my journal 2

[Liquid validity]

Replying as I read.

"I know you run a residential server with redundancy".
Server at home with backup power, connection and hardware
What is that? I run a server at a friend's house, and, as a job, he runs an ISP company, so it is totally reliable from all points of view. I am not doing high-frequency trading anyway.

"Obviously you prefer it to a commercial VPS". What is that? Probably I answered this next question in the paragraph above.
VPS= virtual private server:
PhotonVPS - MB Trading
Commercial Network Services
Running this for better connection stability etc:
Microsoft Windows Server | Hyper-V | Virtualization | Virtual Server | HyperV Workload Performance

You're using a lot of terms I ignore, but once again I can confirm to you that my friend, who runs an ISP company, and has my server at his house, is totally reliable. In fact we're now using a laptop, so to avoid any power outage problems, and we have a secondary connection, too.

Let me know if you have more questions, and please phrase things more simply.

Sorry I keep forgetting english is not your first language, probably because your english is so good

 

[Yamato]

Thanks for the detailed explanation. Yes, English is my second language. Thanks for your compliments.

 

[bbmac]

Okay, so just so I am clear:

Your definition of when the present systems enabled combo has failed (from your previous recent postings on the matter) is when it reaches- 200% of max relativized historical drawdown of this combo -or- at the uncle point whichever comes sooner (in the present case the uncle point is -42k whilst the above is at -54k so it will be the uncle point ?)

Please Consider this: If your definition the failure of a systems enabled combo is @ -200%of the max relativized historical drawdown, then if the current drawdown lasts 200% of maximum historical drawdown duration for this combo, would that too be a failure point ? and if not why does the -200% rule apply re drawdown amount but not drawdown duration ?

Further, please consider this: Per your post replicated below, you say, contrary to my own assertion that actually you think that the greater probability of this prevailing systems enabled combo is that ' ...we will exceed the previous 37k peak, in less than three months...' but let us assume that it does not and the equity curve stays above the uncle point but below the last accounnt hi - just meandering around, this would then be greater than 400% of max historical drawdown duration for this combo - so at what durational point would you change it ? (given that in the long run we are all dead ! )

BBmac.

Replying as I read.

Correct. Reticence from me and even more so since the investors totally agree with me on the concept of non-intervention, despite the fact that their discretionary trading (on another account) is profitable.

I cannot have a lower tolerance because, as i said before, the systems always perform worse in the future (cfr. world cup example and limited accuracy in predicting the future based on the past), so that worse performance doesn't mean we've made a mistake. If anything, we should increase the tolerance, and accordingly the capital to sacrifice.

Yes, it has been well exceeded, and now to some extent I am involved in a 'hope trade'. I can hope all I want provided that I act in a disciplined way. It's disciplined trading, and then praying to god that things go right. There are no hope trades. Usual misinterpretation, and exaggeration.

No, I disagree. The probability is that we will exceed the previous 37k peak, in less than three months. Then when we'll scale up again, we will remember to compensate the silver systems with smaller futures. As i said before, enabling the big silver systems was a mistake at this point. Too much leverage. We will be able to compensate in the future.

Nothing has stopped working. The combo cannot stop working. The mistake can only be that we enabled a few unprofitable system, but this is not the case either. None of them has exceeded its previous drawdown. What happened is just that a whole bunch of systems lost money at the same time. That is why we're not even thinking about disabling anything.

You cannot have a situation where a whole combination is wrong. You can only have a situation where in a combo there's a great number of systems that should not have been selected, but even then, it will be a minority of the systems. This combo was not that great, but the majority of the systems are still very good, and none of them has failed yet. Yes, we could have chosen a more selective combo. Yes, we exceeded with silver.

Bring it on. As long as you have the patience, I will, too.

 

[Yamato]

Replying as I read.

Yes, it would be uncle point, but it would be wrong, if it will be so.

Yes, the combination would fail but that would only mean that I didn't choose systems wisely but not that they all failed, and it still would have to be established which ones failed.

The systems' equity line moves so fast that it will very quickly either rise above 37k or fall below -5k. If it doesn't do it within six months, we'll start changing some stuff.

