Semi-Automated S&R Trading
- Thread starter PetaDollar
- Start date
Mini-update:
- Last week net was about +10 NQ points ($200), 1 contract
- This week so far (Mon + Tues) is about even
- I really need to whip something up to process all those lovely xml output files! Annotated chart + results.
- On the systems side, everything is running smoothly
- Transferring more cash in to trade more contracts
- I'm now using the Java Logger class for logs. It's very easy... just add a "private static final" logger to each relevant class, and each instance can call the static logger. E.g. logger.info("submitting entry") gets you a timestamp, calling class, and calling method along with your message. With Linux I just direct the output to the log file "> out.txt 2>&1".
I'm selling NQ today... any resistance at yesterday's close (= pre-market high) and breakout of the pre-market lows.
Trouble: the program moved the stops up right away after the short entry. This doesn't happen in the automated tests, so it should be interesting to find the problem. (And learn something new about testing 
) I've added some logging to the StopMover class so we can see what it's thinking. Re-running to short the highs thus far, just one trade. The stops will move up right away but I'll have a good log which will help the debugging.
) I've added some logging to the StopMover class so we can see what it's thinking. Re-running to short the highs thus far, just one trade. The stops will move up right away but I'll have a good log which will help the debugging.A very interesting issue. I'm still using price polling for the StopMover code, as opposed to the new asynchronous price feed I recently introduced for entry searches. It works, but it takes a few seconds for the first few ticks to come in. In the meantime, the price is zero and the StopMover thinks it's sitting on a profit when short. When long, it was not an issue. I didn't discover the issue in testing because the prices were instantly available in the StopMover tests.
I'll just change the StopMover class to use the new asynchronous price feed this evening, and will be back trading tomorrow. Again, this is where an object has ticks fed to it as they come in, and does nothing otherwise (the right way to do it). This is in contrast to polling some price data structure at some periodic interval (the wrong way to do it). It is somewhat counter-intuitive because human traders work by polling. What's neat is you can put filters on the ticks before they come in, which vastly simplifies any code that is conditional on certain prices. In many cases (such as this one, where we are waiting for a certain price and will move the stop once we get it), the class doesn't even need to know the value of the tick at that point. So all of the price checking code completely drops out of the class.
I'll just change the StopMover class to use the new asynchronous price feed this evening, and will be back trading tomorrow. Again, this is where an object has ticks fed to it as they come in, and does nothing otherwise (the right way to do it). This is in contrast to polling some price data structure at some periodic interval (the wrong way to do it). It is somewhat counter-intuitive because human traders work by polling. What's neat is you can put filters on the ticks before they come in, which vastly simplifies any code that is conditional on certain prices. In many cases (such as this one, where we are waiting for a certain price and will move the stop once we get it), the class doesn't even need to know the value of the tick at that point. So all of the price checking code completely drops out of the class.
A little late to the party this morning, but same plan as yesterday. Selling around the highs thus far and selling a breakout of the premarket lows. As I was writing this, the program when short at 2746.25 and got stopped out at 2751.00 about 13 minutes later. So the bug I mentioned yesterday is fixed, just in time to lose a little money.
I went with a quick patch due to time constraints and plan on upgrading the trade management code the next time I can put together an 8 hour day to work on it. The goal is to have 100% of the important parts of the code under automated tests, and we're not quite there yet because the trade management code is a little funky (its modules are tested, but it is not involved in any integration tests).
I went with a quick patch due to time constraints and plan on upgrading the trade management code the next time I can put together an 8 hour day to work on it. The goal is to have 100% of the important parts of the code under automated tests, and we're not quite there yet because the trade management code is a little funky (its modules are tested, but it is not involved in any integration tests).
Selling above yesterday's high? I don't normally do such things. 
A break in the routine is not a good thing (see: "late to party" comment). Anyway, I'm a buyer around 2750 now.
