How much would you risk on a 90% trading system | Page 2 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

How much would you invest in a CTA w/ 30% returns? 1 reply

What would you do if you had a very good system? 7 replies

how much would you NEED??? 7 replies

How much would you pay for an MT4 ECN bridge? 3 replies

Would you trade this system? New system #2. 10 replies

Attachments: How much would you risk on a 90% trading system
Exit Attachments

How much would you risk on a 90% trading system

  • #21
  • Jul 3, 2014 4:28am Jul 3, 2014 4:28am
Quoting jiggz
ill put it simple from a personal point of view, if the trade that im about to take is a 90% winner, then I dont mind risking 50% or more. im just talking about a sible trade, not so sure about systems that has that winning percentage,
Wrong reasoning. 90% winning trades doesn't mean that the system is profitable.
MT4/MT5 EAs/Indicators/Alerts coder
  • #22
  • Jul 3, 2014 4:58am Jul 3, 2014 4:58am
Wow! Thanks guys and possibly girls for all the answers, I have a lot of reading and thinking to do. I guess I have to explain my self. the thing is, the sl is based on volatility, so it is allways quite big because the system trades only in high volatility with a smaller tp, it plays on the probability that the tp is hit before the sl is higher, hence the 90% winrate. therefore the risk to reward goes from 4:1 negative! at the most, to about 1:1, crazy huh? it's rarely positive tho, with the dynamic nature (negative) of the stoploss, it's a bit hard to calculate, how much to risk on the system overall, ofcourse it would be easy to just state that I will never risk more than 1 or 2% and that would surely work just fine, but since the system have a 90% winrate, I feel it would simply under perform, I know it can handle the losses without problem as I have forward tested over 1000 trades, and it appear that the losses isn't big enough to draw the profitabillity down. I might add that I didn't use any kind of money management in the tests, I was thinking maybe I could use an average over the losses? but I don't really got that part yet
Bulls are stupid Animals!especially when Im short!
  • #23
  • Jul 3, 2014 5:00am Jul 3, 2014 5:00am
Quoting fxsafe6
{quote} While we're at the Kelly, do you have an online calculator or an excel spreadsheet I can use? The site I used went offline and I'm clueless with excel
Kelly's formula is Maximum Risk = account * (winrate * (1 + RR) - 1) / RR
Note 1: This is a MAXIMUM not a recommendation. That's why broketrader divides the result by some factor, usually 3 or 4.
Note 2: This formula is pretty much useless in real life trading. Like CholiPop and Akh12851 said it depends on the distribution of the results. There is no system with perfectly fixed winrate and RR.
Hanover is also right that Kelly is too aggressive in the sense that the risk of ruin is not really taken into account. It happens that the rule of thumb of 1% is actually a good empirical solution in general.
No greed. No fear. Just maths.
  • #24
  • Jul 3, 2014 5:13am Jul 3, 2014 5:13am
Quoting roughtrader
Wow! Thanks guys and possibly girls for all the answers, I have a lot of reading and thinking to do. I guess I have to explain my self. the thing is, the sl is based on volatility, so it is allways quite big because the system trades only in high volatility with a smaller tp, it plays on the probability that the tp is hit before the sl is higher, hence the 90% winrate. therefore the risk to reward goes from 4:1 negative! at the most, to about 1:1, crazy huh? it's rarely positive tho, with the dynamic nature (negative) of the stoploss, it's a bit...
Can you share the list of the trades results? Only two columns: the pair and the result in pips (not in $$ nor in %)
(the pair is needed because the pip value and the volatilities are different)
No greed. No fear. Just maths.
  • #25
  • Jul 3, 2014 5:26am Jul 3, 2014 5:26am
Quoting PipMeUp
{quote} It is just a rewriting of the well known simplified Kelly criterion http://en.wikipedia.org/wiki/Kelly_criterion
Why Kelly criterion, everyone knows that its not efficient in the forex markets and way to agressive.As you said the market is not normally distributed, so the WIN/LOSS event will also be not normally distributed.

