A brief history of US Treasury bond arbitrage, presented by a recovering bond trader, for the benefit of probably nobody at all.

In the 80s, there were full %age pts of “true” arbitrage opportunities in the yield curve, often lasting months before convergence.

“True” arbitrage” = genuine mathematical mispricing, with no liquidity or capital cost or regulatory or nickel/steamroller caveats.

You would run models overnight, then compute daily changes with pencil and paper. My first boss did just that, at Salomon Arb.

In the mid 90s, the opportunity diminished to mere 10s of basis points, lasting mere weeks. Excel was the state of the art.

By the late 90s and earlys 00s (post LTCM, when I started trading) opptys were usually single digit bps, lasting a few days at most.

And even that was dwindling. In response, we built one of the first “real-time” yield curve arb systems at my hedge fund.

By the mid 00s, our system was capturing <1bp opptys which would last hours at most, minutes more often.

We automated everything. Price feeds, model calcs, portfolio rebalancing, trade recommendations in real time.

Today, even 0.3bp is hard to find. Model-driven electronic trading has gone from <10% of market to >90%. From 3 firms to 300.

Fascinating how every generation of bond traders sees opptys reduce by a factor of 10 both in magnitude and in persistence.

Technology (Moore’s law), spread of knowledge, availability of risk capital, deregulation all contribute to way more efficient market.

Compared to mid 1980s, mid 2010s bond arb has 1/1000 smaller opptys lasting for 1/1000 the time.

Amazing “progress” albeit in the service of an ever more obscure and meaningless goal.