QuoteI have read your paper quite quickly, it's a big oneDespite the fact I am not a quant I found it very interesting. For the good point, I found it very clear, well written and mathematically simple (it changes from paper with stochastic volatility or paper with laplace transform for high path dependant derivative, with semimartigale stuff under change of probability measure over a filtration where at the end the reader understand nothing).Thanks! Really apreciate the good feedback.QuoteYou presented an arbitrage opportunities model with info asymmetry, lag time for the platforms and inventory effects for the market makers. We (theorietical physicists ) love ordering, so my question is what is the main effect among these three ?Well, it really depends on the parameters of the model (beta, theta, volatilities of inovations, etc). But, if I had to guess based on a realistic microstructure model I would say that the main factor is the value of k, which is the lag update on the quotes.QuoteIn your model you consider a perfect liquid market, do you think it's possible to model arbitrage opportunities in illiquid market ?i speak as a newb but it would be interesting to extend your model with opportunity trading (I mean that all the trades hav not the time to be executed, I think that the big ones are prior but I'm not sure). Possibly. I believe that in my case this could be approached as the second lag of the trade being executed some time in the future (e.g. t+1), instead of t. This way there might be a price change between t and t+1 and this is another risk faced by the arbitrageur. Also, this is related to another idea which I've been exploring. This is to look at the time intervals for the update of the whole process as a random variable itself (e.g. duration\intensity models). So, if the market is experiencing a high activity where price and quotes are updated more frequenty, then the execution risk would be higher as the arbitrageur would have to be very quick in closing the trades. This is interesting because a high market activity is exactly the scenario for an arbitrage opportunity (as argumented in the paper). This implies some sort of balance between risk and return.For pure execution risk, there is a paper in the literature that looks at this type of risk (see Kozhan & Tham, 2009), but the underlying model is very different from what I used.QuoteFor your result I was really surprise to see that outlier italian bond which has an arbitrage opportunity of more than 900 s ! (15 min !) . Yes, but I think that it is actually a problem with the data and those were not actual tradeable quotes. One hipothesisi would be an operational error on the way the quotes were recorded. If the time stamps are recorded wrongly then this would overestimate drastically the arbitrage opportunities, which seems to be the case for that particular bond. QuoteI will read your paper again more carefully next week but at first sight congrats it is a good one Thanks for the feedback!
Last edited by msperlin
on June 11th, 2010, 10:00 pm, edited 1 time in total.