My auto trading system generate hundreds of trades a day, but...

Discussion in 'Automated Trading' started by tzlf, Mar 24, 2007.

  1. tzlf

    tzlf

    Hello, everyone. I am a trading system developer, I had developed various indicators and automated trading systems for futures market and forex market. In the latest years I reach a trading system for forex market. The system is a super short term trader, it enter the market for only a few seconds to a few minutes and then exit. Statistics show that the average profit per trade is about 0.7-0.9 pip on EUR/USD, depends on system setup. The system generates a few hundreds trades a day. The stability of the system is excellent, as I'd performed a lot of validation test with out of sample data, it turns out that it perform nearly as well as with the sample data.

    So, current the problem is the average profit per trade is a little low, at least for an individual trader like me. because the 0.7-0.9 pip per trade don't consider spread, all retail forex broker's spread is above 1 pip. But is very likely that a system like this may work for some institution traders, because the spread of interbank forex market can be as low as 0.5 pip. Think that my system generates hundreds of trades per day, and the excellent stability of system performance, even a small edge can generates big cash flow continuously.

    Currently I am looking for some one, that have enough resources (money, historical data, market access, etc.) to help me to push things ahead. I need to verify the system with the historical data that form interbank market without filtering and smoothing. Some of the technical tests I performed shows that filtering and smoothing of price data hurt the performance of the system dramatically. Currently I am using data from a well known forex broker, but it is very likely that their data is being smoothed in some way. Seem like with interbank market data, I can get more then 1.0 pip or more per trade, and even more trade number per day.

    The design of the system is not limited to forex market, I'd like to see how it works on index futures like ES, NQ, or YM if I can get high quality historical data of these market, especially bid/ask historical data.

    Please contact me by email or msn if you are interested in working with me to put the trading system in action.

    email me at tzlfjz@yahoo.com.cn
    MSN: tzlfjz@asheville.com
     
  2. 0.7 to 0.9 pips of profit per trade doesn't leave you much room to play.

    First of all, if you go the way of Currenex, your spread on EURUSD will be 1 pip many times. You also have to add the commission you pay on top of that. When you move away from the retail arena of the brokers you typically hear about (i.e., the "well known" brokers), you need to pay a commission.

    The commission is quotes as a USD figure per million USD traded. This is not quoted round turn. For example, $20 per million is a rate you might get from a Currenex firm (2nd tier prime broker) to start out if you do enough volume. This figure is $40 per million round turn or $4 per $100,000 USD traded. Since EURUSD is traded in units of EUR, the rate winds up being $4 times the current exchange rate (around 1.3300) or $5.32 (over half a pip).

    These commissions make it impossible for you to use a 2nd tier prime broker.

    If you have enough money (at least $1M or usually even more) to set up your own prime brokerage account with someone like RBS, etc., your commissions might be cut in half or more, so instead of half a pip round turn, you'll pay around 0.25 pips or maybe less.

    You'll need a consistent spread of 0.5 pips on EURUSD for this to work (assuming that your average profit is 0.9 pips - 0.7 pips would be too low). This makes it almost impossible to pull off.

    Another option is to use an ECN where instead of buying at the Offer and selling at the Bid (and always having the full spread as an expense), you might design your system to join the Bid to buy and the Offer to sell. This will eliminate the spread, but will also decrease your chances of getting filled, since someone would have to take your order after that and your order might also get a delayed fill because there will be other bids or offers ahead of you.

    It's a tricky situation. Maybe someone else has a different opinion.
     
  3. If you want, PM me your contact info and I'll call you from the office on Monday.

    I work with scalpers and algorithmic traders and have eperience with many of the liquidity pools out there.
     
  4. I just get a feeling you need to be more strict with your system assesment. I'm guessing from your requests that you're still in the newbie stage of development. Here's a few questions:

    1. How much slippage did you include in your tests? If you're going to tell me that slippage rarely occurs, then you're not testing it realistically.

    - I would simply like to know whether you are testing your models correctly / realistically. Sorry... actually you're not because you haven't included the commission(cost of trade) to start off with...

