#market microstructure

[diegocom]

Institutional Liquidity and Software Resiliency: A DeFi Lending Infrastructure Audit

Analyzing the intersection of decentralized liquidity protocols and enterprise software architecture, with a focus on risk mitigation, technical SEO for discovery, and the transition of DeFi to institutional rails.

Read the full article

[storentez]

Crypto Long & Short: The Emergence of Prime Brokers Adds Resilience but Also Risk

Crypto Long & Short: The Emergence of Prime Brokers Adds Resilience but Also Risk

There are many types of hockey-stick evolutions in new industries. Startups dream of the lying-on-its-back L shape for their sales growth. Entrepreneurs chase the “escape velocity” funding momentum. And a subsector can simmer along with a low hum of activity until boom, years of progress and feverish months of under-the-surface connecting explode in a series of announcements and launches that…

View On WordPress

[thecryptoreport]

We Think 10-Minute Settlement Is Slow but That’s Crazy, Feat. CoinDesk Research’s Noelle Acheson

We Think 10-Minute Settlement Is Slow but That’s Crazy, Feat. CoinDesk Research’s Noelle Acheson

CoinDesk Head of Research Noelle Acheson joins the Amun State of Crypto crew to discuss the advantages, disadvantages and eccentricities of crypto markets, exchanges, settlement and more.

In this deep-dive conversation, the group discusses the differences in the market microstructure between the crypto asset industry and traditional capital markets. The guests drew from their experience within…

View On WordPress

[privatesectorbureaucrat-blog]

Simple R Implementation of Roll's Model

So I've been trying to learn R in order to implement some of the stuff I've been reading in my microstructure books and had my first success today implementing Roll's model of a dealer market.

Background of Roll's model

Roll's model models a market in which all trades go through a dealer, pricing is competitive, and prices follow a random walk. The efficient price (the security's "real" value) is denoted as and is assumed to be a random walk which follows this equation:

Where is a zero-mean random variable (I just used R's built in rnorm function because I'm lazy).

We will call the best dealer quotes at time t: and (bid and ask, respectively). The dealer assumes a long run cost of c per transaction. Since the market is efficient, we can assume the bid and asks are:

Knowing all of this, we can model the transaction price as:

Where is a directional indicator set to +1 if the customer is buying and -1 if the customer is selling. Buys and sells are assumed to be equally likely, serially independent, and that agents buy or sell independent of drift ().

Money Shot: The chart

Blue is the bid, red is the ask, and green is the transaction price.

The above is a chart of a security following Roll's model with a starting efficient price of $10 and a c of $1 (Really high, but it makes it easy to see what is going on).

The chart makes it really easy to see some of the cooler conclusions. The bid ask spread in a dealer market is a constant (2c) related to dealer costs (here it's the gap between the blue and red lines) and the bid-ask bounce are readily apparent. (Since green is the transaction price, and trades must occur at the bid or ask, the price fluctuates between the two lines. Some extrapolation from this and you can get the basics of the first high frequency trading strategies.)

R Code

##Time of course t <- 0:50 ##Drift in our model Ut <- rnorm(51) ##Seed Our efficient price function with $10 as starting price of security Mt <- c(10) ##Generate Efficient Price for (n in 2:51){ x <- Mt[n-1] Mt[n] <- x + Ut[n] } ##Create Directional Indicator Qt BuyorSell <- c(1, -1) Qt <- sample(BuyorSell, 51, replace=TRUE) ##Set transaction costs / long run cost of competitive dealer per transaction c <- 1 ##Generate bids Bt <- vector() for (n in 1:51){ Bt[n] <- Mt[n] - c } ##Generate Asks At <- vector() for (n in 1:51){ At[n] <- Mt[n] + c } ##Generate Transaction Price Pt <- vector() for (n in 1:51){ Pt[n] <- Mt[n] + (Qt[n] * c) } ##plot transaction prices V Time xlim <- range(t) ylim <- range(c(At, Bt)) plot(t, At, type="l", col="red", xlim=xlim, ylim=ylim, ann=FALSE) lines(t, Bt, col="blue") lines(t, Pt, col="green", lty="dashed") title(main="Roll's Model: Price V Time", xlab="Time", ylab="Price")

[allthequiettigers-blog]

Of Mark Cuban's dumbassery and financial reform

Yes, I know it's a beautiful day out.  I'm writing outside in New Jersey today.