A break in the routine is not a good thing (see: "late to party" comment). Anyway, I'm a buyer around 2750 now.Selling/buying the pre-market high/low, plus buy/selling breakouts above/below. I'm glad the computer is doing it and not me.
Now that I'm confident the program is well-behaved, a cash infusion is on the way-- enough to trade 3 different futures contracts. I can start that next week.
I'm also gathering up some bucks for another account. I think it's an important safety measure in electronic trading to have a few accounts. This way if you accidentally blow one of them up, you still have something. I purposely said "electronic trading" instead of "automated trading". There is enough risk in electronic trading to wipe out an account. For example, if one of your trades gets busted and you don't have a human broker looking out for you to cover your position, you could be in real trouble.
Now that I'm confident the program is well-behaved, a cash infusion is on the way-- enough to trade 3 different futures contracts. I can start that next week.
I'm also gathering up some bucks for another account. I think it's an important safety measure in electronic trading to have a few accounts. This way if you accidentally blow one of them up, you still have something. I purposely said "electronic trading" instead of "automated trading". There is enough risk in electronic trading to wipe out an account. For example, if one of your trades gets busted and you don't have a human broker looking out for you to cover your position, you could be in real trouble.
... and today I saw the ugly side of this decision, where I had three -5.00 losses for a total of -15.00 shorting 2768.50. Previous to 'previously', it would have been just one loss of -5.00.
Again, the issue is: once a S/R level is violated, do you trade it again (in the same direction)?
I still think yes. If price breaks through where there was previous resistance, spends a little time up there, and then drops back down into the original zone-- it seems there are now even more, and more recent, trades sitting on a loss.
It was good to see a big loss today as I am now keenly aware that it *is* possible to hit the max loss on every trade with this semi-automated system. If you look back in the journal, there were a few days where I had it completely wrong but still lost very little. So what made today different?
I'll just say this because it applies to any kind of trading: the cause was lack of imagination and conscientiousness. Here's what I saw this morning:
**********
band 3
**********
band 2 <---- premarket price in here
**********
band 1
**********
What made it a little different was that although the bands were quite distinct, there were pretty close together. The top of band 3 was yesterday's high, ~2774. The bottom of band 1 was 2760. Only 14 NQ points is a tight squeeze for 3 bands.
The strategy I set up: short the top of band 2, buy the breakout of band 3. I never considered what would happen if the price rattled around in band 3: I'd get a bunch of really bad short entries. And in fact, that's what happened.
Of course, in any strategy I set up there's going to be some nightmare scenario where every trade gets smacked for the max loss. The real question is: how likely is it?
In this case, I think too likely. Nothing really crazy happened. It was a "melt up" day on a Friday after a big week. The price stayed in well-defined bands that were clearly visible in the pre-market.
---
Goal for the weekend: sexy automated performance reports. Equity curves including simulated future equity! In the meantime, you'll have to settle for this:
Performance thus far:
Week 1: +$16
Week 2: +$200
Week 3: -$500
------------------
Net: -$284.00
It appears that you have not fully worked out all the bugs with this system and you're still "testing it"....but with real money on the line. This is a recipe for disaster.
Because simulators give you the ability to do so, I personally test all of my ideas (because that's what you have right now...only an idea) on a simulator before committing even a dime of my money. This is the real edge in trading - ie., unlike most businesses you don't have to invest real capital to see if your business will work.
Oh but wait, you don't have a simulator because your software was built from the ground up :eek:
I posted this in another thread, but the same exact wording applies to you as well:
Are you sure that's why you lost $16?
Could it be that your edge might not be materializing with real money? Has the edge disappeared?
Whenever your real money results are deviating from what you expect, this could be a very strong indication that what you thought was an edge was only fools gold...just something to think about.
Personally, I would get more defensive with the strategies and drop my position size considerably with the goal being not to make any substantial amount of money but test whether the edge still exists.
http://www.elitetrader.com/vb/showthread.php?s=&postid=3601960#post3601960