ask Dcifrths...well, anything...about finance/mkts/ports that is.

[DcifrThs]

[ QUOTE ]
1) You are a standup guy for disclosing the adversity of not fitting at that job...I am a big fan of candor and disclosing weaknesses.

[/ QUOTE ]

thanks. i appreciate your kind words.

[ QUOTE ]

2) You write this:
in terms of prediction, they are pretty worthless. but ironically, the use of the fundamental drivers that come from understanding what is trying to be conveyed by predictive models can be hugely useful in trading. especially where there are non-profit seeking participants.

the fed, though, does a pretty good job as far as managemnet goes and they use predictive models to think about this stuff, so that is aplus for "economists."


I am not sure I understand...to the extent that understanding the efforts of predictive models are useful in trading...doesn't it follow that they themselves need to be predictive? How can understanding what is being attempted by non-predictive economists lead to a predictive strategy?

SM

[/ QUOTE ]

there are a few points above butt he main one seems to be "how does using a non-predictive model help in trading since it doesn't do a good job of predicting?"

the reason imo is that the model may have some important factors that do get a relationship correct. but as it gets more and more complex, it starts to fall apart. depending on the precision you want your predictions to have, this can be fatal.

but if you want to dig into a specific relationship, say between X and Y. you have two models, one that predicts X and one that predicts Y. both models are rarely right.

but what happens when you want to trade X against Y? well that diff trade can be constructed by taking the things in X that also affect Y and the things in the Y model that also affect X. (this is easier said than done obviously).

what you then have, is a model that doesn't try to predict the absolute level or X nor Y but a model that can give you a signal as whether the difference between X and Y will increase, decrease, or possibly stay the same.

so understanding a non-predictive model can still be useful in trading as above.

some predictive models though do a pretty good job. the fed's models for instance seem to work rather well in terms of economic management, but not so well in predicting specific micro #s.

thanks again for your participation and kind words. i hope this helps convey the way i was thinking about the answer i provided above. let me know if it doesn't (things may land differently for you than how i intend them when i write).

thanks,
Barron

[DcifrThs]

[ QUOTE ]
If you thought you were the best (pure/skilled/technological/whatever) trader in the world with $10B AUM, but could only trade in one market, what would you trade?

[/ QUOTE ]

obviously the market im the best at trading. no "best in the world" trader is best at all markets.

Barron

[Pokerstuff2]

The most important things to know before going to work for a global long/sort fund?

[DcifrThs]

[ QUOTE ]
The most important things to know before going to work for a global long/sort fund?

[/ QUOTE ]

i don't know. never worked for a global long/short fund. basic financial calcuations i'd think you should know though. portfolio math and the like.

Barron

[edtost]

[ QUOTE ]
[ QUOTE ]
barron,

wrt managing a portfolio of assets, how do you view the relative importances of predicting returns, predicting covariances, and optimizing the portfolio? how is this influenced by the relative difficulties of each of these and the amount of error even doing them well imparts on the process?

thanks,
ed

[/ QUOTE ]

the answer varies if you are talking about a passive long only portfolio or an active managed portfolio.

in the latter, it depends on the leeway you have as a manger (can you short? use leverage? use derivatives?)

i'll take the passive portfolio since it is easier to explain.

the most important thing is thinking about how an asset class HAS VARIED in the past by itself (and why), how it HAS VARIED in the past against everything else in your portfolio (and why), and most importantly, how you logically think they can and will vary in the future both individually and together.

predicting returns is far easier in the asset class example. an easy way to go about it is to assume each asset class has a sharpe ratio of between 0.2 and 0.3 (which makes sense and is backed by data)...all asset classes except commodities that is (i'd give them a 0.10-0.15).

then you simply back out the "expected return" of that asset class from the volatility estimates you've worked hard to think about (if ever you're not sure, you can increase/decrease these estimates & see the effect on your expected portfolio #s).

from there, you can optimize your portfolio. most people do this with a simple mean-variance optimization (i.e. run the calcs one time). i think this is a big mistake since the optimal allocations are highly sensitive to small changes in correlations & variances. therefore, you'd want to do a 10k run monte carlo with ranges around each of those factors (and since the "expected return" is a function of "expected volatility" that will back itself out). the ranges you can tweak as well to be wider around assumptions you are not as sure about.

in this way, you can try to minimize big and common mistakes in constructing an optimal portfolio.

this is FAR more complicated if you're talking about an actively managed portfolio.

hope this helps. did i asnwer your question? or did you want me to give an opinion as to which part of this is the hardest?? imo that isn't all that important b/c all of it needs to be done right .