 

[bbmac]

Could the uncle point come then not as a result of monetary drawdown but as a result of duration of drawdown, ie in the present example if no new account hi within the 6 months you mention below...might the investor (s) get fed up with lack of progress and walk ?

G/L

Replying as I read.

Yes, it would be uncle point, but it would be wrong, if it will be so.

Yes, the combination would fail but that would only mean that I didn't choose systems wisely but not that they all failed, and it still would have to be established which ones failed.

The systems' equity line moves so fast that it will very quickly either rise above 37k or fall below -5k. If it doesn't do it within six months, we'll start changing some stuff.

 

[Yamato]

Yes, of course. If nothing happened for six months, which is very unlikely (that's why we never even talked about it), I would be ok if people stopped trading my systems. We never even mentioned the possibility.

In the night we lost 7k. Now we're 1k away from uncle point. So now it's more likely than ever that we will stop trading due to reaching uncle point, rather than because of the systems going nowhere for six months.

Actually we had a period where they went nowhere for six months, actually more than 6 months. But since we did not reach uncle point we kept going. So I guess this answers your question. The "time" uncle point is more than 6 months, or at least it has been that way in the past. We waited for things to happen from June 2010 till December 2010.

 

[Yamato]

Today it might be my last day trading with these people. We've never been so low. Now we're only 1300 dollars away from giving up. I have been given instructions to stop trading in case we go below this level.

 

[bbmac]

Eurusd has hit a previous weekly swing hi/lo zone...may see some demand here - but daily/weekly on that instrument [that is directly and strongly correlated to the prevailing systems enabled combo] looks very bearish at the moment. '...It was the fed that did it..'

G/L

Today it might be my last day trading with these people. We've never been so low. Now we're only 1300 dollars away from giving up. I have been given instructions to stop trading in case we go below this level.

 

[megamuel]

How's it going? Still in the game?

Just wanted to ask you a quick question about out samples if you don't mind - Do you ever use more than one? I have done accidentally and my results weren't very good!

What happened was, I optimised from 2001-2006, then out of about 10,000 optimisation results I then backtested everything that made over £2000 (trading 0.1 lots) and had lower than 15% drawdown over 2006-2011. If again they made over £2000 with a drawdown lower than 15% on the out sample then I noted down the settings and these were my candidates to go live with. I then did a full back test to see which settings worked best on the in and out samples but I did it with all the data I have so from 1999-today. Now, on everything I have tested so far they systems do not make money from 1999-2001. Which I guess is another out sample. It's pretty disheartening because now it seems that I have just picked curve fitted settings even for the first out sample. I still have about 10 back tests to go but it is not looking hopeful.

So have you ever used 2 out samples to make sure you haven't picked settings that are curve fit to the first out sample? If this happened to you what would you do?

Thanks,

Sam.

 

[Yamato]

Replying as I read.

Yes, still in the game for another 650 dollars.

No, i never use more than one.

This using more than one out-of-sample described by you seems to me like peeking at the out-of-sample. Going live is not equivalent to a third out-of-sample, because you will not have as much time to test performance.

As i said before, the risk of curve fitting, if you use the out-of-sample, is minimal and it can be ignored.

 

[Yamato]

Ok, great recovery in the last few hours, so we will keep trading as usual up until tomorrow included.

There was a point where the ongoing trades got us to as low as 650 dollars from uncle point, but now we're again 2500 dollars from uncle point.

If it was just a drawdown, then we might recover. If it was that I enabled a bunch of unprofitable systems, then we have a problem, and sooner or later we will hit uncle point.

 

[Yamato]

It went ****ing well. We were down to 600 dollars away from uncle point and now we're more than 6000 dollars away from it. Really amazing day, with a recovery of 5500 dollars.

Now I am really starting to hope that we will make it.

If I have enabled a bunch of unprofitable systems, then we're screwed anyway. But if I did not and this is a regular drawdown, then soon it may end.

In a few days my vacation will be over. Luckily this mess happened during my vacation and we all kept calm. The investors kept calm because they always do. I kept calm because I was having the best time of my life.

If this had happened during my regular office routine, I might not have survived it, getting into endless arguments with the investors.

 

[Yamato]

Further recovery thanks to the overnight trades and now we're 8k from uncle point.