If i assume that the 90% winrate is the average winrate, and the distribution of that is a T distrubition which is very similar to the market's distribution then if a big tail event occurs your money gets flushed down the toilet.

I mean lets calculate that for a second if the winrate is because he has 1:9 RR and 90% then the kelly bet is 0, which is logical since this is a no edge situation - transaction cost.

However lets imagine if you really have an edge of 1 reward over risk so be that 1:8 RR and 90% WR with spreads included, then the kelly pops out 10%, yes, with a high risk situation like this even if you have an edge kelly tells you to risk 10%.

So maybe you can get away with 10% in ranging market but if a trend or a breakout happens and you are on the wrong side (in a fat tail event) 5 consecutive wins can be easily imagined, is already a 50% DD, how are you going to recover from that?

My analysis showed that 5 standard deviation events are regular in the markets so expect the unexpected, even the best winrate strategy could easily have a 20-25 consecutive loss scenario after about 10-15000 trades.

So my Risk management is certainly conservative on this and adaptive, i would recomend risking only about 2-3% of the account in "median" trading conditions, but when the consecutive losses exceed 3 then logarithmically decrease the risk.All profesional HF traders do this, then know that the fat tails are inevitable so you have to work your Risk management around it

It is the exact inverse of a martingale where you exponentially increase the risk when you lose, here we logarithmically decrease the risk when we lose, and set it back after the deviations get back to the mean
"There's a sucker born every minute" - P.T. Barnum
  • #26
  • Jul 3, 2014 5:27am Jul 3, 2014 5:27am
Attached Image (click to enlarge)
Click to Enlarge Name: forex.PNG Size: 44 KB
  • #27
  • Jul 3, 2014 5:36am Jul 3, 2014 5:36am
Quoting Proximus
Why Kelly criterion, everyone knows that its not efficient in the forex markets and way to agressive.
I and Hanover just said that the formula given by broketrader was Kelly. Both of us recommend to not us it ;-)

Quote
As you said the market is not normally distributed, so the WIN/LOSS event will also be not normally distributed
Strangely the results I get, winner or not, always look like a gaussian or the sum of two gaussians with one mode at the average winner and the other mode at the average loser...
No greed. No fear. Just maths.
  • #28
  • Jul 3, 2014 5:43am Jul 3, 2014 5:43am
Quoting PipMeUp
{quote} Can you share the list of the trades results? Only two columns: the pair and the result in pips (not in $$ nor in %) (the pair is needed because the pip value and the volatilities are different)
Attached Image (click to enlarge)
Click to Enlarge Name: eurusdh1.png Size: 44 KB


2 august 2012, an upper wick broke the 5th standard deviation border of the weekly median.In a normal distributed population this should hapen once every 4700 years, yet it happens 1-2 times a year in EUR/USD.

With a Kelly Criterion of 10% risk of that 90% system, these events would probably devastate an account.Forex markets are very unpredictable
"There's a sucker born every minute" - P.T. Barnum
  • #29
  • Jul 3, 2014 5:58am Jul 3, 2014 5:58am
Go with a fixed lot size.
Double your lot size every time you triple your account and you will be fine
  • #30
  • Jul 3, 2014 6:17am Jul 3, 2014 6:17am
Quoting Proximus
{quote} {image} 2 august 2012, an upper wick broke the 5th standard deviation border of the weekly median.In a normal distributed population this should hapen once every 4700 years, yet it happens 1-2 times a year in EUR/USD.
If it wasn't completely off topic in this thread I would show you that even if the markets were normally distributed your calculation of the standard deviation on the posted chart is wrong.
No greed. No fear. Just maths.
  • #31
  • Jul 3, 2014 6:20am Jul 3, 2014 6:20am
subbed for later to download Hanovers excel sheet


