can't believe this method have become so popular in just few days.
cool result.but would you trade live without SL?:
I ran another test and 4 lots, 10k Jan1 - feb 9, No SL, TP 50, GBP/USD.
Results Below.
I think there might be a bug in the EA as it opened a buy order on 2/8/07 @ 1:00 for no reason. Thats why the last order is down 5000. If thats fixed it would be a sell and would have close with 50pips.
This thing is amazing if you have 10 to throw at it without feelings.
Hey NKE! The 'half kelly" is a money management system. Essentially the kelly system is a bet sizing system. It was made popular by the MIT prof. Edward Thorp, the man who inveted hi-lo card counting for blackjack and broke Vegas. After cleaning up, he published his card counting strategy (resulting in the prosecution of card counters from that day on). He used the Kelly System for placing his bets.
He started a hedge fund and used the "helf kelly" system for placing trades... basically using the kelly but the initial position is half of a normal kelly bet. His success brought this system popularity in securities trading and variations of it have been used by traders and fund managers ever since!
I use a half kelly system on my initial trade position and add to my position as the account increases... still using the kelly system to either get out of my trades or figure out initial position size.
Here's the link for the Kelly:
www.bjmath.com/bjmath/thorp/paper.htm
and you can always google Kelly System
Happy Trading!
mwilkinw can you kindly explain as simple as possible the practical approach of using the formulas when it comes to forex trading? I think this is very important, it deals with position sizing to maximize investment growth but I just cant undertand it.
-Mikhail
mwilkinw can you kindly explain as simple as possible the practical approach of using the formulas when it comes to forex trading? I think this is very important, it deals with position sizing to maximize investment growth but I just cant undertand it.
-Mikhail
Kelly% = P - [(1-P)/(W/L]
For example, take Pouria's trades for Jan-Feb 2006 (I'm just starting optimizing this method...). If you'd use a 46 p. SL and a 33 p. TP target (spread not taken into account), you'd have 5 losses and 33 wins (P = .87). So, Kelly% is
Kelly% = .87 - [(1-.87)/(33/46)] = .69.
This means, that you can risk 69% of your capital on a single trade. If you have a 1000 p. capital, and you can risk 690 p., you can you can open 690/46 = 15 positions (again, spread and margin considerations nottaken into account).
Since Kelly% is based on historical performance, and history never repeats itself (at least, the better parts of history, the worst ones tend to repeat over and over again), and since using exact Kelly% leads to huge drawdowns, and since most people don't have the nerve for such a risk, it's customary to use half the Kelly's percent, i.e., in our example, open 7 positions. I don't even have the nerve for that, and usually don't risk more than 15% simulteneously...
mwilkinw, please correct me if I got something wrong. I teach statistics, but always have all my calculus messed up...
sans peur et sans reproche
Indeed, this is of utmost importance. Kelly's formula provides you with the optimal percent of your capital you can risk on a single trade. To use the formula, you have to know the history of a trading system performance: what is the size of a win W (or average win size if it varies), what is the size of a loss L (or its average) and the proportion P of wins out of total number of bets. The assumption is that you place one bet (trade) at a time. Kelly's percent is calculated then using this simple formula:
Kelly% = P - [(1-P)/(W/L]
For example, take Pouria's trades for Jan-Feb 2006 (I'm just starting optimizing this method...). If you'd use a 46 p. SL and a 33 p. TP target (spread not taken into account), you'd have 5 losses and 33 wins (P = .87). So, Kelly% is
Kelly% = .87 - [(1-.87)/(33/46)] = .69.
This means, that you can risk 69% of your capital on a single trade. If you have a 1000 p. capital, and you can risk 690 p., you can you can open 690/46 = 15 positions (again, spread and margin considerations nottaken into account).
Since Kelly% is based on historical performance, and history never repeats itself (at least, the better parts of history, the worst ones tend to repeat over and over again), and since using exact Kelly% leads to huge drawdowns, and since most people don't have the nerve for such a risk, it's customary to use half the Kelly's percent, i.e., in our example, open 7 positions. I don't even have the nerve for that, and usually don't risk more than 15% simulteneously...
mwilkinw, please correct me if I got something wrong. I teach statistics, but always have all my calculus messed up...
