Grey1
Senior member
- Messages
- 2,190
- Likes
- 202
Johny,
Quote "After how many trades am I statistically likely to know that I'm within say +/- 5% of my long term win/loss ratio?How many trades until I'm within +/- 1%?
"
Best way would be to run the simulation for as many as possible to find the statistically meaningful result and then reduce the simulation runs to see if you are within your required +,- limit.
Statistically 95% of runs should fall within 2sd...
Ideally smaller runs should still be close enough to larger samples and the deviation should be as small as possible.
regards
Quote "After how many trades am I statistically likely to know that I'm within say +/- 5% of my long term win/loss ratio?How many trades until I'm within +/- 1%?
"
Best way would be to run the simulation for as many as possible to find the statistically meaningful result and then reduce the simulation runs to see if you are within your required +,- limit.
Statistically 95% of runs should fall within 2sd...
Ideally smaller runs should still be close enough to larger samples and the deviation should be as small as possible.
regards
This is an excellent tool. Thanks.
I do have a question though. Im sure it's obvious, but I'm having a mental block.
The tool assumes that each trade is a winner or a loser, in that every trade either makes £1000 or loses £500 (assuming 0.05 is entered for Risk, and 2 for Reward/Risk). Therefore, are we better using average values for our past trades? e.g If I have 50 winning trades that in total made 50,000, then my average win is 1000. If I am risking 2% of 10,000, but my average loss is say 100, then am I better off putting in 10 for r/r (reflecting the actual data, or my 'on paper' system that would have 5 for r/r? Replies appreciated.
I assume the first.
I do have a question though. Im sure it's obvious, but I'm having a mental block.
The tool assumes that each trade is a winner or a loser, in that every trade either makes £1000 or loses £500 (assuming 0.05 is entered for Risk, and 2 for Reward/Risk). Therefore, are we better using average values for our past trades? e.g If I have 50 winning trades that in total made 50,000, then my average win is 1000. If I am risking 2% of 10,000, but my average loss is say 100, then am I better off putting in 10 for r/r (reflecting the actual data, or my 'on paper' system that would have 5 for r/r? Replies appreciated.
I assume the first.
jonnyy40
Experienced member
- Messages
- 1,329
- Likes
- 11
Grey1,could you please give a brief explanation of what the Monte Carlo/Rand No.s are and how they interact with the other assumptions to give the result.(I assume the Rand No. 'randomises' the winners and losers)
I trade futures and there's a limit on the number of contracts I can/want to trade.Therefore my risk will not remain constant as a percentage of my equity,it will decrease.
if at all possible,could you please give me the calcs (if you have them available) or could you point me in the right direction of where to look so that I can run simulations that incorporate reducing risk as a percentage of equity.
I'm sorry ,I'm completely maths illiterate.
If my previous 100 trades have been 50% win/loss how statistically significant is this?
If it is within +/-5% of my long term win/loss ratio.How often will I need to repeat the process to allow for changing market conditions?i.e should I re-calculate my win/loss ratio on a rolling 100 trade basis.Or can I assume a 100 trade sample will give a reasonable fit for the next 1000 say?
I asked on another thread about Bracket-Trader.This seems to replicate in Sim. mode the actual fill delay times one experiences in futures.Is this just clever programming on the developer's part or a true reflection of what would have happened in live mode?
I trade futures and there's a limit on the number of contracts I can/want to trade.Therefore my risk will not remain constant as a percentage of my equity,it will decrease.
if at all possible,could you please give me the calcs (if you have them available) or could you point me in the right direction of where to look so that I can run simulations that incorporate reducing risk as a percentage of equity.
I'm sorry ,I'm completely maths illiterate.
If my previous 100 trades have been 50% win/loss how statistically significant is this?
If it is within +/-5% of my long term win/loss ratio.How often will I need to repeat the process to allow for changing market conditions?i.e should I re-calculate my win/loss ratio on a rolling 100 trade basis.Or can I assume a 100 trade sample will give a reasonable fit for the next 1000 say?
I asked on another thread about Bracket-Trader.This seems to replicate in Sim. mode the actual fill delay times one experiences in futures.Is this just clever programming on the developer's part or a true reflection of what would have happened in live mode?
