Treasury bonds with nominal model


#1

The nominal model seems to work pretty well with Treasury bonds. Looks like the 18M cycle is dominant with maybe a little shortening in the most recent turn. Expect a retest of the February price lows this summer, which looks to be at least a 54M low. This would be consistent with more Fed tightening and a final move up in stocks after the upcoming 9 month low (although the long-term correlation between stocks and bonds has broken down recently). This would set the stage for higher Treasury bond prices and lower stock prices in the second half of the year and into 2019.


#2

Curt,
My (ZB.F) weekly model calls for higher prices in July. I tried standard Hurst model similar to yoursbut composite did not tested well on past data. Confidence level in my model still is not high enough to recommend it, need to spend bit more time testing in other frames; monthly, daily.


#3

Hi Gary,

The above analysis suggests a trough in May or June and higher prices thereafter. I hear you about the model being questionable though. Hadn’t really looked at Treasury prices using Hurst’s nominal model and was surprised to see it actually made some sense (I’ve done a reversal and now prefer to use Treasury futures or ETF prices rather than yields). I ran spectrum and a Bartels test on TLT and it shows the 42 month (Kitchin) cycle to be very dominant, so if you were building a model from scratch this is what you would probably use as a starting point. This is consistent with your thoughts about alternate models that might backtest better (PQ Wall, etc.).

Curt


#4

Problem is, monthly model does not make any sense and calls more downside and yes Bartels id’s Kitchin as strongest cycle.
bonds30_fft


#6

9.3 year cycle FLD seems to support price on monthly time frame.


#7

3 to 1 harmonics change the interactions but this still makes sense to me (see diagram from David H. below). Expected low is the week of 6/7/2018. IEF (7-10 yr Treasury ETF) might hold at the 12/2013 level, which is the 9 year FLD. This is only about a 3% price drop and equates to a yield of about 3.05% on the 10 year from 2.82% today. The long bond might get much uglier. As your chart shows for the futures, the 9 year FLD is about 15% below the current price of TLT. That would be a yield of about 3.95% versus 3.06% today. This upcoming price trough might actually be an 18 year low.


#8

weekly 9.3y(red) and 38 m (orange) FPL


daily mod and want to see 14th low confirmed before deciding if daily model is any good


#9

It’s starting to look like stocks and bonds will both fall between now and May/June.


#10

My mods both weekly and monthly (stocks) point down at list in to the summer, but I would like to see May as a key month.


#11

At least bonds will be a buy after this upcoming trough (18 yr low?). Stocks have a long way to go before they are attractive on a longer time horizon, although there will be great trading opportunities with this volatility. Buy and hold for stocks (and Bitcoin), is dead :stuck_out_tongue_winking_eye:.


#12

Many so-called traders will find it hard to cope with prices that move in two directions.


#13

Bond mods are holding nice so far. Daily trend is up and away just on time. Next buying opportunity some time in the second week of May at the nominal 10 weeks cycle low.


#14

The last low needs to hold or May 10 week cycle low maybe more severe that model has initially forcusted . In any event it looks at best flat and May seems to be a better opportunity.


#15

Below is the ST long term view for bonds. The low is expected in early to mid-June. What I find interesting is that bonds are due to peak at a new all-time high in 8/2020, which is the expected low in stocks.

Short term view. This final move up in rates and down in bonds will likely mark the end for the economy and stocks in this cycle.

Bonds like the default nominal model, which makes sense since bonds and stocks are so highly correlated.


#16

8/2020 is to far away. I am not sure that ST or any other composite can be trusted this far in to the future. I would need to build monthly model but the dataset is too short.


#17

Agreed. I was just thinking I don’t believe in coincidences.

The above long term view is a daily analysis with a monthly chart (the 54m low in 2013 probably should be pinned a few months later but I didn’t bother). It will be interesting to do a longer analysis that includes the 9 and 18 year cycles back to the late 70s when I get time. As you showed above, it might be more bearish because the 9 yr FLD has consistently provided support in this long bull run in bonds.


#18

Standard Hurst model does not back test well and I could not yet construct any other model that is stable, especially on the weekly time frame and higher. The recent price action may imply that we already have seen 10 weeks cycle high. Below are the dominant cycles in Bonds.
image
The question is if bonds get a bounce from the 424 days cycle low.


#19

Curt,

Following weeky mod tested reasonably well in the past. It would be interesting to see ST composit line with same parameters.
11,24,63,181,485,936 (w)


#20

Interesting. It’s kind of a Wall/Hurst hybrid model. I assume the time frame is as far back as we have bond futures data, which is the late 70s. As we know, “past performance is no guarantee of future results”.

Busy with other things right now but I’ll take a look when I get time.


#21

Yes, “past performance is no guarantee of future results” but without backtesting we have nothing to objectively verify any model.