What Happens Next

[spinola]

Traders Matching Principle - Approach To Markets

If you believe the market is random, then you shouldn't be using a series to muddy the approach at all.

Fix a stop and limit (to be determined by trader) each day, and flip. That is your trade that day.

Place it once you get your trade reason. Job done. Place next trade when that one concludes.

Trader is now trading with a correct matching principle of his perceived belief of random market using the correct random method approach to trading random markets. Which , after all , is what the trader believes.

This is called the ' Matching Principle '


Trader will be happy because he is using an approach based on his character and perception of a market. If he attempts this whilst using a mismatching principle of an approach counter to his own random perception beliefs, this will only create internal conflict as the 2 principles of mind and market are at polar opposites.

The trading gods will not be pleased, and neither will you. Why would or should you be happy? If there is , in actuality, in YOU , in any trading individual, a mismatch ?

Hence I would say first step is for the trade to sort out his own matching principle first.

And without further ado ,l ets roll .

If the link below is a random coin flip.

One side says

The Other

http://www.random.org/coins/?num=1&cur=20-novelty.decision-maker


I might scatter some humour in the above, but it's genuinely what I feel is right or passes the common sense test.


Big fat query all the above of course, just some thoughts.

But again, I would say first step is for the trade to sort out his own matching principle first. That's part of the challenge for anyone, all of us. And it's unique to each individual.

You will be free then to master and evolve your experience. You will know which path to take.

Hence I would say first step is for the trade to sort out his own matching principle first.

We each stand at the fork in the road .Random Road forks left, methodical road forks right.

 

[scholfield]

except those spanish stops nutters

He's getting a bit of a following on twitter, I see.
Pretending that he's an Ex J.P.Morgan trader now. Oh.....and he's now a hot girl, too! lol
https://twitter.com/Big_Bum_Babe


I hear that the market is less random than academia believes and more random than traders believe.

 

[Liquid validity]

If you believe the market is random, then you shouldn't be using a series to muddy the approach at all.

Fix a stop and limit (to be determined by trader) each day, and flip. That is your trade that day.

Place it once you get your trade reason. Job done. Place next trade when that one concludes.

Trader is now trading with a correct matching principle of his perceived belief of random market using the correct random method approach to trading random markets. Which , after all , is what the trader believes.

This is called the ' Matching Principle '


Trader will be happy because he is using an approach based on his character and perception of a market. If he attempts this whilst using a mismatching principle of an approach counter to his own random perception beliefs, this will only create internal conflict as the 2 principles of mind and market are at polar opposites.

The trading gods will not be pleased, and neither will you. Why would or should you be happy? If there is , in actuality, in YOU , in any trading individual, a mismatch ?

Hence I would say first step is for the trade to sort out his own matching principle first.

And without further ado ,l ets roll .

If the link below is a random coin flip.

One side says

The Other

http://www.random.org/coins/?num=1&cur=20-novelty.decision-maker


I might scatter some humour in the above, but it's genuinely what I feel is right or passes the common sense test.


Big fat query all the above of course, just some thoughts.

But again, I would say first step is for the trade to sort out his own matching principle first. That's part of the challenge for anyone, all of us. And it's unique to each individual.

You will be free then to master and evolve your experience. You will know which path to take.

Hence I would say first step is for the trade to sort out his own matching principle first.

We each stand at the fork in the road .Random Road forks left, methodical road forks right.

Yeah I spose I would agree with that.
Both are possible, random is not as efficient, but does work.
That is comparing to someone who can identify the times when
market randomness is highly reduced or almost negligible, and sit on their hands
when that isn't the case.

About the only thing I would possibly contend is random trading with
fixed stops and limits.
Maybe its possible, but I personally have not found any value there.
Dynamic stops and targets - mainly running winners, I found to be better.

As I say though, just because I didn't personally find any value in fixed stops
and targets hardly means its definitive.
It can be instrument dependent, entry method dependent and so on.

 

[Liquid validity]

He's getting a bit of a following on twitter, I see.
Pretending that he's an Ex J.P.Morgan trader now. Oh.....and he's now a hot girl, too! lol
https://twitter.com/Big_Bum_Babe


I hear that the market is less random than academia believes and more random than traders believe.

Gotta love spanish, top class tard

As for random, thats the biggest problem, it isn't empirical.
The presence and mix of it constantly changes each day.
No one truly has that cracked.
As a rough example - min bid rate / NFP / FOMC - almost always
quiet and ranging before, and explosive to some degree after.
That is not random, although you do have to contend with increased volatility.
There are certainly days and times of day where randomness is highly reduced.

