Forums  > Basics  > bid ask bounce  
     
Page 1 of 1
Display using:  

Omega


Total Posts: 426
Joined: Jul 2006
 
Posted: 2006-10-12 16:14
CAn someone tell me in a nutshell what this is exactly

FDAXHunter
Founding Member

Total Posts: 8356
Joined: Mar 2004
 
Posted: 2006-10-12 16:58
Bidhask was a forest troll in Middle Earth. He was notorious for dropping from trees onto passing travellers to squash them. He would jump from a branch, bounce off the happless people below him (Hobbits, for the most part) and then cackle evilly. The Hobbits referred to this as the "Bidhask Bounce".

Seriously though: Bid/Ask Bounce is the effect of perceived volatility because the price changed from bid to ask (or vice versa). It looks like there is movement, but because it's only inside the bid-ask spread, you can't necessarily say that it moved (there's some technicalities there but we'll leave it at that).

The term "bounce" comes from the visual image as the price curve bounces back and forth between the bid and ask.
See below an illustration of what Bid/Ask Bounce looks like:




Specifically, Bid/Ask Bounce refers to the phenomenon of vastly exaggerated volatility calculations at high frequencies. A simple volatility calculation (i.e. plain old standard deviation) treats a move of 1 tick per second the same as a 60 tick move in hour (Sqrt(3600)).
This means that if the market just trades goes up-down every two seconds, a simple standard deviation calculation will show a volatility equivalent to that of a 60 tick move. But the market hasn't even left the bid-ask spread and therefore it would be not that wrong to say that there was no movement at all. Yet you get a very large volatility for this period.

If we do simple volatility calculations at various frequencies, the result looks like this:



The high volatility readings on the left are due to bid/ask bounce.

Hope this helps.


The Figs Protocol.

Omega


Total Posts: 426
Joined: Jul 2006
 
Posted: 2006-10-12 17:28

Nice explanation FDAX ....a couple of questions though. From your explanation it seems that the bid-ask bounce refers to the fact that the "price" bounces between between a bid-ask spread over a certain period of time. eg in a minute we could have bid-ask of 98-100 and have the price bouncing around this limit within teh time frame of a minute..ie measuring the price for each second within this minute we would have high vol when instead the bid-ask prices have not budged...

If this is correct then what is the "price" measuring and what is causing it to bounce around like that


FDAXHunter
Founding Member

Total Posts: 8356
Joined: Mar 2004
 
Posted: 2006-10-12 17:38
The price you are measuring in this particular case (the one that leads us to obvserve bid/ask bounce) is the last trade.

All financial markets operate under the notion of a minimum price increment. This is known as a tick (there are actually half-ticks and quarter ticks, but thats just an oxymoron that the financial industry has managed to perpetuate).
You cannot trade in increments smaller than a tick. Depending on how volatile an instrument is compared to its minimum price increment, this can cause the market to virtually sit at the same bid and ask all day (or even several days). Fed Fund Futures are a great example of this.

A buyer looking for an immediate buy must buy at the offer and a seller looking to sell right this moment has to sell at the bid.

The bouncing around is caused by sellers and buyers hitting the bids and lifting the offers, respectively.

The Figs Protocol.

Omega


Total Posts: 426
Joined: Jul 2006
 
Posted: 2006-10-12 17:44

Nice one FDAX


saffron


Total Posts: 181
Joined: May 2006
 
Posted: 2006-10-12 19:48
I remeber seeing this strange phenomenon in spot usd/jpy during asian session a couple of years ago. Price was swinging 1 pip within bid/ask spread and for quite some time. I tried to understand what's the logic behind this but I didn't know much so I guessed it was just two bored traders playing chicken :) See who would give up first... Of-course, I was using a platform from one of them internet market-makers (I think oanda), so maybe they were just manipulating the data...

