 5:57am Mar 30, 2015 5:57am
Quoting PipMeUp
The flaw #1 is in the first post. When you write P(TP)=SL/(TP+SL) --I fixed the typo-- you're assuming a random walk with no edge. Therefore, per definition, you can't have any strategy even slighly profitable. From there, there is no point in optimizing the SL and TP: the expectancy is always 0 before spread and commission.
Thanks for the hint that there seems to be something wrong. Did you change something in my post? Because what there stands now is not right either:

Correct would be:

P (SL) = TP / (SL + TP) = 20 / (1 + 20) = 20 / 21 = 95%

And in the next sentences there is a mistake, too, It must be:

Spoon-fed: Stop loss Take Profit in pips divided by the sum of take profit in pips and stop loss in pips equals the ratio of hitting the stop. If your stop loss is 1 pip and your take profit is 20 pips than it is likely about 95% that you are being stopped out.

If you are able to change this, I would appreciate if you could do so.


I am not assuming a random walk. I use random generated trade series and I pick out those one that are at least a bit profitable.
  • #22
  • Mar 30, 2015 6:08am Mar 30, 2015 6:08am
Quoting stt
JT4xPro, I think looking at the stoploss vs profit has some flaws - even though it is largely correct. I can show the flaw by giving a simple counter example. support you have a strategy (for now lets just assume it is just one signal and our position sizing etc is equal) that predicts the move of 50 pips very accurately when you look at 1 hours fwd but in next 5 mins, the price goes against it by say 10 pips in many cases. Now the relationship of stoploss size vs TP size...
Hi stt,

that is a good point. In your example a stop loss smaller than 10 pips would lead to a worse result. The stop loss cannot be minimized further than the strategy is allowing it. It must stay profitable. For the other direction it is the same. The stop loss cannot be maximized further than the strategy is allowing it. With a bigger stop loss the losses will grow bigger, too. There will be a point when the losses in sum will outweigh the winners in sum. And before that there might be a point of balance for the stop loss where the profit is at its maximum.
  • #23
  • Mar 30, 2015 6:40am Mar 30, 2015 6:40am
Quoting JT4xPro
I am not assuming a random walk. I use random generated trade series and I pick out those one that are at least a bit profitable.
You do so implicitely by using the formula. It assumes no edge at all. The average profit you make is the sum of the money you make when you win and the money you lose when you lose, which is negative. You win with probability P and lose with probability 1-P. The average is TPxP - SLx(1-P). Assuming no edge this sum is zero. Solving in P gives:
Inserted Code
TPxP - SLx(1-P) = 0
TPxP = SLx(1-P)
TPxP = SL - SLxP
TPxP + SLxP = SL
P = SL / (SL+TP)

And 1-P = TP/(SL+TP)

Assuming both the formula is correct and the trader has got an edge is a contradiction.

PS: I can't modify you posts!
No greed. No fear. Just maths.
  • #24
  • Mar 30, 2015 7:31am Mar 30, 2015 7:31am
Quoting PipMeUp
{quote} You do so implicitely by using the formula. It assumes no edge at all. The average profit you make is the sum of the money you make when you win and the money you lose when you lose, which is negative. You win with probability P and lose with probability 1-P. The average is TPxP - SLx(1-P). Assuming no edge this sum is zero. Solving in P gives: TPxP - SLx(1-P) = 0 TPxP = SLx(1-P) TPxP = SL - SLxP TPxP + SLxP = SL P = SL / (SL+TP) And 1-P = TP/(SL+TP) Assuming both the formula is correct and the trader has got an edge is a contradiction....
I think now I understand what you mean. And you are right, the formula assumes no edge at all. I use the formula to plot the curve "Chance to hit SL...". The curve shows that the stop loss is hit more often the smaller the stop loss is in relation to the take profit.

I do not use the formula for the curve "Drawdown from 1000 trades...". That curve shows that the drawdown is smaller when the stop loss is smaller. That is hard to get in my opinion because one would probably think the less often you are stopped out the less the drawdown will be (at least that was my assumption when I started to deal with that topic).