Grey1
Senior member
- Messages
- 2,190
- Likes
- 202
Johhny,
Lets say you have a strategy called STRATEGY Z.. you want to trade the market using this strategy ..
The very first thing you would want to know is how much your maximum draw down would be if things go against you .. This is your maximum risk . This is the money you can afford to lose with out having a heart attack.. you have to drag this information out of your strategy Z..
You have wo ways..
1) Technical
In Technical you run your strategy over the past few days/month/years ( back testing ) and note the maximum risk or drawdown ..
2) Statistical ( Monte Carlo simulation )
You tell the simulation technique what your strategy edge is (Example 60% win rate , or what ever your edge is ) and the Excel program calculates how much your worse case drawdown would be.. Based upon this info you can sit back and think if you are comfortable with this drawdown ..
Monte Carlo simulation is an statistical measure of the drawdown in this EXCEL program..
Quote "If my previous 100 trades have been 50% win/loss how statistically significant is this? "
Simple.. Run the Excel program with 1000 run and compare the results with the 100 run then you get a good idea.
Quote " If it is within +/-5% of my long term win/loss ratio.How often will I need to repeat the process to allow for changing market conditions?i.e should I re-calculate my win/loss ratio on a rolling 100 trade basis.Or can I assume a 100 trade sample will give a reasonable fit for the next 1000 say?
Just once.. over the long run the probability distribution curve will have all market conditions in it..
Lets say you have a strategy called STRATEGY Z.. you want to trade the market using this strategy ..
The very first thing you would want to know is how much your maximum draw down would be if things go against you .. This is your maximum risk . This is the money you can afford to lose with out having a heart attack.. you have to drag this information out of your strategy Z..
You have wo ways..
1) Technical
In Technical you run your strategy over the past few days/month/years ( back testing ) and note the maximum risk or drawdown ..
2) Statistical ( Monte Carlo simulation )
You tell the simulation technique what your strategy edge is (Example 60% win rate , or what ever your edge is ) and the Excel program calculates how much your worse case drawdown would be.. Based upon this info you can sit back and think if you are comfortable with this drawdown ..
Monte Carlo simulation is an statistical measure of the drawdown in this EXCEL program..
Quote "If my previous 100 trades have been 50% win/loss how statistically significant is this? "
Simple.. Run the Excel program with 1000 run and compare the results with the 100 run then you get a good idea.
Quote " If it is within +/-5% of my long term win/loss ratio.How often will I need to repeat the process to allow for changing market conditions?i.e should I re-calculate my win/loss ratio on a rolling 100 trade basis.Or can I assume a 100 trade sample will give a reasonable fit for the next 1000 say?
Just once.. over the long run the probability distribution curve will have all market conditions in it..
In reply to the reference to traders with a sub 50% success rate I have an interesting example.
A NY based (technical) currency trader friend of mine has traded every day since 1991. All of his trades are stored in a database and he recently allowed me to examine them.
One of my conclusions was that he is profitable on 42.3% of his trades yet his mean gross returns have been just over 18%.
This is a function of a disciplined, and objective as it is purely systematised, approach to money management. The stops are unadjustable.
Notably he only ever risks 60 basis points of capital on any one trade, his risk being measured as the distance between the spot price and his stop.
A NY based (technical) currency trader friend of mine has traded every day since 1991. All of his trades are stored in a database and he recently allowed me to examine them.
One of my conclusions was that he is profitable on 42.3% of his trades yet his mean gross returns have been just over 18%.
This is a function of a disciplined, and objective as it is purely systematised, approach to money management. The stops are unadjustable.
Notably he only ever risks 60 basis points of capital on any one trade, his risk being measured as the distance between the spot price and his stop.
P.S,
If anyone is interested in options valuation models using Monte Carlo processes I have put together a few spreadsheets on this.
They are *true*, lognormal, random (rather than numerical sequence etc) so you sometimes get some clustering but it is interesting none the less.
If anyone wants copies let me know.
Cheers,
VN
If anyone is interested in options valuation models using Monte Carlo processes I have put together a few spreadsheets on this.
They are *true*, lognormal, random (rather than numerical sequence etc) so you sometimes get some clustering but it is interesting none the less.
If anyone wants copies let me know.
Cheers,
VN
