Traders' Roundtable

It could well be that i am merely getting hung up on semantics and that this goes nowhere, but let me lay this out anyway....

In a related thread I caught myself making a statement to the effect of:
Less Things(i.e. timeseries,products,strats etc) allow for greater potential performance at the expense of stability, More Things allow for greater stability at the expense of greater mediocrity.
i.e the benefits and costs of diversification.

But then i thought that this could get shotdown by proponents of "Large Basket, Small Subset" (e.g. Sluggo) who could claim that by exposing themselves to a large number of products(beyond their means to finance perhaps), but then limiting actual trading to a Dynamic Subset Portfolio(DSP tm), they are able to materially improve potential performance through diversification, contrary to my assertion.

This got me thinking: is this DSP really "diversification" as I and perhaps others widely recognize the term, or is it something of a derivative, or perhaps altogether different phenomenon?

For example ,assuming for simplicity we have a single trade strategy and a Global Basket of 300 products, but at any one time was limited to 20 positions, then you could arbitrarily assign each current position to a "equity curve slot" within the DSP.
Whilst it is true that over time the constituent of each of these 20 Equity Curves will vary among products or cash as trades come and go, you still only have 20 equity curves at any one time, and therefore actual diversification of 20.
This is of course different from the Global population portfolio which would be all 300 product equity curves all of the time, and which would have diversification of 300. Being the population, this would represent average performance[lets ignore weightings issues for the time being please].

It is my conjecture that any outperformance achieved by the DSP is therefore the result of:
1) positive selection bias when new trades replace old trades on a first come first served basis(or any other basis for that matter).
2) Opportunity Benefit. More potential products within a population give rise to more frequent opportunities for positive selection bias to occur.

Note that expanding the global Basket population does not directly imply that potential outperformance will be increased(though it is likely, especially when starting out in expansion).
e.g. you could increase the global population from 300 to 400, but if the new products are all universally total dogs FOR YOUR GIVEN STRATEGY, then it is natural that they would crowd out some of your existing positive selection bias and reduce your outperformance.

To really blow your mind, imagine multiple, or godforbid, all possible trade strategies, applied to all possible products. you would then have the total universe population of individual equity curves. now pick 20 and rotate on a firstcome firstserved basis. I almost had to reboot my brain when i started thinking about how one would actually implement such a Demon!


Anyway. If we were to conclude that the DSP effects are not "Diversification" by the dictionary definition or whatever, what would we call it? am i missing something obvious?

Positive Selection Bias of a Dynamic Subset Portfolio through Opportunity Benefit Improvement

...is a bit of a mouthful!

Check out the Parking Lot block in the Blox Marketplace (link), which I think was originally inspired by Dean Hoffman's product called "Relativity". There are a couple threads about Relativity as well.

Also there's a fun thread called Number of Parameters which you may enjoy, including

isn't it true that the decision to simulate system X includes within it, the decision not to simulate systems Y, Z, A, B, C, ...? Therefore aren't those "hidden parameters"? Every system you've ever heard about, is a parameter of the system you're working with now, right? After all you have the freedom to include or exclude those other systems as you please; these are "degrees of freedom," and that makes them parameters. Is it not so?

In the case of the Bollinger Breakout, someone decided to give the indicator named Bollinger Bands a 100% weighting, the indicator named ADX a 0% weighting, the indicator named RSI a 0% weighting, the indicator named CCI a 0% weighting, the indicator named ROC a 0% weighting, and the indicator named RegressionSlope a 0% weighting. Aren't those weights "hidden parameters"? Tradestation contains 200 indicators and textbooks discuss even more. You / John Bollinger / the developer of your system had the degrees of freedom to put in, or leave out, each of these indicators. Therefore they are hidden parameters, as plain as day.

So each and every system, no matter how simple it may appear, actually contains more than a thousand hidden parameters, reflecting not only what is in the system, but also what was deliberately left out of the system by conscious choice.

Ugh.

Edit - after several search failures, found and added quote above

Yeah i was aware of DHP. Relativity (capital R) is already used as a common scientific term for something much more impressive, no need to confuse things. I didn't mention it because "parking lot" doesn't appeal to my aesthetics of nomenclature, yeah , typical Brit huh!. Besides it is just the tool by which this process plays out.
I seem to recall some people were even irritated that Mr Hoffman appeared to present this concept as something special and proprietary to his systems when many others were already doing it, or something to that effect. Either way , no discredit to Dean for espousing it in his approach. I just think the surface is only being scratched so far. What we are talking about is something which apparently adds "alpha" , and yet is not a trading rule as conventionally applied to a product, but some form of meta-trading rule. It also can smooth returns, if only by the fact that it is selecting against poor ones, but it is doing this without requiring additional capital commitment, unlike standard diversification. It may be a form of "holy grail" when sensibly applied, but it may also have many cans of worms hidden in its cupboards if we apply it naively. If we overcomplexify it( instead of using First-come first-serve, we use an alternative positive selector) do we on balance run greater risks of screwing up.

Either way it deserves to be elevated in status and differentiated from standard diversification. It isn't a product for sale either. It is a statistical effect of an underlying selective process. It isn't the same thing as smoothing by diversification, and leads to confusion if it is not differentiated, in my experience.
(e.g. "how the F*** is Sluggo getting that level of performance in that test by trading everything all the time? he can't be"... [penny drops]... "oh, wait, he isn't doing it the way i presumed")

pure and simple "quantity" diversification: rainbow colours all blending together to make blurrghh. depressing.

DSP/parking lot effect: a rainbow which changes it's colours. ooh shiney!

hey, even if Dynamic subset portfolio and parking lot are the best we can come up with as names so far, it would still be nice to discuss this topic further in this thread.

I'm wondering
1)how it can be isolated and measured in terms of it's relative contribution to CAGR or whatever, and what are the best ways of expressing that?
2) how do we control for order of population expansion and underlying trade rules?
3)when does this effect not manifest a positive result and how do we anticipate it going wrong?
4)In what way do we control the systematic risks of this approach in conjunction with it's underlying strategy(overtrading and picking dogs does worse that just a small static portfolio).
5)is it too much of a nightmare to run sector specific strategies/parameters within a DSP(e.g. different parameter choices for lengths yield different rates of opportunity)
6)etc.

that's it for me for now. brain needs rest. been a busy day on this forum for me!

Thanks for the update sluggo. i did like that thread when i read it at the time. I have read pretty much every thread here, just, there is so much good content, i forget where everything was or if it was there, and the search function is not always helpful.

hmm, this thread was right under my nose. good ole klatt!