A good human judgement call, or HFT speed increases the odds of negating the effect,
simply by identifying points in the market that are anything but random,
whereas HFT removes the random element not so much with trade accuracy,
they just counter it with their speed edge.

 

[random12345]

Yeah I spose I would agree with that.
Both are possible, random is not as efficient, but does work.
That is comparing to someone who can identify the times when
market randomness is highly reduced or almost negligible, and sit on their hands
when that isn't the case.

About the only thing I would possibly contend is random trading with
fixed stops and limits.
Maybe its possible, but I personally have not found any value there.
Dynamic stops and targets - mainly running winners, I found to be better.

As I say though, just because I didn't personally find any value in fixed stops
and targets hardly means its definitive.
It can be instrument dependent, entry method dependent and so on.

Agree - William Eckhardt contended many years ago that degrees of freedom were all important in his systems running alongside Dennis' discretionary mob and that any more than 5 parameters of input were excessive - though these were the days when trend days were more common than ranging days, the point still stands. This is why trailing stops tend to have better implied future value on a curve fit system than profit targets.

I'm not sure about dynamic stop losses myself though - I tend to prefer a static instrument based stop (sampled to relatively pertinent volatility) as once you get into say extreme trend points and the like (ATR or whatever) you can find yourself with enormous stops below or above the spike of the last move, though forex is perhaps better for this because you will see value buying or selling by institutions with stops in that area acting as pretty decent stop defence, though a news move might slash you up in both directions, especially if you're always in.

 

[random12345]

Agree - William Eckhardt contended many years ago that degrees of freedom were all important in his systems running alongside Dennis' discretionary mob and that any more than 5 parameters of input were excessive - though these were the days when trend days were more common than ranging days, the point still stands. This is why trailing stops tend to have better implied future value on a curve fit system than profit targets.

I'm not sure about dynamic stop losses myself though - I tend to prefer a static instrument based stop (sampled to relatively pertinent volatility) as once you get into say extreme trend points and the like (ATR or whatever) you can find yourself with enormous stops below or above the spike of the last move, though forex is perhaps better for this because you will see value buying or selling by institutions with stops in that area acting as pretty decent stop defence, though a news move might slash you up in both directions, especially if you're always in.

Also LV, you'll know that targets missed by just the spread alone are actually quite commonplace, something that is much less of a bottom line problem with trailing stops.

 

[Purple Brain]

So if gbp/jpy were to make 155.87 before it made 160.01 (currently 157.71)within the next x hours - would that be totally random?

 

[Triggerfish]

Is the OP suggesting his random chart is a true representation of the markets?

.....Despite that I would still say that markets themselves are not random.....

The random element mainly applies to trade outcome, no one ever knows in advance which trade will be a loss, scratch or win.
If you did, that would by definition imply a 100% strike rate, which no one has.....

If ones view of the markets are not random then is it not then possible that it can be known, if it is not known, then surely not random cannot be applied? It would only be an assumption?
Thoughts?

 

[random12345]

So if gbp/jpy were to make 155.87 before it made 160.01 (currently 157.71)within the next x hours - would that be totally random?

The underlying price represents the collective 'opinion' of the market participants, which obviously isn't random.

The movement in-between such prices is "random" enough to stop the majority of non random approaches from working, as would be necessary for any one group of speculators to generally profit while others generally lose.

For example, if you wanted to be purely qualitative, you could choose just to trade the major news events where a change is expected or any result off forecast would cause a move for the rest of the year, with a stop order bracket - you can't do this because you will be pushed into both positions by the spread before the release if OTC, or spikes if futures or even potentially miss the majority of your ave fill by thin liquidity. The mechanics are just random enough. You don't have to use the word random - whatever you prefer. Some people are offended by the word - just use difficulty instead.

There is always some function of the market that makes obvious money making approaches difficult and randomness is used to describe the collective nature of these functions I believe.

 

[Liquid validity]

Agree - William Eckhardt contended many years ago that degrees of freedom were all important in his systems running alongside Dennis' discretionary mob and that any more than 5 parameters of input were excessive - though these were the days when trend days were more common than ranging days, the point still stands. This is why trailing stops tend to have better implied future value on a curve fit system than profit targets.

I'm not sure about dynamic stop losses myself though - I tend to prefer a static instrument based stop (sampled to relatively pertinent volatility) as once you get into say extreme trend points and the like (ATR or whatever) you can find yourself with enormous stops below or above the spike of the last move, though forex is perhaps better for this because you will see value buying or selling by institutions with stops in that area acting as pretty decent stop defence, though a news move might slash you up in both directions, especially if you're always in.