    2. What's the % profitability? Risk/Reward Ratio? I think most 3rd party backtesting software gives these figures... Also what type of trade orders did you use? Market? Stop? Limit? (Limit orders ignoring FIFO???)

    - I'm also guessing that you haven't coded the tests properly.

    Anyways... you're still new in the development business. There's nothing wrong with letting a model go and starting on a new one... Unless... you are willing to push something that most likely won't work... placing your career on line with hope...

    (What's the point of system development?)
     
  5. ccooper

    ccooper

    The comments by danger66 are absolutely correct. I have just started up with a Currenex broker, and am still using IB. The commissions are the same at $20 per million. Actually, with IB you can usually see 0.5 pip spreads in EURUSD during active trading hours, whereas on Currenex the spread is 1 pip. There are reasons not to use IB, but for what you are doing, and where you are in development, I would suggest doing your forex trading there. You can also get fairly realistic historical forex data from them, but in practice your fills will never be as good as the model indicates due to the way they route orders.
     
  6. If you cannot make more than that, you should throw away the system. And if you cannot make a better system you should look for something else to do.
    I know that this is not what you would like to hear, but sometimes reality can be very hard and frustrating.
    I've been trading forex for 5 year and i'm trading the ES now.

    Just out of curiosity i checked 03/22/2007 on EUR/USD.
    My system generated two trades between 0800 am and 0800 pm where i live:

    short 1.3372 out at 1.3354
    short 1.3360 out at 1.3320

    58 pips from which you have to deduct the spread and the commissions. That's what i call a system. Jumping in and out hundreds of times a day is to me irreallistic, very stressful, a waste of spreads and commissions and has nothing to do with a system. On top of that the profit margin is so small that the slightest miscalculation in your testing can give huge differences in reality. 0.1 difference per trade, multiplied with a few hundred trades a day will give a difference of tens of pips a day. In an optimistic test this will give a lot of dollars that will never appear in reality.

    A system has to prove that you can stay for quiet a while in the market without being hurt too much. That's the ultimate proof that you have an idea about what the market is going to do.
     
  7. ccooper

    ccooper

    Look at it this way -- the only way your system can make money is if your trading costs (slippage, commission, fixed costs) are kept to an absolute minimum. You can only do that by trading big size. Yet you can't get executions that will reflect your model except at small size. As you try to increase your size, your executions suffer, and your system profitability goes down the tubes. What you have is untenable.
     
  8. maxpi

    maxpi

    Your system needs to be tested over bar interval. If you slowed it down it might produce trades that overcame your costs and slippage.
     
  9. That' totally inaccurate. "Unrealistic?" It is not. The most successful traders I've ever come across are scalpers. I work with scalpers that are profitable. They exist. Most traders can't scalp though. It is not for everyone.

    "very stressful?" - only for traders that can't make money on it. Many successful scalpers treat what they do more like a sport than anything else. It is a redefinition of a "high." Definitely not stressful for them.

    "a waste of spreads and commission?" Why a waste? That's only so if someone is idiotically trading a gazillion times and not making money. One thing is to scalp and another thing is to scalp successfully. If you do the latter, it is definitely not a waste of time.

    "has nothing to do with a system?" Nonsense. There's a new breed of algorithmic scalpers out there that are making money. I am currently helping one integrate their system to Currenex and it trades thousands (not hundreds) of times a day! Most traders are not at this level and will never be.

    Read about the "world's most successful scalper," Paul Rotter - aka "the Eurex Flipper." And you though that hundreds of trades a day was a lot?

    http://www.trading-naked.com/paul_rotter.htm

    spike500, I'm glad you have a system that's working for you, but let's face it, the longer you keep a position open, the more risk you are taking. Because nothing catastrophic has happened to you trading this way in the past doesn't mean that it won't happen in the future. Scalpers don't run this risk. Since a position is open only a short period of time, the risk due to market exposure is severely diminished. This adds another level of risk control that is absent from other forms of trading and investing. Few traders even work about this type of risk control, but the reality is that it's a huge advantage for capital preservation.

