Today saw a massive short covering trade occur in the US Treasury market. A broad base of traders got short treasuries over the past few days (long end auction setup trades...and the usual group who just MUST fade every move). Yesterdays weak ADP and ISM data brought in a large buyer of bonds (and a medium size seller of stocks). Today the stocks market went sideways while the Treasury market was rocked...the herd of Treasury shorts was caught and fear overtook all rational thought. There was a very high volume short covering event in the morning that lasted for about an hour and saw 100mm/basis point trade in a straight line "up" (this is 4x normal volume for that time period). While all this craziness was taking place in the Treasury market...the stock market was "unch." Its as though stock market participants were completely unaware of what was happening in the Treasury market.

3 weeks ago (March 14th), S&P futures opened 1554 while 10yr Treasuty futures opened @ 130-10. Today, S&P futures closed @ 1555 "unch from 3 weeks ago" while 10yr treasury futures closed @ 132-26 (thats 2 1/2 POINTS on a 7yr piece of paper).

3 weeks ago the relationship between stocks and bonds was priced at fair value. Today, the stock market hasn't moved and the Treasury market has been ROCKED. On a relative value basis, the Treasury market is now 20 basis points rich compared to the stock market. To put that in S&P terms...the S&P is currently trading 1555....if the S&P were to move to accommodate the move in the Treasury market...the S&P would need to trade 1450...that's 100 POINTS lower (or 6.4%). This is the most rich these markets have been relative to each other in years. The current correlation regime has lasted 9 months (treasuries and stocks have been dancing with each other in a stable fashion for the past 9 months). Even adjusting for regime change, this move today is the richest these markets have been even during this current 9 month regime.

So, either the regime between bonds and stocks has just changed...or the market has just given traders with cash on the sidelines a gift. History has taught us that regime changes don't happen very often...and the smart move for a trader has been to fade these attempts to "break the regime." Its true that someday the regime will shift, and we will enter a new regime, and fading a move like this will lose money. Do you think that just happened? Do you really think we just entered a new regime? Its true that Japan just embarked on a new round of QE...and it is a 2.5 times the size of current US QE, and North Korea is threatening war. Should that break the regime of the relationship between US stocks and US bonds? Someday, something will do just that. The question for traders is...do you think the regime just shifted.

Personally, I'm not convinced, and so i faded this move at the close by selling ES and ZN (S&P futures and 10yr Treasury futures) at a ratio of 4 ES vs 11 ZN contracts. If you are trading a 1mm$ account and wanted to put 1/3 of your margin to work on this leveraged trade, then you would sell 40 ES @ 1555 and sell 110 ZN @ 132-26 (thats 317k of margin...so 1/3 of your futures account as an example). The trade to fade this move is to sell both (sell ZN and Sell ES), which is exactly what I've done.

Follow @govttrader on twitter to see what happens with NFP and next weeks long end (10yr and 30yr) treasury auctions. Did the regime just change...or did the mkt just provide me a gift? My expected timeframe for this trade is 1-2 weeks. As always, my expectation for the future will determine when i actually exit this trade.

Disclosure: I am not advising anybody do this trade with me...this is a leveraged trade, and short positions have the theoretical potential for unlimited losses.

govttrader out...

Check the embedded twitter feed over to the right ----->>

## Thursday, April 4, 2013

## Tuesday, April 2, 2013

### Stock vs Bonds (and trading other correlated assets)

I get asked a lot of questions about how i measure 10yr yields vs ES to get a RV valuation. I think i've posted the details on this blog in the past....so here i'm just talking about the basic premise...when you want to trade any pair based on a historical relationship.

When you think 2 securities are correlated..the first thing you normally do is ask...how can i tell if this is true? The first route is to take prices, and graph them over some time interval (5 min, hourly, daily, etc...). In this case, i took closing ES prices and 10yr yields. I noticed that if i look from June 2012 to the present..there seems to be a tracking relationship...but with some deviation. This is exactly what we want to see in a correlated pair...basic tracking...but with enough slippage that the pair mean reverts over time to some "fair value" (both towards and away from...ie...it gets both rich at times..and cheap at other times). I know from experience that the treasury mkt tends to selloff or weaken between the start of the month and long end supply (10yr and 30yr auction supply concession). I also know that treasuries tend to richen towards the end of the month (the month-end extension trade). When i looked at this pair graph, and overlapped these time frames..lo and behold...this tends to be true...not always to the exact date...but generally speaking. So taking a few basic measurements and some time in excel...aong with a few assumptions that i had to verify from the data, i calculated that 4 ES contracts vs 1mm 10yr notes (or 11 ZN contracts) gave me the correct PnL graph to match the graph of the securities levels (after adjusting by a factor to put them in consistent terms). You can do this yourself if you spend some time in excel with the historical data. Then to answer the question "are treasuries rich or cheap vs stocks?" you can look at the graph and put the pair in historical context. I take this a step further and add a "fudge factor" to center the bulk of the historical distribution at zero...the highest and lowest points on the graph are not exactly equal distant from zero...but pretty close. Then i measure the distance from the point i've created as my "fair value zero" to the current mkt level...and violla....rich/cheap. While you could just as easily be measuring the rich cheap of ES...the UST component tends to be the more volatile component...and thus the one i've chosen to base my measurements. I could just as easily be saying that ES is 65 pts too rich vs where treasuries are trading right now...but since i more often just trade the bond leg on its own...i prefer to quote in terms of UST..so basis points (ie..if the 10yr were to selloff 13 bps right now..and ES didn't move...then i would say both are priced at fair value...alternatively..ES could selloff 65 pts and 10yr yields not move...and i would also say both are priced at fair value...given the last 9 months of this relatonship). This is not rocket science...just lots of basic graph interpretation and some basic math...combined with my intuition of what i expect...backed up with 9 months of mkt history.

