Back in Brazil, time for the the second half of Market Microstructure - Confronting Many Viewpoints 2016 (more details at http://market-microstructure.institutlouisbachelier.org/?lng=FR - .WEiTp6IrKbU ).

One presentation I was looking for was, of course, Khalil Dayri (Bloomberg) on predicting the consequences of tick value changes using the Tokyo Stock Exchange pilot program as an experiment. The paper analysing the Tokyo results can be found at https://arxiv.org/abs/1507.07052, and the original paper on optimal tick sizes is https://arxiv.org/abs/1207.6325. The original uncertainty zones model paper by Robert and Rosenbaum was published at http://jfec.oxfordjournals.org/content/9/2/344.short and as a chapter of "Econophysics of Order-driven Markets". There are other references around describing the model (with more or less details), like my work on Tick Sizes, which can be found at https://www.researchgate.net/publication/304011741_Ticks_Coins_and_Traffic_Lights_-_How_the_choices_of_tick_size_and_fee_schedules_can_define_the_market_structure?ev=prf_pub (Chapter 16 of the book I co-wrote is a brief summary of the results).

It's a great model, it works well in describing data and it also predicts well what happens when parameters change. So why the #!?$ was their comment letter to the SEC Tick Size Pilot Program not considered as it should?

<End of rant>

<Breathes deep>

Look at the last KCG paper about tick sizes in Europe! What? Haven't you heard of it? Alexander talked about this last week.

What? Who is Alexander? He is Alexander Laumonier, of course, and I was glad to meet with him again; we met Sebastian Neusüβ and had a great discussion about architecture and data. Check the Eurex presentation here: http://www.eurexchange.com/blob/238346/5e2ce06990dd2a2e108fd2030dfcf5a2/data/presentation_insights-into-trading-system-dynamics_en.pdf and the KCG paper https://www.kcg.com/news-perspectives/article/does-europe-have-a-smarter-tick-proposal . Look also for the non-HFT profile of Alexander, like this book he edited: http://www.zones-sensibles.org/wp-content/uploads/2014/05/ZS_Blois_GrandPrix.png . Life is not only HFT, and it goes by even faster.

Next up was Mark Van Achter (Erasmus University Rotterdam), "Trading Speed Competition: Can the Arms Race Go Too Far?", available at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2779904; it looks at the economics of investing on being faster.

Felix Patzelt (CFM) showed how complex systems can amplify bursts of activity on "Instability from Information Efficiency" ( https://arxiv.org/abs/1511.03732 ).

A brief pause for the rallying chant of the HFT studies:

-What do we want?

-HFT labeled data

-What do we have?

-NASDAQ 2008/2009

"High-Frequency Trading and Extreme Price Movements", presented by Ryan Rordan (Queen's University) continues the spirit of the Brogaard/Kirilenko papers ( https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2531122 ) looking at the NASDAQ dataset and how HFTs trade during periods of extreme price movements.

The round table "Are buy side the new liquidity providers" with Nej Djelal (Barclays), Zoltan Eisler (CFM), Nicolas Megarbane (AMF) and Yazid Sharaiha, (Norges Bank Investment Management) discussed, among other themes, the dark execution caps in Mifid II, Plato Partnership, execution of block-sized trades, etc. Norges Bank has recently also put forward a research programme: https://www.nbim.no/en/investments/research/the-norwegian-finance-initiative-nfi/nfi-research-programme/ .

Anna Obizhaeva (New Economic School) presented "Dimensional Analysis and Market Microstructure Invariance", a very interesting topic ( https://pages.nes.ru/aobizhaeva/Kyle_Obizhaeva_DA.pdf ). There was an interesting discussion afterwards about how the relative tick sizes can explain the different intercepts on Figure 2. Look for a similar analysis about the linear relationship between volatility squared and the number of trades in my work and in Dayri and Rosenbaum.

One of the themes of the conference was cross-impact (e.g. Thomas Guhr), and Michael Benzaquen (CFM) presented "Dissecting Cross-Impact on Stock Markets: An Empirical Analysis" ( https://arxiv.org/abs/1609.02395 ).

