Understanding Fixed Income Relative Value

qwerty_phynatical · Post by **qwerty_phynatical** » Thu Jul 02, 2020 4:40 am

I've been trying to read up on different strategies that funds utilize. I'm a little confused about FI RV. From what I've read Relative Value trading is usually a subset of statistical arbitrage. However I see a lot of fixed income RV funds that are purely discretionary.

However aren't stat arb strategies usually implemented in a systematic setting? Or am I missing something?

Likewise, are equity market neutral which are relative value strategies, implemented systematically only or they too can be discretionary?

In the sense that by systematic I mean the algorithms themselves book the trades whereas in discretionary humans have to actually make the call and book the trades.

Any kind of enlightenment will be really helpful. Thanks.

tbretagn · Post by **tbretagn** » Thu Jul 02, 2020 7:32 am

There are a bit more to the world than direct market access.

For example one RV strat would be to trade bonds versus swaps. Given the broken dates you need to do (ie your swap is not a simple spot starting 10Y but the matched maturity of the bond) it means the machine can't do the swap. And you alos need to repo the bond which might no be machine accessible.

Bit like shorting equities. Someone needs to do the repo, and there is an element of human intervention after all.

Billymazee · Post by **Billymazee** » Thu Jul 02, 2020 3:27 pm

My gathering is that the order mentioned in your second sentance is the other way around. Statistical arbitrage is usually considered a type of relative value, rather than RV being a form of stat arb. You're correct in that most stat arb strategies are done systematically, but not all fixed income RV strategies are discretionary. Strategies in the more liquid markets can be automated pretty easily, especially if you're trading futures or the most liquid cash Treasuries. However, some markets are less liquid and/or require longer time horizons, which might require a discretionary approach.

On the equity market neutral side, the space is pretty divided between stat-arb and discretionary. Most of the largest multi-manager funds (Millennium, Citadel, Balyasny, Point72) run both stat-arb/higher-frequency strategies as well as fundamental discretionary equity RV. This is largely due to the tight risk controls they impose on their PMs, with the hard drawdown limits largely forcing the managers to run market-neutral. The funds that run both benefit since stat-arb and fundamental EMN are (usually) not highly correlated and have different risk exposures due to differences in portfolio concentration, re-balancing frequency, style titls, etc. However, both stat-arb and discretionary equity market strategies are very common.

Hope this helps.

EspressoLover · Post by **EspressoLover** » Sat Jul 11, 2020 6:30 pm

> From what I've read Relative Value trading is usually a subset of statistical arbitrage.

I don't know if I really agree with this. With the caveat that these are ambiguous terms, one of the hallmarks of stat-arb* is that there's a specific forecast over a fixed horizon. Whereas this isn't always the case with RV.

Consider the classic RV trade of buying off-the-run treasuries at a discount to on-the-run. It doesn't involve any directional view on rates, just the belief that the off-the-run bonds are trading too cheap relative to their on-the-run counterparts. The RV trader believes that the spread will inevitably narrow but doesn't necessarily have a specific timeframe. It could be tomorrow, it could be six months. (With regard to fixed income, it certainly helps that RV trades almost always have positive carry.) In contrast a stat-arb strategy generates a signal at a fixed horizon. If you traded OTR stat-arb style, you'd want to forecast something like we expect the excess spread to decay by 25% per month.

This may seem like a distinction without a difference but there's an important subtlety. With forecasts you can systematically blend disparate signals and instruments using a standardized metric. Portfolio construction becomes a purely mechanical function of maximizing return relative to risk. Otherwise you're comparing apples to oranges.

Let's say you're an RV portfolio manager. You see that OTR spreads are two times wider than normal, bond-CDS basis is only 50% wider but carry is very high, cash-futures basis isn't that wide but you expect an upcoming event to catalyze narrowing. How do you decide to allocate capital? How much powder should you keep dry for future opportunities? The only solution is to inject human judgement and intuition. This is why you wind up with a discretionary manager in the loop.