I'm only using dynamic stop as a very loose term, I agree a maximum hard stop is
essential.
An example would be tiered trailing stops with varying step frequency
on each stage, with a tighter hard stop for instance.
Thats mainly what I was referring to in terms of dynamic stop.

I just didn't want to say that in case anyone took it literally, as the most
basic forms of trailing stop implementation can be pretty basic - although it still
works.

Any stop exit I get would be tighter than the max hard stop.
Agree on spikes and ATR lag - thats why I don't use them.

Agree on Eckhardt too, one of my key inspirations.

 

[Liquid validity]

Also LV, you'll know that targets missed by just the spread alone are actually quite commonplace, something that is much less of a bottom line problem with trailing stops.

Agree, virtually every time I've tried with targets, I've just found it to be an
upside limitation and nothing more in terms of equity curve.
Even a trail / target combo has the same effect.

 

[random12345]

I'm only using dynamic stop as a very loose term, I agree a maximum hard stop is
essential.
An example would be tiered trailing stops with varying step frequency
on each stage, with a tighter hard stop for instance.
Thats mainly what I was referring to in terms of dynamic stop.

I just didn't want to say that in case anyone took it literally, as the most
basic forms of trailing stop implementation can be pretty basic - although it still
works.

Any stop exit I get would be tighter than the max hard stop.
Agree on spikes and ATR lag - thats why I don't use them.

Agree on Eckhardt too, one of my key inspirations.

Ja - the problem with modelling trailing stops realistically with limited CPU power is an issue mind you, especially if you tighten them as the floor improves. I tend to find the reality is somewhere between tick magnified and OHLC as your broker won't receive or be able to facilitate the perfect tick by tick stop update as magnified would assume and therefore won't stop out positions as regularly since the stop will lag, but nor will a move grab the whole bar as OHLC can misinterpret.

Without a rack of Xeon Phis though I don't know how this could be better modelled, reducing tick frequency to replicate 'seen' ticks has the same problem with news bars as OHLC does.

 

[Purple Brain]

My point in the gbp/jpy example is that the upper level is 230 pips away and the lower level is 184 pips away. Levels chosen purely at random - honest.

But if anyone suggests the closer target is more likely they have brought probability into the picture. If truly random, both levels have equal probability of being hit.

If anyone suggests the price has visited both levels in the past, but most recent action has been below the lower level therefore the lower level more likely to be hit because of affinity or whatever else you want to call it, that's technical analysis.

If anyone suggests that it made a lower high on the daily Oct 22 therefore probabilities favour a downward move and therefore the lower level, that's probabilities and technical analysis.

My point - laboured I know - is that random takes a lot of effort out of the trading process - all of the effort in fact, but it becomes a binary punt - at best.

My hat is still very much in the ring of the statisticians and technical analysis guys - regardless of which level gets hit first - I get a distinct intuition that the analysts will win out over the long run with the randomistas.

 

[scholfield]

Re. that chart. I guess there's still differences.
Randomly generated data may contain patterns and runs etc, but that does not mean that price data is random despite taking on similar characteristics. A random set of data is normally distributed. Price data is not normally distributed, it has fat tails.

 

[barjon]

Hi guys - back from golf early (rained out halfway)

I was not intending to say that markets are random - you can't say that because a randomly generated chart looks like a market chart, therefore the market must be random.

The point was that TA patterns are not necessarily indicative of market progression since those same patterns can occur in randomly generated movement. It follows that TA should be seen in that context and not as the be all and end all. In my view it mere serves to point up places where you are prepared to make an assumption about future direction. It's what happens next that counts (limiting loss, taking adequate profit, finding momentum etc) and not the TA.

 

[Liquid validity]

Ja - the problem with modelling trailing stops realistically with limited CPU power is an issue mind you, especially if you tighten them as the floor improves. I tend to find the reality is somewhere between tick magnified and OHLC as your broker won't receive or be able to facilitate the perfect tick by tick stop update as magnified would assume and therefore won't stop out positions as regularly since the stop will lag, but nor will a move grab the whole bar as OHLC can misinterpret.

Without a rack of Xeon Phis though I don't know how this could be better modelled, reducing tick frequency to replicate 'seen' ticks has the same problem with news bars as OHLC does.