So, both curves are related to stop loss but only the first one is related to the intended formula. The second curve is estimated as described (algorithm, trades are generated by randomness,...).
  • #25
  • Mar 30, 2015 7:50am Mar 30, 2015 7:50am
Quoting JT4xPro
I do not use the formula for the curve "Drawdown from 1000 trades...". That curve shows that the drawdown is smaller when the stop loss is smaller. That is hard to get in my opinion because one would probably think the less often you are stopped out the less the drawdown will be (at least that was my assumption when I started to deal with that topic)..
I understand that. The problem is that you select only a set of successful random trading sequences. This is called survivorship bias.
No greed. No fear. Just maths.
  • #26
  • Edited 8:48am Mar 30, 2015 8:25am | Edited 8:48am
PipMeUp makes a very valid point here which previously escaped me. Even though you are filtering profitable outcomes from a random sequence of 1000 trades, the subset of profitable outcomes are still random in nature.

However (please correct me if I'm wrong) I would imagine that provided you had a trading approach with a positive expectancy to start with (eg. by setting the distance to stop as the same as distance to profit target and testing over a sufficient sample size and possibly across multiple instruments and ensuring a positive expectancy outcome with a statistical degree of confidence), J's conclusions reached would still hold in that you would be looking for the optimal MINIMUM stop distance in maximising positive expectancy and minimising drawdown as opposed to widening a stop to give room for price movement to breathe which I assume would be conventional wisdom speaking?
  • #27
  • Edited Mar 31, 2015 2:45am Mar 30, 2015 2:11pm | Edited Mar 31, 2015 2:45am
Quoting PipMeUp
{quote} I understand that. The problem is that you select only a set of successful random trading sequences. This is called survivorship bias.
Hey PipMeUp! Thank you for taking the time comprehending my thoughts and giving me some criticism. Only with reflection there can be improvement. To examine it in a kind of scientific way there should be a view not only on the series that end up with a positive result.

Here I did the estimation with
A) positive: Only results that are at least a bit positive
B) negative: Only results that are at least a bit negative
C) both: every result

Attached Image


As I am really tired today, I am not going to write anything more about the graph (for today ). Have a look by yourself if you like to.

I have recoded the algorithm and put it into a spreadsheet. Who likes can perform the estimations with it. That shall make it a bit more transparent and more comprehensible.

You can do a single calculation (1000 trades), 100 runs each with 1000 trades and you can run a series. Therefore you have to put in the range for stop loss, take profit and so on. Have a look by yourself. The red cells with white font are the input fields.
Please let me know if you find bugs or any other mistakes or if it works fine.
Attached File(s)
File Type: zip Spreadsheet DD.zip   87 KB | 333 downloads
  • #28
  • Mar 30, 2015 2:20pm Mar 30, 2015 2:20pm
Quoting Copernicus
PipMeUp makes a very valid point here which previously escaped me. Even though you are filtering profitable outcomes from a random sequence of 1000 trades, the subset of profitable outcomes are still random in nature. However (please correct me if I'm wrong) I would imagine that provided you had a trading approach with a positive expectancy to start with (eg. by setting the distance to stop as the same as distance to profit target and testing over a sufficient sample size and possibly across multiple instruments and ensuring a positive expectancy...
Hi C!
Have a look at the post above. It is also related to your post. In another post you wrote you would like to take part at the comparison between different stop losses but you cant because you are not using an EA for your trend trading approach. If you send me your history (maybe for EURUSD at the beginning) including open time, open price and take profit of the trades I could code an EA that will imitate your trades in a backtest. We could run the backtest with every different stop loss to see how the result would have been affected. We could estimate the optimal stop loss value for your trading style. You can pm me for that purpose if you want.
  • #29
  • Edited 6:44pm Mar 30, 2015 5:59pm | Edited 6:44pm
Hi J

Thanks greatly for the suggestion. I will send you trade details tonight (my time) by this thread. It's all out there anyway, so no need for a PM unless you would prefer it this way. It would be a great exercise for any interested in honing the approach I use.....but please don't go to too much trouble as it is a bit of a selfish desire on my part.

Just with your example in Post 27 alone with the 1000 trades, whether they be random or not, those curves tell a story in themselves that were previously just hunches for me and certainly suggests we should be looking for optimal minimum conditions as opposed to maximal conditions to improve performance (to either minimise losses with a pure radom walk series net of commissions and slippage, or maximise profits net of commissions and slippage) for approaches with a positive performance bias.

Thanks once again.

C
  • #30
  • Mar 30, 2015 10:14pm Mar 30, 2015 10:14pm
Hi J

Below are the trade detail records for the period 1 Jan 2013 to 31 Dec 2014. Unfortunately subsequent to this period, the approach has changed significantly but for the purposes of testing, an analysis of these trades would be greatly appreciated.

The dataset is limited being only 482 trade records and furthermore I only have a narrow data set for each instrument which I have sorted in the spreadsheet. Given the restricted sample size per instru,ment, is there any way the analysis can be done for the entire dataset, or does it need to be restricted to a single instrument.

