Totally random trading system graphs

April 13th, 2011, 08:41 PM

Quoting dutchbookmaker

You can see how pointless it is to backtest .

ThatManFromTexas

There... I fixed that for ya....

I second that! I have traded for several years, but never spent to much time backtesting. According to the backtests, I am supposed to be well up in 9 figures now. If only someone could convey this to the market, I would be most appreciative...

And why someone would backtest without transaction costs escapes me...

April 13th, 2011, 09:13 PM

MXASJ

I'm interested to learn how you did that in Matlab

. I've been spending most of my time with R and a package called quantstrat. I have Matlab so any pointers would be great...

Yea I'm done with R...fuck that.

% Nrand = input('How many trades? ');
Trades = 500;
Starting = 5000;
NumCars = 1;
Iterations = 400;
WinPer = .35;
Commision = 3;

CarsPts = NumCars * 5;
YmW = 15 * CarsPts;
YmL = 5 * CarsPts;

Win = 0;
Loss = 0;
Profit = 0;

tradmatr = zeros(Iterations, Trades);
%tradestation - 2.36 perside
for nloops2=1:Iterations
Profit = 0;
for nloops=1:Trades
Toss = rand(1);

if (Toss < WinPer) %winning
Win = Win + 1;
Profit = Profit + YmW;
end
if (Toss > WinPer) %losing
Loss = Loss + 1;
Profit = Profit - YmL;
end
Profit = Profit - Commision;

% v(nloops) = Profit;
tradmatr(nloops,nloops2)= Profit;
end

end
%
% disp(['win: ', num2str(Win)]);
% disp(['loss: ', num2str(Loss)]);
% disp(['profit: ', num2str(Profit - Starting)]);
% disp(['Commision: ', num2str(Commision)]);
% disp(['%: ', num2str(Loss/Trades)]);
% disp(['-----']);
plot(tradmatr);

April 13th, 2011, 09:23 PM

Plug and play code. Love it...

April 13th, 2011, 09:42 PM

MXASJ

Plug and play code. Love it...

Please pay me back with some insight.

April 13th, 2011, 10:03 PM

dutchbookmaker

Please pay me back with some insight.

Buy low, sell high!

Kidding.

I think we are in the same camp where the coin toss represents a system with no edge, commish means you are paying to toss the coin, so you are doomed to failure no matter what money management system you use if you play the game over time.

You need a system with positive expectancy after commissions and other expenses.

What I've been doing in R (and thanks for making me want to try it in Matlab) is to take the daily lognormal returns of my system under test and see how "randomizing" those results would alter my terminal PnL, my MDD, and the various ratios people look at. I then am trying to work out with what level of confidence will I be making a given level returns, or conversely at what level of confidence will I go bust.

Is that a Monte Carlo sim? I have no idea.

I do know I can't do this natively in Ninja as it is a multi instrument basket, so I don't really use the Monte Carlo sim capabilities of Ninja.

BTW most of the bookmakers (er... equity market liquidity providers) in my town are Dutch.

Totally random trading system graphs

Discussion in Traders Hideout

Totally random trading system graphs