 

[Yamato]

Eurusd has hit a previous weekly swing hi/lo zone...may see some demand here - but daily/weekly on that instrument [that is directly and strongly correlated to the prevailing systems enabled combo] looks very bearish at the moment. '...It was the fed that did it..'

G/L

This is not what I keep saying, and, if you keep twisting, unintentionally, the truth, I need to keep reminding you that you're doing it:
"...directly and strongly correlated to the prevailing systems enabled combo".

1) We have no way of predicting where EUR and ES will go, because if we did, we should simply trade discretionally those two markets, and we would make money much more easily.

2) My systems are not "strongly and directly correlated", and I just said that if both ES and EUR are rising we tend to make more money than if they're both falling. This does not mean that we lose money if they're both falling, but only that we will make less money (which is irrelevant anyway, because of point #1).

Here's the chart of the relationship between my systems and ES and EUR:



This incident reminds me when I told you that almost all my systems made money in forward-testing, and you kept repeating over and over that "some of them did". In both cases you exaggerate their shortcomings, and therefore my shortcomings. It is almost as if you wanted to subtly and imperceptibly downplay my work. Maybe you don't even know it yourself, you're not even conscious about it.

 

[new_trader]

It went ****ing well. We were down to 600 dollars away from uncle point and now we're more than 6000 dollars away from it. Really amazing day, with a recovery of 5500 dollars.

Now I am really starting to hope that we will make it.

If I have enabled a bunch of unprofitable systems, then we're screwed anyway. But if I did not and this is a regular drawdown, then soon it may end.

In a few days my vacation will be over. Luckily this mess happened during my vacation and we all kept calm. The investors kept calm because they always do. I kept calm because I was having the best time of my life.

If this had happened during my regular office routine, I might not have survived it, getting into endless arguments with the investors.

I’m glad to see that things have turned around. The curious thing to me is your use of words like “hope” and “luckily” and saying “If this had happened during my regular office routine, I might not have survived it, getting into endless arguments with the investors”.

Just remind me again, you trade mechanical systems because you have trouble controlling your emotions, right? It seems to me as though it was your vacation that ‘saved’ you.

 

[Yamato]

Yeah, it saved me from a big fight with the investors, whereas instead, being on vacation, I was very easygoing and said: hey, if you want to stop, we can stop at any time. So we didn't start blaming each other for anyone's mistakes. I am very calm and rational right now, as if this huge fall in the equity curve did not matter to me. And right now, being on vacation, it really does not matter.

 

[megamuel]

Replying as I read.

Yes, still in the game for another 650 dollars.

No, i never use more than one.

This using more than one out-of-sample described by you seems to me like peeking at the out-of-sample. Going live is not equivalent to a third out-of-sample, because you will not have as much time to test performance.

As i said before, the risk of curve fitting, if you use the out-of-sample, is minimal and it can be ignored.

That would be pretty cool if you got as low as 650 dollars and then went back up to make a new equity high! Good luck!

Here is an example of my testing:

I had 35 candidates that performed well in both in sample and out sample. Yet when I tested them on the full range of data they all cane out looking like this. Red = In Sample, Green = Out sample, and Blue is the extra bit of data that I included in the complete backtests. In all of my tests the systems don't work in this blue area (1999-2001). Surely if the system 'works' then it should work there also. It seems that I have curve fit to both samples rather than finding something that actually works...

 

[Yamato]

Sorry, but we will need a chat session to clarify what you have done, because what you are saying leaves room for too many misinterpretations. I cannot reply based on the things I have understood (not much). It would be complex to keep on talking about this on a forum. Of course, I would still recommend that you use exactly the method I suggested to you, so anything different is not right. Having said this, I still do not know exactly what you have done.

 

[Yamato]

In the meanwhile... it happened! On my last day of vacation with my friends, it happened!

Yes, yes, yes!

There was a silver trade and it finally delivered what I have been expecting all these weeks. 11k of profit with just one trade. If you add this to the other trades, we're now 20k away from uncle point, 15k in the profit, and virtually safe from whatever ****ing accident will happen in the future. I think we can start thinking that the drawdown is finally over.

 

[bbmac]

Great news...that was some move on Silver/Gold today !

G/L