Mark Cuban's comments about the AP twitter hack fiasco are so utterly misguided - the man is living proof that you need not be smart or possess basic logical reasoning skills to be a billionaire - one really only needs to take outsized risk.  Mark Palmer from Streambase already took him to task, but I thought it necessary to make a couple of extra points about his assertions that Palmer didn't quite touch on.

Cuban: "all of this algorithmically driven trading has created a ton of downstream problems"

"He proposes solutions" to the "problems" that include taxing trades that are held for less than one hour"

Ok - financial transaction taxes are most likely to have disproportionate negative effects on liquidity to whatever nebulous "gains" you might receive.  I think of them as a massive value-detractor.  Rather than offer stability to markets, an added burden in the form of heightened frictional costs will only increase short-term volatility, as fair value is redetermined on a less-frequent basis.  The "blip" caused by the tweet (because let's face it - a 15-pt drop is only considered significant in today's ultra low vol environment) would have likely been significantly worse had market-makers not been able to step in immediately upon figuring out that the report was false.  Additionally, it is likely that they would result in an even further thinning of book depth as smaller market-makers with less significant economies of scale than say, a GETCO, will find their already-tenuous edges eaten into to a point where it makes little sense to continue trading.  Margins for these firms are razor-thin in US equities - one need only look at the financials for a Knight Capital, GETCO, or Virtu to see that this is the case, and this would only serve to push the market into further consolidation.  Isn't this what the regulators allowed ECNs for in the first place?  To create a "competitive landscape"?

It is not the case that markets haven't been roiled by false information before, and program trading is not in and of itself destabilizing in nature.  Central planning and regulation, such as the half-assed implementation of Dodd-Frank and the bizarrely inconsistent regulations that have been imposed on OTC markets are a great example of attempts to rein-in trading that have ultimately just resulted in a reduction of the efficacy of natural price discovery mechanisms.  Financial transaction taxes, while seemingly harmless, will do alot more to damage the already-fragile equity market structure than they will to "level the playing field".

I've long felt that these complaints stem from a feeling of inadequacy - an unwillingness to adapt.  What I find curious is that large institutions have not fundamentally changed their execution methods, by and large, despite evidence of the necessity to.  One way to think about this is in a Darwinian sense - if colocation and advanced execution is available, and one chooses not to utilize it, is it really wrong?  I can colocate my servers for a personal account for roughly $200 per month, and my own funds are certainly not institutional-sized.  What exactly are the "downstream problems" that have been created?  It seems to me that it is more an issue of market infrastructure failing to adapt as quickly as trading strategies have.  What MAY be worth looking into further from a regulatory perspective are the ATS's - such as those at Citadel and Knight.  The fairness of so-called "flash orders" has already been ruled on - the inability  to opt out of having your flow pre-shown to certain market participants to give them the option of improving on the NBBO or passing is NOT okay, but what amounted to systematic front-running has been dampened significantly since proposals in 2009 to ban the practice led to voluntary stoppage by most market makers.

My view, honestly, is that the US equity markets should look to the CME as an example for what microstructure should look like.  The futures exchanges are simpler in design and much more robust as a result.  Algorithmic and high-frequency trading are prevalent in most liquid futures contracts as well - the difference lies in the complexity of microstructure.

Anyway, those of us in the tri-state area should be out and about enjoying the weather this weekend.  Have a beer at Dorrian's or pop downtown to Bar Pitti and meet a cute Euro girl.