A natural question is why don't RV funds all take the stat-arb approach? And I think the answer is because the latter requires much more powerful analytics and sophistication. Discretionarily trading the OTR spread doesn't demand much more than a souped-up Excel sheet, a Bloomberg terminal to pull prices, and a couples years of experience to get a gut feeling for what's "wide" given today's conditions.

In contrast going full systematic stat-arb would require years of pristine historical data (tough in OTC markets), accurate models of liquidity and execution costs (also tough for OTC), and careful training of historical models against a backdrop of sharp regime changes triggered by breaks in banking, regulatory and macroeconomic conditions. To be honest, I think a lot of times this approach isn't really possible for many classic FI RV strategies. AFAIK, RennTech and similar firms, who should be well positioned, don't really swim in those waters.

(*As a postscript, I'll add that "stat-arb" is an overloaded term. It can refers to a generalized set of techniques like signals fitted on historical returns, mean-variance portfolio optimization and orthogonal factor decomposition. It also commonly is used to describe a cluster of equity market neutral strategies that focus on sources of alpha around return reversals, earnings events, analyst forecasts, etc. The two often go together, but not always. For this post, when I say "stat-arb" I'm referring to the former.)

Strange · Post by **Strange** » Sat Jul 11, 2020 9:42 pm

Adding to what @EspressoLover says, FI relative value frequently revolves around a relative value model/view that is overlaid over a strong macro view. In essence, a lot of RV players are trying to assemble a portfolio that has positive expectation overall (carry of some sort) and will perform well in a specific macro scenario.

tabris · Post by **tabris** » Wed Jul 15, 2020 2:32 am

I think what people consider "classic" FI RV is different than what people consider "classic" stat arb as @Strange/@EspressoLove mentioned.

But the two worlds are indeed blending and there are niche folks that do "non classic" FI RV that are closer to market making using "stat arb" type models. These guys have been growing post GFC... so basically all roads lead to Rome in one sense or another

Jurassic · Post by **Jurassic** » Wed Jul 15, 2020 8:55 am

Is there much of a difference between curve shape rv and cash/basis trades, see page 2 this document https://www.schroders.com/getfunddocument/?oid=1.9.2744529

mtsm · Post by **mtsm** » Thu Jul 16, 2020 11:37 pm

Yes.

I don't have time now, but what EL wrote is bs. I can elaborate on this later.

cherk · Post by **cherk** » Tue Jul 21, 2020 6:07 pm

Pls elaborate when you get a chance. Thanks!

tabris · Post by **tabris** » Fri Jul 24, 2020 12:34 am

@Jurassic yes there is a difference...

@cherk I don't want to speak for anyone else but at the end of the day I think it boils down to what people classify as RV/stat arb and the classification will give raise to disagreements about what there can be or cannot be done.

Much like in stat arb, there are equity long short guys who use stat arb + discretionary decisions outside of just their signals to build a portfolio. Likewise, FI RV will have some souped up excel spreadsheet managers who look at off and on the run bond spread using z score/z spread/pca/etc plus their own macro views as to why this spread is there. But there will be guys who would look at FI RV in a completely autonomous fashion where they build a portfolio of RV trades using optimization methods as well.

Just because you haven't heard of it doesn't mean someone isn't doing or trying it. Plus, the guys who are successful at systematically doing FI RV would not tell you they are doing it systematically. Neither would their investors know because all they need to tell them is they are doing FI RV. The details on how to select a bond or systematically size such trades/portfolio optimization of FI RV trades/entry and exits/risk management of such trades are probably secrets kept close to their heart and gives them an edge versus "classic" FI RV guys. Nobody in stat arb space will tell you the stuff mentioned above neither if an investor asks but a stat arb investors just knows that these are probably important in systematically churning out good stat arb trades as opposed to FI RV side.