my journal 2

[Yamato]

That's an interesting one - although the additional volume would strengthen your direction, you would also put yourself in a race to get in and out. One way to keep the advantage would be to keep the system secret and run a "live" calls service.

Yes, I like this idea better. Which basically means keeping it as secret as possible.

 

[Adamus]

yes but I and probably Travis have no desire to run a live calls service nor to sell the systems, it's just a question of whether you're stupid to discuss your system in public or not. Although I'd sell it to you cheap.

A sort of snobbery for the trading classes. Snobismo!

 

[Dommo]

yes but I and probably Travis have no desire to run a live calls service nor to sell the systems, it's just a question of whether you're stupid to discuss your system in public or not. Although I'd sell it to you cheap.

A sort of snobbery for the trading classes. Snobismo!

Thanks, but I wouldn't buy any trading system.

 

[Adamus]

Travis, how did you manage to post that message before me? If you could get your system to do that with orders, you'd be made.

Seriously I had no idea you were into marketing and all that. Or do you see an advantage in having a following of traders who leap onto your signal as soon as you give out? I guess it's front-running. Hmm, in fact maybe I'll do it too.

 

[Yamato]

Yeah, as a creator of systems, I find it inconceivable not just to buy but even receive other people's systems for free. They're all too complex for me and somehow I always think mine are better.

On the other hand, when someone explains to me a trading idea, I usually try it. But maybe not even that right now, because, like Adamus, I have more ideas for systems than time to test and automate them.

 

[Yamato]

Travis, how did you manage to post that message before me? If you could get your system to do that with orders, you'd be made.

Seriously I had no idea you were into marketing and all that. Or do you see an advantage in having a following of traders who leap onto your signal as soon as you give out? I guess it's front-running. Hmm, in fact maybe I'll do it too.

I don't think I posted any message before yours.

It's not that I am into marketing, at all. I like money though... I am into most things that will give me money. You know - that's actually the reason why I got into trading. I didn't understand everything you said above, but I hope I got the general sense.

 

[Yamato]

pivot system

Continuing from here:
http://www.trade2win.com/boards/trading-journals/85510-my-journal-2-a-119.html#post1217344


I was really good, wasn't it. But there is a problem. If I take a sample of two different weeks, July 5th to 23rd, the systems is unprofitable.




That's what I meant by taking it as a quiz. I will now see what was wrong, fix it... what could be wrong, try to fix it, and see if it will work on another 2 weeks.

The good news is that the pivots are calculated correctly also on the OUT sample weeks, and they match on the TWS chart perfectly.

Ok, i drank half a bottle of sparkling wine, made a phone call, and now I am back on the pivot system.

It's hard to see what could have made it fail so badly. Maybe it's perfectly fine. I'll need to download another 2 weeks as OUT sample. Also because it didn't lose that much. On the other hand, the exiting at the end of the session in this case helps profit.

Ok, screw the quiz mode. I can't just download 2 weeks at a time forever and ever. I will now download a whole 3 months, keeping the rest as OUT sample.

There. I have done it. I now have 3 months of data on a 15-minute timeframe, attached:
View attachment EUR_IDEALPRO_20100524_to_20100810.txt

 

[Yamato]

This guy never fails a movie:
http://www.ilovewatchingstuff.org/watch-11855-Keeping-Mum

He's got great taste.

 

[Yamato]

pivot system

Damn.

Here's the data from all 2010, GMT plus 2 (Athens):
View attachment EUR_IDEALPRO_20100104_to_20100810_GMT_plus_2.txt

This unusual time zone is devised so that the session ends exactly at midnight, and tradestation calculates the pivots the same exact way that TWS (and everyone else) does. The only way to achieve this - as far as I know - is to get the session to end at midnight. And to do that I have to either speak Greek or Finnish (Athens/Helsinki timezone).

So guess what.

I worked my ass off and created a system that worked on two weeks. Then I tested it on two other OUT of sample weeks, and it failed in those two other weeks.

Then I downloaded 2 months, and I worked my ass off and created a system that looked beautiful and that worked perfectly on the 2 months, and didn't seem overoptimized.

Then I downloaded 6 months and it lost in all the out of sample months.

Now I will try it for the third time. I don't have much more data available. So this time I won't use all the data I have.

But for now i can say this:

1) god bless the out of sample
2) this quiz is harder than I thought, but even more fascinating
3) I need to use SHORT entries as well, to verify if my SHORT strategy fails when we're in an uptrend and viceversa, which seems to be the case with the versions I've created so far. No wonder the LONG-only strategy made money: the EUR went up for both months I tested it on.
4) I should never again choose an in-sample period where the market goes only in one direction.