Hey Ilanr! Thanks for doing some of the clarifying for me! The Kelly formula is a great formula for traders
That formula is correct. The kelly formula allows you to figure out your position size based on past performace. (that is why I like backtesting so much)... Having your total profit percentage, your average win, and your average loss will allow you to formulate how large your trade or (bet size) will be as a percentage of your equity.
This is a great way to increase your profits, because as your equity grows your bet sizes grow. This allows you to add to your winning trades and minimize your losses with a set SL.
Figuring out where to take your profit and where your SL is... and backtesting to receive your Profit%, Average Win, Average Loss Data... based on your historical performance, gives you a much better system of increasing your profits and managing your downside without burning out your account.
However, as was stated for securities trading I suggest a 'half kelly" system. I use 80% of the kelly... 80% of the kelly figure is practical and will optimize your rate of return. Determine how many trades you are likely to have on at one time and then divide your 80%-Kelly value by that number of trades.
Increasing your position by the Kelly-Value determined by your increased equity will exponentially grow the size of your winners and keep your losers at a steady loss as determined by your SL.
If you use a strict money management system and know your profitibility, average wins and average losses.. theoretically you will be profitable.
This system was made famous by Dr. Edward Thorp and is used to beat the dealer in blackjack... this is how he broke Vegas and eventually made billions with his hedge fund from the 60's to the 80's.
Optimum Risk Percent = W% - [(1-W%)/(W/L)]
(W%) is your winning percentage, (W) is the average size of your winning trades, (L) is the average size of your losing trades
This is a critical system for traders who want optimal rates of return. For practical application, use the 80%-Kelly Value for your trading.
Plug in your data and solve left to right.
Happy Trading!
P.S. Ilanr.. you're a statistician... I'm a mathematician!
- #75
- Feb 11, 2007 3:54am Feb 11, 2007 3:54am
- | Joined Dec 2005 | Status: It's only money | 1,413 Posts
TRADE FOR SWAP
snip... The kelly formula allows you to figure out your position size based on past performace. (that is why I like backtesting so much)... snip... Ilanr.. you're a statistician... I'm a mathematician!
.
sans peur et sans reproche
- #79
- Feb 11, 2007 7:58am Feb 11, 2007 7:58am
- | Joined Dec 2006 | Status: o_O | 1,511 Posts
Ok ive manualy backtested EURUSD for september,october and november,with stop of 14 including the spread and here are the results ( they of course may not be accurate 100%)
W-winners or profitable trades
L-looosers or losses
SEP
when tp=15 you get
15W-10L = +85
when tp=20
13W-12L = +92
OCT
when tp=15
10W-12L= -18
when tp=20
9W-13L = -2
NOV
when tp =15
10W-9L= +24
when tp=20
9W-10L = +40
Max consecutive wins when tp(15)=4, when tp(20)=3
Max consecutive losses when tp(15)=4, when tp(20)=5
So far January showed best results ( posted earlier) for EURUSD , i havent tested any other pairs.
W-winners or profitable trades
L-looosers or losses
SEP
when tp=15 you get
15W-10L = +85
when tp=20
13W-12L = +92
OCT
when tp=15
10W-12L= -18
when tp=20
9W-13L = -2
NOV
when tp =15
10W-9L= +24
when tp=20
9W-10L = +40
Max consecutive wins when tp(15)=4, when tp(20)=3
Max consecutive losses when tp(15)=4, when tp(20)=5
So far January showed best results ( posted earlier) for EURUSD , i havent tested any other pairs.
Ok ive manualy backtested EURUSD for september,october and november,with stop of 14 including the spread and here are the results ( they of course may not be accurate 100%)
W-winners or profitable trades
L-looosers or losses
SEP
when tp=15 you get
15W-10L = +85
when tp=20
13W-12L = +92
OCT
when tp=15
10W-12L= -18
when tp=20
9W-13L = -2
NOV
when tp =15
10W-9L= +24
when tp=20
9W-10L = +40
Max consecutive wins when tp(15)=4, when tp(20)=3
Max consecutive losses when tp(15)=4, when tp(20)=5
So far January showed best results ( posted earlier) for EURUSD , i havent tested any other pairs.