View Single Post
  #6  
Old 10-10-2007, 10:11 PM
DcifrThs DcifrThs is offline
Senior Member
 
Join Date: Aug 2003
Location: Spewin them chips
Posts: 10,115
Default Re: Trade Idea Generation. the process of writing and thinking..LETS D

[ QUOTE ]
I'm afraid I don't have anything to add as don't tend to consider the fundamentals when looking for trades.

I'm just wondering how this fits into your ideas of efficient markets. I don't understand how you can find trades if the markets are efficient.

I'm also interested in how you manage these positions if you were to take them. I know you're interested in the research/analysis side of things, but if you were given trade ideas to execute, how would you manage your risk and positions?

[/ QUOTE ]

first point: i don't think markets are sufficiently efficient and i've divered my attention from my discussion w/ phonebooth to this. when ig et more time i'll go back and continue that discussion.

second point: i take one thing at a time. lets say i condense all that down into 3 trades:

1) maltese lira vs. euro
2) short US 2yr long US10yr (yield curve flattens)
3) long JPY/USD

i'd imagine that my entire portfolio was allocated to these 3 positions and plug in some vol & correlation #s. i'd expect 1 to be uncorrelated with 2 and 3 but i'd think that there may be some correlation between the yen strengthening and the yield curve flattening. namely, if the US 2yr yield falls (long the 2 yr), then the Yen should strengthen so if i get that leg of the trade right, the yen should strengthen vs. the dollar since interest rate diffs accound for the plurality of currency returns. so i'd be conservative and plug in a 50% correlation.

then i'd look at the volatility of these positions over time (or simlar positions in the case of maltese lira vs. euro). currently i don't have the data to dig into this but i'm just taking you through how i'd go about it.

from there i'd assign a maximum % of the portfolio that each bet could take on assuming it has a 100% absolute value signal. that takes into account correlations and volatilities. this is done so that i can hit a target dollar volatility very easily relatively speaking by adjusting the overall capital in the portfolio.

then i'd look at my signal. how confident am i in each of these bets. if 1=50%, 2=20%, 3=80% i'd want to allocate my portfolio accordingly.

then you simply divide the capital according to the max allocation and multiply by the signal you have. so if i had $100 and if each bet was at a max signal it would be:

1) takes up 40%
2) takes up 30%
3) takes up 30%

and my signals were:

1) 50%
2) 20%
3) 80%

then my capital allocated to each position would be:

1) $20
2) $6
3) $24

in reality that is much trickier because for each bet you have to know exactly what you're taking on...and there are costs to taking on these bets... and some are more costly than others, AND most annoyingly, the market doesn't give you a $6 jpy/usd futures contract lol.

further, the short 2yr/long 10yr bet can't just be put on as 1 unit short the 2yr 1 unit short the 10yr. you have to adjust for duration. if the 2yr has about a 1.5 year duration and 10yr has a 7.5 year duration then your position must be long 7.5/1.5 units of the 2yr for every 1 short unit of the 10yr you take on. otherwise, you'd have an odd bet that doesn't match what you want to take on in terms of yield curve exposure. if you just did 1 unit short 2 yr 1 unit long 10yr, the bet would be weighted 7.5/1.5 times more towards the 10yr than the 2yr since the 10yr has that many more times the duration.

similarly, you have to match the volatilities for the diffbets of anything you do (i.e. if you had a diff bet short JGBs vs. long Gilts...japanese rates rise relative to UK rates...then you have to adjust the bet size so that you end up with a balanced position...the JGB is far more volatile than the Gilt). i used 2 simple examples (1and 3) and one semi complex example (2) to show that a lot goes into managing these bets.

anyways, i hope this answers your questions.

i also hope i haven't made any silly mistakes, though taht is possible since my brain is fried at the moment.

using matlab can be a pain in the freaking ass when you have nested loops of crap and you forget a semicolon and the program blows up [img]/images/graemlins/frown.gif[/img]

Barron
Reply With Quote