I know what you mean, but thats deeper than I have looked at it personally.
Maybe its not an issue for me as trade freq per day is 0-2, so stops are wider anyway.
I have to be honest I haven't used OHLC calc on bar close as trail trigger,
I just use native platform tick based stops, but then I have no choice as I don't use the API, so mine are not server side with the broker - which is a weakness,
although its rarely a problem.

I take it by using the API you have tick based trails server side, or have I
misunderstood?

 

[random12345]

I know what you mean, but thats deeper than I have looked at it personally.
Maybe its not an issue for me as trade freq per day is 0-2, so stops are wider anyway.
I have to be honest I haven't used OHLC calc on bar close as trail trigger,
I just use native platform tick based stops, but then I have no choice as I don't use the API, so mine are not server side with the broker - which is a weakness,
although its rarely a problem.

I take it by using the API you have tick based trails server side, or have I
misunderstood?

I've always found that broker historical ticks are bs both in terms of their own tick time-stamping and ticks received outside of Equinix co-location (in LMAX's case) so I don't know the answer re what's best to use. I don't think API trails are emulated at broker level as isn't that a broker determined exit? I don't think they have high level functionality on the matching/cluster engine?

To be honest I model multi floors with OHLC using historical live perf to adjust expected ROA so long as the trail percentage is realistic (i.e. when invoked is not just a few pts off the spread difference). I once had an amazing week when I accidentally used 5% trails on a low floor after coming back on Sunday eve in a drunken stupour - because of the stop would be trigger rejections until a viable exit, I effectively rode the price up with my own mini unsophisticated HFT. Never tried it again though!

 

[Liquid validity]

I've always found that broker historical ticks are bs both in terms of their own tick time-stamping and ticks received outside of Equinix co-location (in LMAX's case) so I don't know the answer re what's best to use. I don't think API trails are emulated at broker level as isn't that a broker determined exit? I don't think they have high level functionality on the matching/cluster engine?

To be honest I model multi floors with OHLC using historical live perf to adjust expected ROA so long as the trail percentage is realistic (i.e. when invoked is not just a few pts off the spread difference). I once had an amazing week when I accidentally used 5% trails on a low floor after coming back on Sunday eve in a drunken stupour - because of the stop would be trigger rejections until a viable exit, I effectively rode the price up with my own mini unsophisticated HFT. Never tried it again though!

Yeah server side OCO and trails do seem to be mostly native.
I'm guessing its purely to limit CPU load their end, not a technical impossibility.

As for the tick issue, agree.
You can see the difference between LMAX and MBT on those 10 sec charts
I posted on walters thread, at tick level the variation would be even more
obvious.

Its never been an issue for me as I set out to strictly limit intra day frequency
right from the start, so the tick variance is not a major factor.
Although I have no doubt some trades have been stopped with MBT ticks
that would not have been with LMAX and vice versa.
I doubt its a massive difference but I can't say what the true impact has been.
Obviously with tighter stops and increased frequency that effect increases.

 

[new_trader]

Hi guys - back from golf early (rained out halfway)

I was not intending to say that markets are random - you can't say that because a randomly generated chart looks like a market chart, therefore the market must be random.

Who would say that anyway? People see recognizable shapes in cloud formations like faces, cars etc. However there are trillions of cloud formations that don't look like anything. Pi as far as we know is an irrational number with an infinite number of decimal places. The digits are randomly distributed yet you will find the year you were born in there, in the proper sequence, and even your phone number. When you have infinite the chances of finding something recognizable in there is high. Scientists and astronomers postulate that since there are only a finite way in which atoms can be arranged, in an infinite universe there will be another planet exactly like earth with someone exactly like you living on it! :-0

 

[random12345]

Yeah server side OCO and trails do seem to be mostly native.
I'm guessing its purely to limit CPU load their end, not a technical impossibility.

As for the tick issue, agree.
You can see the difference between LMAX and MBT on those 10 sec charts
I posted on walters thread, at tick level the variation would be even more
obvious.

Its never been an issue for me as I set out to strictly limit intra day frequency
right from the start, so the tick variance is not a major factor.
Although I have no doubt some trades have been stopped with MBT ticks
that would not have been with LMAX and vice versa.
I doubt its a massive difference but I can't say what the true impact has been.
Obviously with tighter stops and increased frequency that effect increases.

I get you.

In calling them outright BS, I meant in terms of the frequency seen by the matching engine vs those dispatched to clients by api.lmaxtrader.com - now I have such a low latency solution I can still see a big difference, the engine must see twice as many than their system can cope with, which is an issue with acceleration methods that I have raised with them. You basically have to collect your own 2 years of ticks with whatever relevant connection you're using in order to have a pertinent tick sample - yikes.