Forums  > University  > Theoretical papers on the limit order book  
     
Page 1 of 1
Display using:  

trialanderror


Total Posts: 31
Joined: Feb 2019
 
Posted: 2019-10-09 19:41
Hello,

Since my last post here in the university section, and after discussion with my professor, the aim of my thesis has changed somewhat. I am now looking at the limit order book for 30 stocks and getting descriptive statistics from there. Currently, I am replicating this paper: https://www.jstor.org/stable/2329330?seq=8#metadata_info_tab_contents

However, I do not feel like part II. A. where the authors visualize the order book adds much to the analysis. Does anyone have a suggestion on what I could do instead, maybe testing some theory or similar, of the simple averaging visualization that might add more to the paper? Maybe some interesting papers I could read to get some inspiration?

Regards,
trialanderror

Nous promettons selon nos espérances, et nous tenons selon nos craintes.

nikol


Total Posts: 794
Joined: Jun 2005
 
Posted: 2019-10-09 21:38
could you share the papers?

trialanderror


Total Posts: 31
Joined: Feb 2019
 
Posted: 2019-10-10 06:53
Attached File: LimitOrderBookInvestigation.pdf

Nous promettons selon nos espérances, et nous tenons selon nos craintes.

bullero


Total Posts: 57
Joined: Feb 2018
 
Posted: 2019-10-10 08:32
https://arxiv.org/pdf/1809.08060.pdf

ronin


Total Posts: 483
Joined: May 2006
 
Posted: 2019-10-10 10:59
> I do not feel like part II. A. where the authors visualize the order book adds much to the analysis

Actually, it does.

> a suggestion on what I could do instead, maybe testing some theory or similar, of the simple averaging visualization that might add more to the paper?

When all is said and done, the basic questions to do with the order book are:

1. Can it predict the future price trajectory
2. Can it tell you when an order will be filled
3. Can it tell you what the impact will be of filling certain quantity.

You can't do much about 3, but either 1 or 2 should be perfectly feasible. And, as an added bous, you will work out what the point of cumulative order book graphs is.


"There is a SIX am?" -- Arthur

nikol


Total Posts: 794
Joined: Jun 2005
 
Posted: 2019-10-10 15:31
+
4 how your predictions will change if you know that the last limit order arrived with size of V and price P.
5. and your maximum order size << V

If all above is answered - determine parameter of the order by maximizing PnL distribution (mean and sharpe ratio?)

trialanderror


Total Posts: 31
Joined: Feb 2019
 
Posted: 2019-10-12 09:32
@bullero

Thanks. I'll have a look at it.

@ronin & @nikol

Thanks to you too. I'll dare to take a stab at the reasoning based on @ronin point 1. and I guess 3. as well. Would the slope of the order book then point to how far an order of a certain size, in general, will be able to push the price up/down, thus predicting price trajectory and at the same time giving the impact of filling a certain quantity? However, I must say that points 1. and 2. feel somewhat connected. If I know the impact of filling a certain quantity, does that not also imply that I know something about the future price trajectory?

Also another question. The authors say "(...) the shape of the order book may reflect the correlation in the value of the security to various bidders on the same side of the market and the shading of bids compared to the underlying reservation values. Alternatively, if the slope of the book arises from informational asymmetries, then our findings suggest that the marginal information content of trades is decreasing with trade size.".

Knowing that the findings consist of:
1) the bid-ask spread is larger than twice the difference between the adjacent quotes on each side of the book,
2) provision of liquidity at the bid-ask quotes is only a small portion of the overall depth of the order book and
3) the price schedule is steeper at the quotes than away from the quotes

What does the authors' statement actually mean in light of the findings?

Nous promettons selon nos espérances, et nous tenons selon nos craintes.

ronin


Total Posts: 483
Joined: May 2006
 
Posted: 2019-10-12 10:24
My 2 is a weak version of my 1. They are linked, of course, but 2 should be easier than 1. For 3 you need deeper information - you need to know who is posting what and filling what.

As for your 1,2,3, isn't that what you should be working out for your project?

To start you off, you seem to be saying that the bid-ask is 2-3 ticks wide. The mms seem to think that 1 tick spread isn't worth it. There are some crazy mofos putting small quantities at 2-3 ticks to see if anybody bites, but the mms are staying at 4 ticks and above. Why?

Then, moving away from the touch. Imagine you have some large quantity Q to execute. You have to cut it up in small chunks, and get those chunks filled at a steady rate. How would you place them given the shape of the order book?

Then you kind-of know how people's behaviour affects the order book. Then you have to invert that. Given the shape of the order book, what does that tell you about people's behaviour?

That shuld be plenty for a thesis, I would think.

> The authors say "(...) the shape of the order book may reflect the correlation in the value of the security to various bidders on the same side of the market and the shading of bids compared to the underlying reservation values. Alternatively, if the slope of the book arises from informational asymmetries, then our findings suggest that the marginal information content of trades is decreasing with trade size.".

I wouldn't go down that route. That sort of thinking is based on some models which take the efficient market hypothesis too far. In the real world, there are buyers and sellers, and nobody knows too much. This notion of price==information is unrealistic even on much longer time scales, let alone intraday.

"There is a SIX am?" -- Arthur

trialanderror


Total Posts: 31
Joined: Feb 2019
 
Posted: 2019-10-14 18:22
@ronin

Your 1 and 2: Alright I understand.

My 1,2,3: I'm actually replicating a paper by previous authors who have done the same study in 1995, I'm just doing it now and for a different market. That's where those results come from.

Your question 1: The mms are staying above because they don't make enough of a margin at 2-3 ticks? I would also think that this has something to do with the vol of the market in general. Given a higher vol, I guess the mms would widen their spreads (I also presume you got this from my point 2 and 3)

Your question 2: Given the shape of the order book I would imagine it being hard to fill at a steady rate far away from the quotes since volume is much higher away from them. Therefore, I would say you'd need to distribute the chunks randomly close to the quotes to get a steady fill rate, how close would depend on vol I guess. I remember from my last post that there are models for this(?)

Your question 3: Can't come up with anything substantial here except for there being, as you say, some crazy mofos as well as probably more sophisticated mms in the market. No insights for me there so if you'd like to fill in feel free

The authors say...: That's what I read in the thesis I'm replicating. I asked because I didn't understand what that meant, not really my own opinion or anything

Nous promettons selon nos espérances, et nous tenons selon nos craintes.
Previous Thread :: Next Thread 
Page 1 of 1