Simulation vs Real Cash

Hello,

I've been programming for a while now. And out of curiosity I've been building a program that trades the E-mini with a demo/api account from OEC.

I've recorded a short movie about it. http://obeleh.thruhere.net/sharemap/20120103_2202_37.avi

The fundamental thing that it does is buys for current bidprice and sells for the current ask price. When accounted for commission (which the api doesnt show) the program makes an average between 1$ and 3$ profit per transaction. And it can make a whole lot of transactions per day.

Now the thing is. Since it's all simulated I can't be sure whether the orders get filled in the same way as they do in the simulation. Can anyone have a look?

The link didn't work for me.


Real$$ is always very different from simulated, as most sim environments give you benefit on Bid/Ask.

The noise level on ES is 2-3 tics so you should be able to get

at least $20 profit per trade.........

And 1 tic on es is $12.5 so how do you compute?
Scalp 1 with $5 in commish, $5 out commish and $2.5 profit?

That probably won't work in real-life ... one of the key elements your simulated account cannot reflect with integrity is where your orders get in the queue, so you probably get lots of unrealistic fills.

If you really wanted to do this, you'd need a CME seat so that your trading comms get as small as they can, considering the volume you would do, and co-locate your trading system so that you have really minimum delay for both getting the order-book data & sending/updating your orders.

My 2 cents
D.

As far as I can see it simulates the orders to be entered on the bottom of the stack and executed when on top. Orders don't get filled immediately and if I place big orders they don't get filled all at once or filled completely at all. That is why I think the filling is quite realistic but ofcourse I haven't seen it in the "real world" ;)

This is how I compute profit/loss

Begin situation:
Realized: 87380 en Transactions: 181

End situation:
Realized: 101117 en Transactions: 2556


Income: 13737
Transactions: 2375
Worst Case Transaction Costs: 2.4
Expenses: 5700
Profit: 8037
Time: 7min
Profit per transaction: 3.384

You could make a limited test in a real account, making 100 trades - while doing the exact same on a sim account. Then you have some ground to compare real vs sim.

Who knows, you might even make some money doing so!

Obeleh,
Moving from test to live test always interesting.

Use a test control.....let the program execute 10 transactions and see what happens......you don't say what your risk per transaction is so I guess you'll see.

If you don't know the code or how to control risk with the code, then you shouldn't move to "live" test until you have control. It doesn't even sound like you're clear on the logic.

Thats a lot of transactions in 7 mins......the average volume for es for every two tics is around 2400.....so whatever testing live testing you do should be controlled....to make sure you get reasonable fills.

O,
I read your top statement again and it sounds like you are using limit orders.....buy at bid and sell at ask. Most simulated environments will give you this trade but the market won't if the tic trend is moving opposite. And with your projected volume, I suspect you'll be fortunate to get 30%.....maybe....and your fail rate will increase as well.

Try switching to market orders and see what the result is.

Most of my stuff is event based and its market in & out. If the event is wrong then its wrong.....but its always in and out immediately.