Barron

[/ QUOTE ]

you came pretty close to what i was looking for ... i was mostly just looking for what your philosophy was wrt portfolio management in general.

to make the example a bit more complicated though, say that instead of allocating between asset classes (where looking at returns/covariances is somewhat trivial), you're managing something like stock picking or an arb strategy that may be going away, where returns and covs are likely to change from historical, and you may be subject to highly non-normal return distributions and "black swan" type unpredictable events.

knowing that everything needs to be done right, but doing any of them right will take all the time/effort you have and then some, and still may not have a definitive answer, on what step of the process would you focus your energy?

[chisness]

Barron,

I'm in college and have read that interviews will often ask about particular stocks or how the economy is "doing" in general. The former seems pretty easy to research and understand, but how do you concisely explain the state of the entire economy? Pick a few major recent events and their effects? Talk about market direction and interest rates?

[APXG]

Barron,

Where do you feel "knowledge accumulation" is more efficient - entry level at hedge fund or reading everything you can get your hands on at home + occasionally seeking out smart people to discuss / learn? I ask this b.c. I am beginning to see a parallel with the entry level job as a more advanced+focused extension of college, which is a highly inferior learning mechanism b.c. there are too many stupid demands and distractions for optimal learning(i.e. GPA, people who love to contagiously waste time, and classes on English literature).

Assume you have financial flexibility from poker that the pay from any job is irrelevant, as well as enough $$$ to start a small personal fund when the "knowledge accumulation" has hit a desired level -- and then make back any losses through poker. The stop would be pretty tight the first time [img]/images/graemlins/wink.gif[/img]

[Groty]

[ QUOTE ]
Here is a n00b question that maybe you can answer.

As I understand it, a company goes public to raise capital.
Now let's say I buy 10% of the share so I'm basically owning 10% of that company. So how is it that some years after the IPO, the company can decide to "release more shares" to raise more capital. Surely that's not fair to the current shareholders because once they do that I own less than 10% of the company.

[/ QUOTE ]

The certificate of incorporation filed with the secretary of state in the state of incorporation lists the maximum number of shares the corporation is authorized to issue. (Any balance sheet of a publicly traded firm will list the number of shares authorized, the number issued, and the number outstanding at the balance sheet date.)

The board of directors has the power to vote to issue more shares up to the authorized number without first obtaining shareholder approval. However, to increase the authorized number of shares requires shareholder approval (as defined in the certificate of incorporation).

[DcifrThs]

[ QUOTE ]
[ QUOTE ]
[ QUOTE ]
barron,

wrt managing a portfolio of assets, how do you view the relative importances of predicting returns, predicting covariances, and optimizing the portfolio? how is this influenced by the relative difficulties of each of these and the amount of error even doing them well imparts on the process?

thanks,
ed

[/ QUOTE ]

the answer varies if you are talking about a passive long only portfolio or an active managed portfolio.

in the latter, it depends on the leeway you have as a manger (can you short? use leverage? use derivatives?)

i'll take the passive portfolio since it is easier to explain.

the most important thing is thinking about how an asset class HAS VARIED in the past by itself (and why), how it HAS VARIED in the past against everything else in your portfolio (and why), and most importantly, how you logically think they can and will vary in the future both individually and together.

predicting returns is far easier in the asset class example. an easy way to go about it is to assume each asset class has a sharpe ratio of between 0.2 and 0.3 (which makes sense and is backed by data)...all asset classes except commodities that is (i'd give them a 0.10-0.15).

then you simply back out the "expected return" of that asset class from the volatility estimates you've worked hard to think about (if ever you're not sure, you can increase/decrease these estimates & see the effect on your expected portfolio #s).

from there, you can optimize your portfolio. most people do this with a simple mean-variance optimization (i.e. run the calcs one time). i think this is a big mistake since the optimal allocations are highly sensitive to small changes in correlations & variances. therefore, you'd want to do a 10k run monte carlo with ranges around each of those factors (and since the "expected return" is a function of "expected volatility" that will back itself out). the ranges you can tweak as well to be wider around assumptions you are not as sure about.

in this way, you can try to minimize big and common mistakes in constructing an optimal portfolio.

this is FAR more complicated if you're talking about an actively managed portfolio.

hope this helps. did i asnwer your question? or did you want me to give an opinion as to which part of this is the hardest?? imo that isn't all that important b/c all of it needs to be done right .