If you can't achieve anything out of this data, then don't worry about it and I can revisit this when I collect a far larger sample over time for a single instrument.

Just a word on the technique:

  1. It is a trend following technique with no profit target and a trade capital risk of 0.5% per trade. The 'no profit target' might be a problem for you. If it is, then don't worry about analysing.
  2. The time is recorded as GMT+1 during this period (I think). The broker has subsequently changed time frames for trade records to GMT +10;
  3. The data is based on trade by charts functionality so the trade entry values and trade exit values are mid-way between the spread. For the purposes of testing, just use the values as is, and don't worry about spread.

Let me know if you can do anything with this and once again, please don't bust a gut on this. Only do it if you can easily apply the data. I am more than happy to simply use your statistical methods and data to make inferences on how this might be applied to my approach.

Thanks mate

C

Attached File(s)
File Type: xlsx Copernicus Trade Details.xlsx   88 KB | 314 downloads
  • #31
  • Mar 30, 2015 11:24pm Mar 30, 2015 11:24pm
I would say you are doing so far. I was watching your trading performance and can say this simple theory is working! how long?? no idea... but as long as things keep changing up and down and mostly in a profit, then you are on the right track.
Cheers,
  • #32
  • Mar 31, 2015 6:32am Mar 31, 2015 6:32am
I don't think it works forever. I have been testing so many EA (Coded by myself), one of the testing was few pips SL and good TP with aiming different RR ratio scenario. I found that it works if the market environment is permitted, however it suffer DD and blow the account once the market change to volatile. So in my opinion, it will work unless you know when to use them.

After many years of studying, analysing, testing, i finally can make profit, but must adapt to change in pertaining to market environment. I am trading manually now, but will slowly to convert my strategy into full robot, however, my mistake of letting the trades goes gave me big DD, as you check my trade explorer. Overall my trading strategy is still good. Now got to build EA to help me manage SL and TP once i manually enter the trades. The key for my success is average daily range.

Just my 2cents. Anyway, wish you good luck.
  • #33
  • Mar 31, 2015 11:44am Mar 31, 2015 11:44am
Quoting FxMegaCB
I would say you are doing so far. I was watching your trading performance and can say this simple theory is working! how long?? no idea... but as long as things keep changing up and down and mostly in a profit, then you are on the right track. Cheers,

I canīt tell you how long it will work. That is a question I was already asking myself. But I can tell you how long it still works. I tested from 2007 till now. From about 2011 the strategy started to become profitable with sl and tp in different setups. Before it was only breakeven or mostly unprofitable. But from 2011 till now backtests are proofing profitability. Cheers!
  • #34
  • Mar 31, 2015 11:54am Mar 31, 2015 11:54am
Quoting ~bull.bear~
I don't think it works forever. I have been testing so many EA (Coded by myself), one of the testing was few pips SL and good TP with aiming different RR ratio scenario. I found that it works if the market environment is permitted, however it suffer DD and blow the account once the market change to volatile. So in my opinion, it will work unless you know when to use them. After many years of studying, analysing, testing, i finally can make profit, but must adapt to change in pertaining to market environment. I am trading manually now, but will slowly...
I do not think too, that it will work forever. An interesting sight on market behavior and developement deliveres the adaptive market hypothesis by Andrew Lo (http://en.wikipedia.org/wiki/Adaptive_market_hypothesis wiki link. Searching for the originial paper and reading is worth it!). Strategies are evolving but they do not live forever. Market is changing and strategies have to do so as well when the profitability shall remain.
I trained the strategy on a time sample in 2013. The out of sample test shows growing profits the nearer the out of sample testing period approaches to the training period. That is an interesting behaviour I think.

