8 decks! Never tried playing Black Jack professionally so not sure if you can beat it with 8 decks or not. Shuffling doesn't hurt card counters though does it? Sorry I never was exposed to these card games growing up. Just recently watched the documentary about the MIT blackjack team and saw a connection between that and trading.;-)
Which system is better?
- Thread starter cornholio
- Start date
Thanks man. That formula gave me 6.6 consecutive losses with50% win-rate system out of 100 trades. When I did monte carlo simulation I also got average 6 consecutive losses. Good stuff. You guys think this is a good monte carlo simulation?:
I use a mechanical system to alert me of a potential trade and the rest is my own interpretation of P.A-damn I hate the word as its misused.
Bones, my point in searching for a way to figure out what a potential consecutive losing streak ones system can have is not to be able to have a "sure thing"(nothing is 100% in life) but to be able to know what is more or less average so that 1. when the losing streak comes you have confidence to keep trading knowing it is normal and 2. you can know what to risk per trade so that when the losing streak comes the draw-down doesn't take away 50% of your account. So thanks to penny for the formula and dom for telling us about Monte Carlo.
However, using the monte carlo, it shows that there was a run with 13 losing streaks. So the '6' is just an average I guess. So now the question is: Do I determine what % to risk per trade based on the worst losing streak run(13) or the average(6)? But also, the worst losing streak run keeps changing and is not constant. Once it is 13, next simulation run it is 10, etc.
For me personally I would base my max acceptable draw-down on the average losing streak while keeping in mind that 10 losing streaks is possible. Because I think if I where to base it on the worst losing streak the performance of the account would suffer greatly in anticipation of the 10 losing streak.
Now about the size of the consecutive losses as one does more trades:
50% win-rate over 100 trades you get 6:
=(LN(100)/-(LN(1-0.5))) = 6
So most would think after 1000 trades it would be 6*10 so 66 right? Nope. It is actually only 10!:
=(LN(1000)/-(LN(1-0.5))) =10
So dom was right in saying that the more trades you take the bigger the size of the losing streaks. But I knew that it doesn't grow like like 6*10. It's crazy that its 10 and not 66. I have to try this on monte carlo to see if I gets a number close to 10 also. But I already think it is true because if it was that you got 66 consecutive losses during your 1000 trades I think 100% of people would quit and there wouldn't be a trading industry. Risking 1% a trade would give you a 49% draw-down after 66 consecutive losses.
Ok I need a break.;-)
Cheers,
corn
However, using the monte carlo, it shows that there was a run with 13 losing streaks. So the '6' is just an average I guess. So now the question is: Do I determine what % to risk per trade based on the worst losing streak run(13) or the average(6)? But also, the worst losing streak run keeps changing and is not constant. Once it is 13, next simulation run it is 10, etc.
For me personally I would base my max acceptable draw-down on the average losing streak while keeping in mind that 10 losing streaks is possible. Because I think if I where to base it on the worst losing streak the performance of the account would suffer greatly in anticipation of the 10 losing streak.
Now about the size of the consecutive losses as one does more trades:
50% win-rate over 100 trades you get 6:
=(LN(100)/-(LN(1-0.5))) = 6
So most would think after 1000 trades it would be 6*10 so 66 right? Nope. It is actually only 10!:
=(LN(1000)/-(LN(1-0.5))) =10
So dom was right in saying that the more trades you take the bigger the size of the losing streaks. But I knew that it doesn't grow like like 6*10. It's crazy that its 10 and not 66. I have to try this on monte carlo to see if I gets a number close to 10 also. But I already think it is true because if it was that you got 66 consecutive losses during your 1000 trades I think 100% of people would quit and there wouldn't be a trading industry. Risking 1% a trade would give you a 49% draw-down after 66 consecutive losses.
Ok I need a break.;-)
Cheers,
corn
For me a mechanical system is the only way. When I started out I had a mechanical system that I would enter the trade, set SL and TP and then go to bed. I made money this way. Then I tried day-trading/reading "PA" and it was amazing how I was able to lose so consistently. I coudn't handle it emotionally. Ticks up: hope, Ticks down: fear. It was torture. I would close out the trade 90% of the time with a small loss or 10% of the time with a small profit. Unfortunately it took me a long time to figure out what was wrong with my game and that day-trading was not for me. But like they say: Better late than never.
