Average NBBO size database

TraDaToR

TraDaToR

Hello,

Does someone know a website where you can find the average NBBO size of a given security( in ticks or cents )? Like you type AAPL and you get the average spread...I am talking about displayed liquidity.

It would be great to get the average spread in the hour after the open and then during the rest of the session, but I am perhaps asking for too much here...LOL

Thanks.
 
TraDaToR said:
Hello,

Does someone know a website where you can find the average NBBO size of a given security
More...

No website that I know of but you can pretty accurately derive the average spread from the day's one minute bars (ohlc or really all you need are the 1min closes) which are freely available from yahoo or google finance.

The insight is that the variance atributable to movement in the mid price should scale lineraly with sampling interval (sigma with sqrt(T)) while the portion of the variance atributable to spread should stay constant. I don't remember the exact extration method but you should be able to google it, it is pretty well known. IIRC it was a maximum-likelihood fit, easy enough to do in excel.
 
I completely disagree with your assumptions. You can't reliably estimate the spread between bid and offer from looking at 1min compressed time series and thinking that the price volatility in a 1 minute window has any bearing on the spread during that time. I can disprove that rather quickly actually: take two securities, or actually the same, look at different 1minute bars and the distance between high and low. You can find many examples where the distance matches but spreads are different. Maybe not in the same day all too often but definitely even within a week.

Kevin Schmit said:
No website that I know of but you can pretty accurately derive the average spread from the day's one minute bars (ohlc or really all you need are the 1min closes) which are freely available from yahoo or google finance.

The insight is that the variance atributable to movement in the mid price should scale lineraly with sampling interval (sigma with sqrt(T)) while the portion of the variance atributable to spread should stay constant. I don't remember the exact extration method but you should be able to google it, it is pretty well known. IIRC it was a maximum-likelihood fit, easy enough to do in excel.
More...
 
Zzzz1 said:
I completely disagree with your assumptions. You can't reliably estimate the spread between bid and offer from looking at 1min compressed time series and thinking that the price volatility in a 1 minute window has any bearing on the spread during that time. I can disprove that rather quickly actually: take two securities, or actually the same, look at different 1minute bars and the distance between high and low. You can find many examples where the distance matches but spreads are different. Maybe not in the same day all too often but definitely even within a week.
More...
I don't think you understood my post. That the distance between high and low matches between two stocks with dissimilar spreads is not relevant. The relevant metric is how that distance scales with sampling frequencies (for example scales among 1,2,4,16,32... minute bars) for the two stocks.
 
Last edited:
OK. I will read this. I hate academic papers and formulas, but sometimes you have to make an effort. I understand the logic behind for sure.Thanks.
 
That's an even stranger assertion. What makes you think it would scale at all? So the spread at an observed 1minute bar was on avg 0.2 pips. What does that tell you about the 1 hour or 24 hour bar? Are you sure you are not confused on this?

By the way, let's stay with spreads, I feel you are trying to wiggle your way out of this by trying to explain how volatility scales with time. That was never the point. You suggested bid offer spreads can be estimated as a function of time and volatility. That volatility scales with time and why, I have learned in my first or second class session in stoch Calc.

Kevin Schmit said:
I don't think you understood my post. That the distance between high and low matches between two stocks with dissimilar spreads is not relevant. The relevant metric is how that distance scales with sampling frequencies (for example scales among 1,2,4,16,32... minute bars) for the two stocks.
More...
 
Last edited:
One of the worst papers I have ever come across. Not only does it make outlandish and unproven assumptions in the abstract already but the conclusions do not flow at all from the main arguments and hypotheses, stated in the main body.

Not trying to shoot you down but your claim does not even make sense upon cursory thought. And this "research" you cited fails to prove your (and its own) stated points even in the conclusion section.

In case we are totally talking about different topics and you might be talking about volatility, only, then it probably does not address the OP either. Volatility scales with time on average but it is a very bad estimator for specific volatility clusters, especially around the open or close of trading sessions.


Kevin Schmit said:
To save myself further explanation, here attached is a paper describing how to estimate spread from high/low data. The paper deals with daily bars, but is easily extensible to intraday bars.
More...
 
Last edited:
Just like to confirm, are you referring to the bid-offer spread here when you say spread? Or were you talking about volatility such as high low spread?

TraDaToR said:
Hello,

Does someone know a website where you can find the average NBBO size of a given security( in ticks or cents )? Like you type AAPL and you get the average spread...I am talking about displayed liquidity.

It would be great to get the average spread in the hour after the open and then during the rest of the session, but I am perhaps asking for too much here...LOL

Thanks.
More...
 
You must log in or register to reply here.
Back
Top