Quoting Proximus
{quote} Why Kelly criterion, everyone knows that its not efficient in the forex markets and way to agressive.As you said the market is not normally distributed, so the WIN/LOSS event will also be not normally distributed. If i assume that the 90% winrate is the average winrate, and the distribution of that is a T distrubition which is very similar to the market's distribution then if a big tail event occurs your money gets flushed down the toilet. I mean lets calculate that for a second if the winrate is because he has 1:9 RR and 90% then the kelly bet is 0, which is logical since this is a no edge situation - transaction cost. However lets imagine if you really have an edge of 1 reward over risk so be that 1:8 RR and 90% WR with spreads included, then the kelly pops out 10%, yes, with a high risk situation like this even if you have an edge kelly tells you to risk 10%. So maybe you can get away with 10% in ranging market but if a trend or a breakout happens and you are on the wrong side (in a fat tail event) 5 consecutive wins can be easily imagined, is already a 50% DD, how are you going to recover from that? My analysis showed that 5 standard deviation events are regular in the markets so expect the unexpected, even the best winrate strategy could easily have a 20-25 consecutive loss scenario after about 10-15000 trades. So my Risk management is certainly conservative on this and adaptive, i would recomend risking only about 2-3% of the account in "median" trading conditions, but when the consecutive losses exceed 3 then logarithmically decrease the risk.All profesional HF traders do this, then know that the fat tails are inevitable so you have to work your Risk management around it It is the exact inverse of a martingale where you exponentially increase the risk when you lose, here we logarithmically decrease the risk when we lose, and set it back after the deviations get back to the mean
Nobody uses the full Kelly criteria, it is normally some multiple lower. A guy on BMTF created a neat spreadsheet that combined kelly criterion with the risk or ruin. Very nice spreadsheet.
  • #32
  • Jul 3, 2014 6:47am Jul 3, 2014 6:47am
Quoting PipMeUp
{quote} Kelly's formula is Maximum Risk = account * (winrate * (1 + RR) - 1) / RR Note 1: This is a MAXIMUM not a recommendation. That's why broketrader divides the result by some factor, usually 3 or 4. Note 2: This formula is pretty much useless in real life trading. Like CholiPop and Akh12851 said it depends on the distribution of the results. There is no system with perfectly fixed winrate and RR. Hanover is also right that Kelly is too aggressive in the sense that the risk of ruin is not really taken into account. It happens that the rule of...
I am aware of the issues with Kelly, I used 1/8 usually. But it's a fast way to calculate how much to risk when dealing with multiple systems and changing parameters (like when systems go into period of losses etc)...

So if anyone can give me a spreadsheet of it, it would be much appreciated. No idea how to ''program'' it into Excel.
  • #33
  • Jul 3, 2014 7:07am Jul 3, 2014 7:07am
Quoting PipMeUp
{quote} If it wasn't completely off topic in this thread I would show you that even if the markets were normally distributed your calculation of the standard deviation on the posted chart is wrong.
Can you please elaborate that.I dont use bollinger bands since it is a very bad representation of standard deviation, i use directly the formula which calculates it.The population size is 1 week, aka 120 bars in H1 and i use the corrected standard deviation formula based on a sample size.

Attached Image


To my knowledge it works in any distribution.
"There's a sucker born every minute" - P.T. Barnum
  • #34
  • Jul 3, 2014 8:55am Jul 3, 2014 8:55am
Quoting roughtrader
Wow! Thanks guys and possibly girls for all the answers, I have a lot of reading and thinking to do. I guess I have to explain my self. the thing is, the sl is based on volatility, so it is allways quite big because the system trades only in high volatility with a smaller tp, it plays on the probability that the tp is hit before the sl is higher, hence the 90% winrate. therefore the risk to reward goes from 4:1 negative! at the most, to about 1:1, crazy huh? it's rarely positive tho, with the dynamic nature (negative) of the stoploss, it's a bit...
Hi Roughtrader,