Barron

[/ QUOTE ]

you came pretty close to what i was looking for ... i was mostly just looking for what your philosophy was wrt portfolio management in general.

to make the example a bit more complicated though, say that instead of allocating between asset classes (where looking at returns/covariances is somewhat trivial), you're managing something like stock picking or an arb strategy that may be going away, where returns and covs are likely to change from historical, and you may be subject to highly non-normal return distributions and "black swan" type unpredictable events.

knowing that everything needs to be done right, but doing any of them right will take all the time/effort you have and then some, and still may not have a definitive answer, on what step of the process would you focus your energy?

[/ QUOTE ]

ok so lets just be explicit here.

we are now talking about running an actively managed portfolio in one asset class, lets take an M&A Arb strategy in equities.

keep in mind that we are now branching into a specific area in which i am not educated.

in order to construct an optimal portfolio we need the following:

1) variance estimates & ranges around them: st.dev
2) covariance estimates and ranges aroudn them: correlations
3) expected return estimates and ranges around them.
4) as large a sample of profitable trades one can find.

the latter is fairly abundant in today's climate. the main problem is constructing the optimal capital allocation to these trades (while at the same time minimizing transaction costs). lets assume we are dealing with only liquid securities (not a small assumption for many big managers).

in this case, first and foremost i'd concentrate on getting the trades right. that is #4. the more trades you have, the better off you are. for M&A arb, that is probably pretty hard considering that a sudden rise of risk aversion will drive correlations of deals not getting done up. if you're betting some will get done and some won't then you're better off (again, im no expert here. i've never worked for an M&A arb fund nor do i know specifics about how they tackle this). so once you have those bets i'd do the following.

just throw some base st.dev, corr, & exp. ret #s in there and create a monte carlo'd portfolio of these trades.

then, see where the biggest risk gets allocated (keep in mind you need to think about RISK allocation and work backwards to capital allocation). then i'd want to see what would happen to my portfolio if the worst thing that could ever happen in this bet happened (i.e. if i was betting on a convergence in share prices lets say one went bankrupt and the other hit the jackpot). could i blow up? if yes, take some risk away.

do the same down the line with all trades.

we get to see how difficult this is especially since we just started with some base #s. the main thing i'd want to make sure i could do is weather the worst and still have capital to make it to the best. that is the key with alpha strategies that are concentrated and highly susceptible to black swan type events imo.

you can never plan on everything.

could the top 3 fund i worked for blow up? $30b down the drain? sure. if all correlations of their 150+ trade types went to 1 and they all went precisely against them (some would have to move up , some would have to move down since most of the portfolio is in diff trades not directional trades). the goal is to make sure that no one bucket or area can hurt you disproportionately. you want to be able to gain the best of it while weathering the worst of it.

specifically, if you have a very high conviction bet (high expected return and to you, a low st.dev) and it moves against you without changing the fundamentals/flows etc. you want to be able to put even more into that trade or at LEAST not have to pare back your position.

in most cases, i think people allocate too much to those 5-15 killer trades than they should b/c if they moved against you you'll be in a position of reducing risk exposure not putting more on.

i know i didn't give you a clear answer, but it isn't something that is clear. you need to think about everything but the MOST IMPORTANT thing to think about is what could cause you to blow up or have to pare back positions given some basic assumptions. the accuracy of those assumptions aren't as important as making sure your portfolio of alpha trades can stand up to some big moves against it and still be in a position to profit from your views.

and you DEFINITELY need to be thinking about any big move that happens while you're managing this portfolio. if your trade moves 1.5st.devs against you, are you still sure you have the fundamentals right? did you miss something? SHOULD you pare it back b/c you've missed something?

LTCM doubled down when they *should have* realized that all they were doing is buying the riskier asset while shorting out the risk-free asset...in every market. in every conceivable way. then they got into areas in which they didn't have an expertise, or an edge, where big moves would KILL their portfolio.

not once did they go through the above practice whole heartedly. it was a failure of risk management above all else. i know you can critique my analysis here b/c hindsight is 20/20...but they never did really get into that type of risk management assessment of their bets. that one thing could have helped them weather the storm.

you dont want that type of thing to ever happen to an active portfolio.

hope this helps. can somebody who actually manages an active portfolio liek the one above (m&a arb, or equity type long/short) chime in? am i missing something?

thanks,
Barron

[edtost]

thanks. i was looking more for your thought process than a clear answer, so that was perfect.