You are the man. Can't believe this video is free. Watch this video if you want, you can learn a lot from it and its also entertaining.
: http://www.youtube.com/watch?v=QflVqavHHM0
cornholio:
The attached is an Excel template, working with the YASAI free MonteCarlo add-on for Excel (http://www.yasai.rutgers.edu/download.html).
It is a lightweight version of what I use.
The key is to study the drawdown characteristics, not the max. number of consecutive losses.
The tool gives you the average & stdev for whatever you set it to give you, here the P&L max DD.
Once you DD gets N stdev away from the average, your edge is likely gone. Unfortunately, trading outcome distributions are far from Gaussian, so you'll have to use a fairly large N (I use 5, most of the time) to avoid being "stopped" by some noise.
The attached is an Excel template, working with the YASAI free MonteCarlo add-on for Excel (http://www.yasai.rutgers.edu/download.html).
It is a lightweight version of what I use.
The key is to study the drawdown characteristics, not the max. number of consecutive losses.
The tool gives you the average & stdev for whatever you set it to give you, here the P&L max DD.
Once you DD gets N stdev away from the average, your edge is likely gone. Unfortunately, trading outcome distributions are far from Gaussian, so you'll have to use a fairly large N (I use 5, most of the time) to avoid being "stopped" by some noise.
Attached is the output of my MC sim for my newest system, CL151.
The first one corresponds to 1-year worth of trading (240 trades).
The second-one corresponds to 10-years worth of trading (2400 trades).
I run 10,000 iterations for each model.
I also attached the per-year view of that system, out of backtesting.
In the MC sim, scenario 1 is using the trade distribution straight from backtesting, scenario 2 is a modified trade distribution with 1/2 performance. As you can see, the yearly DD on this backtest ranges from avg - 1.6 st-dev, to +1.3 std-dev of the 240-trades simulation, and is avg - 0.7 stdev of the 2400-trades simulation (10 years). Clearly, I can expect to see larger DD in the future than on this backtest.
Sadly, when you hit DD in excess of the avg DD (even the 10-years version), there is no way (that I know of) to tell if this is a temporary rough time, or it is the beginning of the end.
And to those that say "trade your P&L curve", using a 30-trades (or whatever length) MA on the P&L curve, and tell you to stop trading under the MA & resume above the MA, I can say that I did simulate that to the Nth degree (using MC simulations), and the results are always of a severe impact on the ending P&L, with always a worse DD than just trading the system as long as the DD doesn't exceed whatever limit you have chosen for it.
The first one corresponds to 1-year worth of trading (240 trades).
The second-one corresponds to 10-years worth of trading (2400 trades).
I run 10,000 iterations for each model.
I also attached the per-year view of that system, out of backtesting.
In the MC sim, scenario 1 is using the trade distribution straight from backtesting, scenario 2 is a modified trade distribution with 1/2 performance. As you can see, the yearly DD on this backtest ranges from avg - 1.6 st-dev, to +1.3 std-dev of the 240-trades simulation, and is avg - 0.7 stdev of the 2400-trades simulation (10 years). Clearly, I can expect to see larger DD in the future than on this backtest.
Sadly, when you hit DD in excess of the avg DD (even the 10-years version), there is no way (that I know of) to tell if this is a temporary rough time, or it is the beginning of the end.
And to those that say "trade your P&L curve", using a 30-trades (or whatever length) MA on the P&L curve, and tell you to stop trading under the MA & resume above the MA, I can say that I did simulate that to the Nth degree (using MC simulations), and the results are always of a severe impact on the ending P&L, with always a worse DD than just trading the system as long as the DD doesn't exceed whatever limit you have chosen for it.
Hi Dom993,
Very interesting study, indeed.
I don't understand why the maximum drawdown gets bigger when you stop trading if the equity curve is below a certain moving average, though ?!?
It should be the opposite, no? After all, you are preventing major drawdowns from ever happening in the first place