    These are unique advantages of scalping:

    * Lessened exposure limits risk - A brief exposure to the market diminishes the probability of running into an adverse event.

    * Smaller moves are easier to obtain - A bigger imbalance of supply and demand is needed to warrant bigger price changes. It is easier for a currency to make a 3-pip move than it is to make a 50-pip move.

    * Smaller moves are more frequent than larger ones - Even during relatively quiet markets there are many small movements that a scalper can exploit.

    Don't get discouraged tzlf. If I forget to connect via MSN later on today (Monday), send me another PM.
     
  10. That' totally inaccurate. "Unrealistic?" It is not. The most successful traders I've ever come across are scalpers. I work with scalpers that are profitable. They exist. Most traders can't scalp though. It is not for everyone.

    If you work with scalpers you know what they can, but you don’t know what other traders can. I don’t know of any successful trader that manages billions of dollars who is scalping. The reason is simple: you cannot manage big amounts of money in scalping because scalping has to do with fast execution.


    "very stressful?" - only for traders that can't make money on it. Many successful scalpers treat what they do more like a sport than anything else. It is a redefinition of a "high." Definitely not stressful for them.

    I transformed scalpers into trend following traders; they all agree that they trade much more relaxed than before. So it is their conclusion, not mine. And most of them make more money than before (although this can be because they were bad scalpers, but that’s not sure per definition).


    "a waste of spreads and commission?" Why a waste? That's only so if someone is idiotically trading a gazillion times and not making money. One thing is to scalp and another thing is to scalp successfully. If you do the latter, it is definitely not a waste of time.

    It has nothing to do with successful or not. It is all about the ratio between profits and cost. My cost per trade is at least ten to hundred times lower than the cost for a scalper.


    "has nothing to do with a system?" Nonsense. There's a new breed of algorithmic scalpers out there that are making money. I am currently helping one integrate their system to Currenex and it trades thousands (not hundreds) of times a day! Most traders are not at this level and will never be.

    If you trade only 1000 times within an 8 hours session you have to make a RT every 28.8 seconds without interruption. So if it is a multiple of 1000, as you say, the system should make a RT probably within 5-10 secs average. How much profit can you make in 5-10 secs? And how big can the position be, knowing that you have to have a full fill before you can close the trade with a second transaction?


    * Lessened exposure limits risk - A brief exposure to the market diminishes the probability of running into an adverse event.

    Not true, not only the length in time of the exposure defines the risk, the direction of the markets is far more important. If you scalp against a very strong strength your 15 secs open trade is riskier than if I go with the trend for 5 minutes. The environment of the market is far more important than the length in time of the exposure.

    Who will have the biggest risk of freezing to death: someone who is running naked outside for 15 secs or the one who is running outside for at least 10 hours?
    The answer is: the one who is running in the middle of the winter, because the one who is running in the middle of the summer can run around for days without having any risk of freezing to death. So not the length of the exposure is the main risk factor, but the circumstances ( market conditions) are. You need to know when it is summer and when it is winter. That’s what I call have a system. Scalpers are fast running in and out all the time, summer and winter, because they do not see the difference between summer and winter. I make long walks in summer and hop in and out in winter.

    * Smaller moves are easier to obtain - A bigger imbalance of supply and demand is needed to warrant bigger price changes. It is easier for a currency to make a 3-pip move than it is to make a 50-pip move.

    The success of a system can be measured by what is taken out of the market. If you can take what the market offers, you have an excellent system. If the markets moves 10 pips your system should (ideally) take the 10 pips; if the markets moves 250 pips, your system should take 250 pips. Your system has to adapt itself to the market conditions. Scalpers will take the 10 pips, but never the 250 pips, because they are not able to take the big moves. You should compare the net results per year of the “thousands of trades a day system” with those of the other systems. If you can take a few big moves, the scalper will need thousands or even millions of scalps (depending on the average profit per trade) before he will have the same profit.
     
    #10     Mar 26, 2007