Thanks for your wishes :-)
  • #35
  • Mar 31, 2015 12:02pm Mar 31, 2015 12:02pm
Quoting Copernicus
Hi J Below are the trade detail records for the period 1 Jan 2013 to 31 Dec 2014. Unfortunately subsequent to this period, the approach has changed significantly but for the purposes of testing, an analysis of these trades would be greatly appreciated. The dataset is limited being only 482 trade records and furthermore I only have a narrow data set for each instrument which I have sorted in the spreadsheet. Given the restricted sample size per instru,ment, is there any way the analysis can be done for the entire dataset, or does it need to be restricted...
I come back to you as quick as possible. But it will take some days as eastern is coming and I am on a little journey for the holidays.
  • #36
  • Mar 31, 2015 12:27pm Mar 31, 2015 12:27pm
All due respect this is all bla bla bullsh*t. If you have a trading idea, or a strategy or an EA just spit it out already and we will test it ourselves with different ratios.
If we are talking about random position opening with different SL TP scenarios - then it is also BS.
  • #37
  • Edited 1:15pm Mar 31, 2015 1:05pm | Edited 1:15pm
Quoting JT4xPro
{quote} I do not think too, that it will work forever. An interesting sight on market behavior and developement deliveres the adaptive market hypothesis by Andrew Lo (http://en.wikipedia.org/wiki/Adaptive_market_hypothesis wiki link. Searching for the originial paper and reading is worth it!). Strategies are evolving but they do not live forever. Market is changing and strategies have to do so as well when the profitability shall remain. I trained the strategy on a time sample in 2013. The out of sample test shows growing profits the...
Ok, here is my opinion. This is not a knock on you or anything that you posted so please do not be offended: I am SO SICK of academics claiming things like this. They usually always test data on a wider timeframe (daily) and they do "dumb" systems like MA crossovers or 200MA cross, factor models or something. They DO NOT take into account that you could have risk control on a higher frame (say daily) and use a smaller frame to enter in within that framework; The daily might not look profitable, but the smaller scales they almost never mention are.

What you DO want: Sadly enough, I would liken a lot of this to gambling theory and analogies: IE if you find an edge, can know when you are in your edge, then keep generating the random event process that gives you your edge. So in effect, you are just re-drawing from the same favorable distribution of P&L results, and YOUR MISSION becomes to find many processes that can do this as often as possible.

I think it is possible, I just don't think that academics have framed it in even close to the right way- It is mostly "well this first difference" or this second series or these five factors do not reveal as much of a profitable process and LESS about just acknowledging- YOU KNOW NOTHING. But if you have a positive process, keep flipping the coin and hit it as much as possible.

OR BASICALLY: Why would you expect the MACD or 200MA crossover or whatever to be sacred? Just find something that generates favorable events as many times as possible.. IT IS ALL ABOUT THE EVENTS. So basically, the MA or MACD/RSI or whatever just transact less and give you fewer favorable opportunities so why do them? I think very few people remotely approach this in the right way and I CANNOT STAND ACADEMICS. I honestly think if you find support/resistance or whatever you want to call it with market profile, that you can setup a line: If it is above the line, buy and hold. If closes below, you are short. I think you can trade these really well and that only really involves WAITING, controlling risk, and taking chances.

Haha. He is probably an academic now that I have said this, but this is one of the best applied books I have read:
http://www.amazon.com/Modeling-Tradi...+systems+bandy

*End of rant*

EDIT: the academics also never take into account that you could have a hedging position (actual option) or could be using some kind of straddle (forex "hedge"): IE once it confirms on the side of my line, I lift the other side and there is still PLENTY of room for profit. Controlled risk and profit run- but they never usually consider this and I have never really seen them interview actual traders also anyway.. you know, if you can't do something maybe you should talk with someone who can right? Jeez.
  • #38
  • Mar 31, 2015 7:07pm Mar 31, 2015 7:07pm
Quoting JT4xPro
{quote} I come back to you as quick as possible. But it will take some days as eastern is coming and I am on a little journey for the holidays.
Mate, have a great break :-)
  • #39
  • Apr 1, 2015 9:08am Apr 1, 2015 9:08am
Quoting skillz16
{quote} I honestly think if you find support/resistance or whatever you want to call it with market profile, that you can setup a line: If it is above the line, buy and hold. If closes below, you are short. I think you can trade these really well and that only really involves WAITING, controlling risk, and taking chances.
Can you define a setup out of that? Rules that are defining
- the line
- when to open buy / sell
- when to close buy / sell
- controlling risk

Personally I would give it a try as I like simple systems. They are easy to code and we would find out quickly if there is something we could develop out of it.

P.S. Maybe you donīt like academics. And maybe that is ok for you. But please - that is the wrong place to talk about it. Take part at this discussion if you like it. There is no need to spoil it if you donīt.
  • #40
  • Apr 1, 2015 9:11am Apr 1, 2015 9:11am
Quoting B-Naire2B
All due respect this is all bla bla bullsh*t. If you have a trading idea, or a strategy or an EA just spit it out already and we will test it ourselves with different ratios. If we are talking about random position opening with different SL TP scenarios - then it is also BS.
Yeah, I wonder why, but somehow it really did just start to smell a little bit like bullshit.
0 traders viewing now