Thursday closed with a brilliant presenttion by Fabrizio Lillo (Scuola Normale Superiore di Pisa): "Detection of Intensity Bursts Using Hawkes Processes: an Application to High Frequency Financial Data" ( https://arxiv.org/abs/1610.05383 )

The idea of comparing bursts at points where you expect them (macro news events) with other points is very interesting, and the results justify more research.

So, for those interested in modeling microstructure processes, Hawkes processes are a very useful tool; not the truth (remember Derman's distinction between theory and models: http://www.econtalk.org/archives/2012/03/derman_on_theor.html ), but if you want to simulate prices or extract some information, learn it (that's my plan anyway).

The Hawkes vibe continued on Friday, with Aurélien Alfonsi (Ecole Nationale des Ponts et Chaussées), presented "Optimal Execution in a Hawkes Price Model and Calibration" (earlier paper at https://arxiv.org/abs/1506.08740 ).

The questions arising at Rama Cont (Imperial College)'s presentation on "High Frequency Dynamics of Limit Order Markets" (similar presentation at https://workspace.imperial.ac.uk/quantitativefinance/Public/events/SLIDES 3RD IMPERIAL ETH/Rama Cont 1.pdf ) were interesting: Similar states of the limit order book may arise from different events (trade or cancellation) - Can a model of the order book state alone describe adequately trades?

More interesting questions also after Paul Besson (Kepler Cheuvreux) presented "To Cross or not to Cross the Spread: That is the Question!" ( https://www.researchgate.net/publication/308083412_To_Cross_or_Not_to_Cross_the_Spread_That_Is_the_Question ). After showing the results of using order book imbalance to choose whether to be agressive or passive on your child order, Robert Almgren questioned the fact that all estimation was conditioned on a trade happening (my thoughts on that: I agree, such an algorithm might be susceptible to spoofing). My unspoken question was: If the signal is so strong, why does it persist for so long?

Henri Waelbroeck (Portware LLC) presented "How the Market Digests Earnings Announcements: can we Reconcile Options Prices, Fat-Tailed Earnings Shocks and Fair Pricing of Institutional Metaorders?"; some of his conclusions on the impact of cash orders are similar to those in his 2014 presentation ( http://market-microstructure.institutlouisbachelier.org/uploads/91_3 WAELBROECK Paris2014_HWaelbroeck Short.pdf ).

Sophie Moinas (Toulouse School of Economics - https://www.tse-fr.eu/people/sophie-moinas - publications ) presented "The Role of Pre-Opening Mechanisms in Fragmented Markets"; she has been researching price behavior in multiple venues.

Matthew Baron (Cornell) presented "Risk and Return in High-Frequency Trading" ( https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2433118 ); these articles usually have the same Broogard Kirilenko connection.

Albert Menkveld (VU University Amsterdam - http://albertjmenkveld.com/ ) closed the event with "A Network Map of Information Percolation" ( https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2770313 ), looking at the Swiss Franc big move.

The event also had the presence of Donald MacKenzie ( http://www.sps.ed.ac.uk/staff/sociology/mackenzie_donald ), a must read ("An Engine, Not a Camera: How Financial Models Shape Markets"). At the speakers' dinner, he discussed his research on how price prediction became such an algorithmic staple. HIs ATD paper is available here: http://www.sps.ed.ac.uk/__data/assets/pdf_file/0018/216261/ATD35.pdf.

So, after 4 days of high-level pesentations, what can we extract from it?

- Microstructure is complex, because behavior of agents is influenced by the fact that some agents are actively modeling the behavior of the agents

- Math is useful, but data should speak louder - do not become too enamored of your model if it doesn't explain reality too well, otherwise you'll end up with the microstructure equivalent of epicicles

- But there is information to be extracted and some of it can be modeled

- Take a look at Hawkes processes

- Read Dayri and Rosenbaum

- Read Kyle and Obizhaeva (yes, it is interesting) and similar works on trading invariance - even if you don't want to believe in the theory, the patterns in the data are informative

- Keep up with research! Most of it will be on SSRN/Arxiv.

Thanks for reading it all.