 

[Adamus]

Re: pivot system

Hey, bad luck. Since you're having a tough time, I'll let all the spelling and grammar mistakes go this time.

If you think about it, what is so special about that formula for pivot points? It's nothing special at all really, in fact it's no better than Fibonacci numbers or the Golden Ratio. Not that it can't work, it's just that the prices only seemed to bounce off the pivot points whereas in fact they were bouncing 'nearby' which sucks because you'd have to put your entries and your stops further away from the pivot points, cutting down the profit on the reduced move in-between.

I'm playing around with high frequency 5 minute bars and systems for them in NinjaTrader and after a couple of hours of sifting through reams and reams of data, I've just discovered that NinjaTrader has got a bug where my backtest has 3 losing trades in one bar at prices that are way above and below the OHLC. Crazy beta software. Anyway, on the upside, it means my system will backtest a lot better once the bug is fixed.

 

[Yamato]

Good work to you as well. I am thinking now that it will work though. But I don't want to speak too soon. Good work with your tests.

 

[Yamato]

Let's put some music, to cheer us up.

 

[Yamato]

 

[Yamato]

 

[Yamato]

 

[Yamato]

 

[Yamato]

I don't believe it!

It seemed perfect even with the falls, short, long, up, down... in all situations... and yet now that I tried to use it in the out-of-sample period, the usual crap: it loses exactly in the out-of-sample.

It makes me appreciate more the profitable systems I created.

But it is so so so so SO bad... I suck.



You see that part where it goes down? The first half of the chart? That's exactly the out sample. I did this for the third time, and failed again. Each time my out-of-sample trades look like this chart.

 

[Yamato]

pivot system: i give up

I give up due to being persecuted by the out sample. Each time I had the code working perfectly, smoothly and profitably on the in-sample, the out sample failed me. Well, good, because had i not done it, i would have wasted the next year thinking I had a great system, like i did with previous systems.

Here's my last code the way I had it before giving up.

Inputs: test(0), test2(0), test3(0);
variables: Pivot(0), Res1(0), Sup1(0), nearness(0), trailing_trigger(0), ma_entry_value(0), ma_exit_value(0),stoploss_value(0);

Pivot = (HighD(1) + LowD(1) + CloseD(1))/3;
Res1 = 2 * Pivot - LowD(1);
Sup1 = 2 * Pivot - HighD(1);

nearness = 0.0017;
ma_entry_value = 3;

trailing_trigger = 0.0030;
ma_exit_value = 9;
stoploss_value = 0.0030;

If marketposition = 0 and date>=1100105 
and low <= Pivot + nearness and lowd(0) >= Pivot - nearness
and c > AverageFC(c, ma_entry_value)
and lowd(0) > lowd(1)
Then buy("Long") this Bar;


If marketposition = 0 and date>=1100505 
and high >= Pivot - nearness and highd(0) <= Pivot + nearness
and c < AverageFC(c, ma_entry_value)
and highd(0) < highd(1)
Then sell("Short") this Bar;



If c < AverageFC(c, ma_exit_value) and (openpositionprofit > trailing_trigger) then exitlong ("L trailing exit") this bar;
If c > AverageFC(c, ma_exit_value) and (openpositionprofit > trailing_trigger) then exitshort ("S trailing exit") this bar;

If openpositionprofit < -stoploss_value then exitlong ("L stoploss") this bar;
If openpositionprofit < -stoploss_value then exitshort ("S stoploss") this bar;

 

[Yamato]

another one of those sleepless nights

**** them all. It's hot, they don't gratify me, no promotions, no bonuses, ever. I am the only one working all the time without coffee breaks nor phone calls, and they promote everyone else.

**** them all.

I am staying up all night and showing up at work whenever i feel like. If I can't sleep is because they don't provide the right environment for me. If they fire me so much the better - i will rest on my parents' backs.

**** them, too.

What now.

I took 4 types of different sleeping pills and I still can't sleep by how pissed off I am about my unhappy life.

 

[Yamato]

Luckily at least the discretionary nightmare is over. I feel much better now.

However, as I say this, i realize that precisely today I have placed my first discretionary trade in a long long time, because the GBP looked very oversold - my usual bottom-picking, when i think "it can't go any lower" and occasionally I blow out my account.

**** them all.