Welcome to NexusFi: the best trading community on the planet, with over 150,000 members Sign Up Now for Free
Genuine reviews from real traders, not fake reviews from stealth vendors
Quality education from leading professional traders
We are a friendly, helpful, and positive community
We do not tolerate rude behavior, trolling, or vendors advertising in posts
We are here to help, just let us know what you need
You'll need to register in order to view the content of the threads and start contributing to our community. It's free for basic access, or support us by becoming an Elite Member -- see if you qualify for a discount below.
-- Big Mike, Site Administrator
(If you already have an account, login at the top of the page)
I just started messing with matlab and thought this was interesting just because I can remember backtesting without bothering to mess with transaction cost.
The first chart is 1000 trades, repeated 200 times...YM sized, 1 contract, 50% win rate, 5 point target, 5 point stop...Basically coin flips for 25 bucks, win or lose. The second is the same with 3 dollar commission per "trade"..
The 3rd is 56% win rate and 3 dollars a trade.
I remember seeing something like this online but it is rather useless with just one iteration without getting a feel for the distribution.
You can see how pointless it is to backtest without including transaction cost.
Can you help answer these questions from other members on NexusFi?
Here is another simulation using the same YM sized 1 contract type bets, 200 iterations of 500 trades each, 52% win rate but this time a win is 15 YM points and a loss is 5 YM points for the typical 3:1 setups that is standard in the trading literature. 3 dollars of transaction cost per trade. You can see that this is a completely unrealistic and MASSIVE edge, even at only a 52% win rate.
The second is the same but repeated 400 times but this time with only a 35% win rate. Even at such a low win rate, with 3:1 winners size to losers only one system did not make money after 500 trades.
Comeon, no one has an opinion on this? I should add I figured out how to do this yesterday and had been waiting YEARS to be able to plot this stuff....So I'm quite excited to see it.
I would gladly go into more detail if someone just post "makes no sense"...
Well I'm less here to learn stuff than to sharpen my own thoughts...
If you just started trading though, I don't think this thread is for you...
There is a "random walk" / stochastic nature to markets/betting that you have to be a fool to not "accept".
Consider a thought experiment that you just ran a backtest on YM data 1/1/08 - 1/1/10, on intraday data taking 2 trades a day on a system you invented for YM that had metrics of 5 point profit target, 5 point stop, and wins at 50%....
The first graph is 200 traders doing exactly that but not considering transaction cost.
The second is with transaction cost...
The 3rd is with 1:1 win ratio, but the win rate of each of the 200 traders being 56%..Some win huge, some blow.
What you should conclude from this is that "metrics" on a single equity curve are totally meaningless. No matter how well you keep a journal you can't tell if your method is stochastic not in line with the market, over EV or totally random...
The second set of graphs is IMO showing the absurdity of the 3:1 win:loss > 50%...anyone who actually has such an edge should be able to Kelly/Optimal R that edge to retirement in 3 years without a problem...
it makes perfect sense and it is essential that traders understand what you are showing. It all about probabilities and expectancy. The visual representation is fantastic to show this.
Yea but my graphs neither include pyramid/money management ideas or negative dependency in the sense that a string of losers would tip a coin flip negative, as well as a string of winners would make a coin more likely to be a loser because of greed.
Thank you though, I'm just looking to shoot the shit.
I hope to cream the idea of single run "sharpe ratio" in this thread as not making sense at all.
Especially if the Ninjatrader people are watching, and what a complete joke the monte carlo engine they made for 7 is.
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...