You are using an out of date browser. It may not display this or other websites correctly.
You should upgrade or use an alternative browser.
You should upgrade or use an alternative browser.
Block box trading
- Thread starter Jimmy3
- Start date
Maybe -
http://www.tower-research.com/
http://www.janestcapital.com/
Then there are your run of the mill quant funds-
Renaissance, AQR, D. E Shaw, etc
http://www.tower-research.com/
http://www.janestcapital.com/
Then there are your run of the mill quant funds-
Renaissance, AQR, D. E Shaw, etc
Quote from Jimmy3:
What's the point of testing it with live orders ? Back testing programs can account for slippage. It's not a high frequency system so impact slippage is negligable
Then why do so many systems that backtest well utterly fail in live trading? If you haven't run it live then you do not have reliable and verifiable results.
Hedge funds, trading firms and the lik get hundreds of these proposals a year. If it is just backtested, no one will even look at it unless you have a track record from a previous system. If you have out of sample paper traded on a demo account it is BETTER, but not by that much.
Backtested systems can account for slippage, but they have a harder time accounting for other things such as bad ticks and fills. This problem becomes acute when you start doing high frequency algorithmic trading.
So let me give you an example. You have a high frequency ES system, that has an average trade time in the <30 sec range. Just a scalping system, so you want to be in and out of the trade as quick as possible.
Your system gets a signal due to a price change or whatever, and you place a trade, and get filled. Then you place your closing trade order, but you are 500 deep in the queue and your order does not get and the market moves against you a couple of ticks. Trading on a demo account may have filled the order at the price if it does not have L2 data or time and sales.
In the meantime( like 20 sec ), your broker/data feed backfills a tick which would, based on your systems logic, would not have made the trade.
So there you are, losing money in a trade that your never would have taken if you were backtesting and one that you may have exited at a profit if you were trading on a demo account.
And then, some shops won't believe you because they suspect every backtested system is curve fitted.
If you have a system that has run real time for a reasonable amount of time that is fairly profitable, you will not have a problem finding money. Raising funds now is like raising money for a startup company, you need angel investors or trade it from your own cash.
Backtested systems can account for slippage, but they have a harder time accounting for other things such as bad ticks and fills. This problem becomes acute when you start doing high frequency algorithmic trading.
So let me give you an example. You have a high frequency ES system, that has an average trade time in the <30 sec range. Just a scalping system, so you want to be in and out of the trade as quick as possible.
Your system gets a signal due to a price change or whatever, and you place a trade, and get filled. Then you place your closing trade order, but you are 500 deep in the queue and your order does not get and the market moves against you a couple of ticks. Trading on a demo account may have filled the order at the price if it does not have L2 data or time and sales.
In the meantime( like 20 sec ), your broker/data feed backfills a tick which would, based on your systems logic, would not have made the trade.
So there you are, losing money in a trade that your never would have taken if you were backtesting and one that you may have exited at a profit if you were trading on a demo account.
And then, some shops won't believe you because they suspect every backtested system is curve fitted.
If you have a system that has run real time for a reasonable amount of time that is fairly profitable, you will not have a problem finding money. Raising funds now is like raising money for a startup company, you need angel investors or trade it from your own cash.
Quote from Jimmy3:
Are there any prop firms out there that hire algo traders based on back testing results of mechanical systems they want to trade ?
Of course if you're in the position for 30 sec backtest will not work. My system is designed to catch intraday swings. It makes 6 points (2 lots ES) per day on the average usually on 4-5 trades. Max drawdown - about 3K. Since only market orders are used I really don't know how real life results can be different from those from the backtest