As you have forward tested over 1000 trades, i asssume that you already have the complete data of the trade taken, like profit/loss pips per trade, or in percentage profit/loss per trade as per your trading system, for amount of 1000 trades personally i will consider the data is quite reliable, then from the data, maybe you can try to calculate based on fact what is the largest drawdown it ever had, "for example" if the largest drawdown it had in pips is 500pips cumulative, then you can adjust it with the drawdown percentage that suits you, if the largest drawdown you are willing to accept is 15 percent then perhaps you can try 5000pips mm, so if you had that 500pips drawdown, it's 10percent of your account, you still have 5 percent spare from your acceptable drawdown limit, if it reach 750pips or 15%, maybe you could stop for a while for evaluation..

In percentage if the largest drawdown it had is "for example" 10 percent cumulative for 1% risk per trade mm (the data had maximum 10 consecutive loss trades), then you can raise your risk to a little below 1,5% if your limit drawdown is 15%, of if your dd limit is 30%, you can raise your dd to a little below 3% per trade, maybe 2% or 2,5%, just adjust it according to the test data and your limit drawdown..

the key is you have to be sure your 1000 trades data is quite reliable, usually backtest data is more profitable than reality..

just share my thought here, sorry if my language is not clear..
live your life heartfully..
  • #35
  • Jul 3, 2014 9:50am Jul 3, 2014 9:50am
Quoting roughtrader
Wow! Thanks guys and possibly girls for all the answers, I have a lot of reading and thinking to do. I guess I have to explain my self. the thing is, the sl is based on volatility, so it is allways quite big because the system trades only in high volatility with a smaller tp, it plays on the probability that the tp is hit before the sl is higher, hence the 90% winrate. therefore the risk to reward goes from 4:1 negative! at the most, to about 1:1, crazy huh? it's rarely positive tho, with the dynamic nature (negative) of the stoploss, it's a bit...
Well well, if you have 1000 trades alteady i assume its on demo, then find out the distribution of the wins and losses. See how "fat" is the tail of the distribution, and also find out the maximum consecutive losses, and the average consecutive losses.

After you know the distribution of the losses which is important how frequently they distribute consecutively after one enother, the risk should not be more in % than the maximum, and also find out that the maximum is how much deviation.

As you see 5 deviation events are frequent so if the maximum consecutive losses are below that, then expect worse upcoming so lower the risk even more.

However the initial risk can easily be the average cons loss adjusted, since if you would try my logarithmic risk decrease approach it would be more eficient to adjust the risk based on that if many consecutive losses come, then to have low risk all time that would minimize profit.

The risk should be flexible i think not constant, if you say you trade based on low volatility then you should avoid day trading and focus on night trading, the sprread is bigger but probably less breakouts and trends during that time.
"There's a sucker born every minute" - P.T. Barnum
  • #36
  • Jul 3, 2014 1:38pm Jul 3, 2014 1:38pm
Quoting chrisyuto
{quote} Hi Roughtrader, As you have forward tested over 1000 trades, i asssume that you already have the complete data of the trade
Unfortunatley I have deleted it, I use Forex Tester 2 for my testing, but thats ok, I can easily repeat the procedure and post some result here,
Thanks for your idea, I have a good gut feeling about it, purhaps we can elaborate it when I get back with some data, it will take me a few days tho.
Bulls are stupid Animals!especially when Im short!
  • #37
  • Last Post: Aug 14, 2014 3:01am Aug 14, 2014 3:01am
Risk comes from not knowing what you’re doing. Never invest in a business you cannot understand. Most people get interested in day-trading currencies/ equities/ futures when everyone else is. The time to get interested is when no one else is. You can’t buy what is popular and do well. For some reason, people take their cues from price action rather than from values. What doesn't work is when you start doing things that you don’t understand or because they worked last week for somebody else. The dumbest reason in the world to buy a currency pair/ equity/
futures contract just because it’s going up. Diversification may preserve wealth, but concentration builds wealth.
0 traders viewing now