If you do this exercise yourself (be prepared to spend some time fiddling in excel and playing with fudge factors to define the relationship)...you should find that the spread between ES and 10yr yields looks almost like a sin wave (from geometry...sin / cosine / tangent...remember those days??). That is a perfect example of a mean reverting relationship if you ask me. And since this is trading..you really never get perfect. If you look back pre-June 2012...this relationship breaks down. I haven't spent the time to figure out what the relationship was back then...but i can say with certainty that the regime shifted in the 1st half of 2012...and the resulting regime gives us this nice relationship to trade. I'll take it for as long as it lasts...and then when it breaks down (which it eventually must)..then i'll move on to something else. Thus is the nature of trading....

I stopped graphing the relationship back in January...since i built a realtime trading system and i'm involved in trading this everyday, i don't need the graphs anymore..so i stopped updating it...but if you do the work you should be able to reproduce what i have and then fill in the blanks for the more recent time period.

govttrader out...

When you think 2 securities are correlated..the first thing you normally do is ask...how can i tell if this is true? The first route is to take prices, and graph them over some time interval (5 min, hourly, daily, etc...). In this case, i took closing ES prices and 10yr yields. I noticed that if i look from June 2012 to the present..there seems to be a tracking relationship...but with some deviation. This is exactly what we want to see in a correlated pair...basic tracking...but with enough slippage that the pair mean reverts over time to some "fair value" (both towards and away from...ie...it gets both rich at times..and cheap at other times). I know from experience that the treasury mkt tends to selloff or weaken between the start of the month and long end supply (10yr and 30yr auction supply concession). I also know that treasuries tend to richen towards the end of the month (the month-end extension trade). When i looked at this pair graph, and overlapped these time frames..lo and behold...this tends to be true...not always to the exact date...but generally speaking. So taking a few basic measurements and some time in excel...aong with a few assumptions that i had to verify from the data, i calculated that 4 ES contracts vs 1mm 10yr notes (or 11 ZN contracts) gave me the correct PnL graph to match the graph of the securities levels (after adjusting by a factor to put them in consistent terms). You can do this yourself if you spend some time in excel with the historical data. Then to answer the question "are treasuries rich or cheap vs stocks?" you can look at the graph and put the pair in historical context. I take this a step further and add a "fudge factor" to center the bulk of the historical distribution at zero...the highest and lowest points on the graph are not exactly equal distant from zero...but pretty close. Then i measure the distance from the point i've created as my "fair value zero" to the current mkt level...and violla....rich/cheap. While you could just as easily be measuring the rich cheap of ES...the UST component tends to be the more volatile component...and thus the one i've chosen to base my measurements. I could just as easily be saying that ES is 65 pts too rich vs where treasuries are trading right now...but since i more often just trade the bond leg on its own...i prefer to quote in terms of UST..so basis points (ie..if the 10yr were to selloff 13 bps right now..and ES didn't move...then i would say both are priced at fair value...alternatively..ES could selloff 65 pts and 10yr yields not move...and i would also say both are priced at fair value...given the last 9 months of this relatonship). This is not rocket science...just lots of basic graph interpretation and some basic math...combined with my intuition of what i expect...backed up with 9 months of mkt history.

If you do this exercise yourself (be prepared to spend some time fiddling in excel and playing with fudge factors to define the relationship)...you should find that the spread between ES and 10yr yields looks almost like a sin wave (from geometry...sin / cosine / tangent...remember those days??). That is a perfect example of a mean reverting relationship if you ask me. And since this is trading..you really never get perfect. If you look back pre-June 2012...this relationship breaks down. I haven't spent the time to figure out what the relationship was back then...but i can say with certainty that the regime shifted in the 1st half of 2012...and the resulting regime gives us this nice relationship to trade. I'll take it for as long as it lasts...and then when it breaks down (which it eventually must)..then i'll move on to something else. Thus is the nature of trading....

I stopped graphing the relationship back in January...since i built a realtime trading system and i'm involved in trading this everyday, i don't need the graphs anymore..so i stopped updating it...but if you do the work you should be able to reproduce what i have and then fill in the blanks for the more recent time period.

govttrader out...

Subscribe to:
Posts (Atom)