Graeme


Total Posts: 1629
Joined: Jun 2004
 
Posted: 2006-10-12 22:16

Here's an example but from the other end of the liquidity spectrum: closing prices of a highly illiquid penny stock over a year. The number of times the bid-offer pair changed was a lot lot less than the number of times the price changed.


Graeme West

ThirdEye


Total Posts: 36
Joined: Oct 2006
 
Posted: 2006-10-13 03:28
Bounce Vol adjustment has been one of the "secrets" among a few option traders for many years. Yes very important for penny stocks.

jaiman


Total Posts: 248
Joined: Oct 2004
 
Posted: 2006-10-13 17:05

what sort of techniques are used to account for the bounce? after reading some of the stuff in the RennTech thread, i'd assume that they have to be filtering this out.

 


FDAXHunter
Founding Member

Total Posts: 8356
Joined: Mar 2004
 
Posted: 2006-10-13 17:17
And we would tell you this why? Like ThirdEye said... it's a secret. Cool

By the way: There's no double-'N' in Renaissance.

The Figs Protocol.

pj


Total Posts: 3353
Joined: Jun 2004
 
Posted: 2006-10-13 17:53
Renntec and Rennttec are the secret evil
brothers of Rentec.

BTW you should definitely try and take some sort of holiday... good for the soul. (Donal Gallagher)

jaiman


Total Posts: 248
Joined: Oct 2004
 
Posted: 2006-10-13 18:11

to me every time i hear rentech i think of these guys first. and it seems to influence my spelling.

i thought it was a "secret" like a double-quote-even-newbs-know-it sort of secret.

How bout this: is it assumed that prices bouncing around a mid point have no value, that only the deviation from that mid point is of interest?


FDAXHunter
Founding Member

Total Posts: 8356
Joined: Mar 2004
 
Posted: 2006-10-13 18:52
jaiman: is it assumed that prices bouncing around a mid point have no value, that only the deviation from that mid point is of interest?

That's a contradiction in itself. If prices bouncing around a point, then by definition they also deviate from that value.

Anyway, you're one of Cockmott's pals, so I ain't talking to you out of sheer principle...

The Figs Protocol.

jaiman


Total Posts: 248
Joined: Oct 2004
 
Posted: 2006-10-13 19:39

That's a contradiction in itself. If prices bouncing around a point, then by definition they also deviate from that value.

Ok, i was unclear. when i said price i meant mid point, so last price bouncing between +1 pip and -1 pip of the same mid point would have no real information value(other then buyers and seller are not overwhelming the market makers, possibly). but if the mid point itself starts moving, that seems to be what people would be interested in. So my question should be how do you differentiate between midprice movement vs. bid-ask bounce in a string of last trade prices?

My guess would be that you need to have bid and ask prices and probably market depth to do it, but i'd be interested in other ways.


Anyway, you're one of Cockmott's pals, so I ain't talking to you out of sheer principle...

ummm based on what? because i considered doing the cqf? or do you have me confused for someone else?


FDAXHunter
Founding Member

Total Posts: 8356
Joined: Mar 2004
 
Posted: 2006-10-13 21:11
I remember you. I even remember your red icon. But maybe I shouldn't hold that against you. Or maybe I should... ah I hate my conscience. Life would be so much easier without one.

Best to use bid/ask data. And the nice thing about that technique is that if you don't have bid/ask data, then you don't need to worry about high-frequency dynamics anyway. Nice how that problem solves itself, isn't it?

Of course, you could also try to estimate the mid-price from last trade data. This is also relatively trivial. A simple trick is to simply take the last trade series and then for each tick take a guess on the bid-ask location from the tick it's two neighboring ticks (left and right) and then calculated the supposed mid-price.
Note that we assume a constant bid-ask spread for this, which will work for very liquid markets but not for penny stocks.
This will underestimate the actual volatility and provide a lower bound for the vol.
You can derive a theoretical adjustment factor (telling you how much higher the actual vol should be), but depending on the market you are looking at that factor might be meaningless and the lower bound will actually be close to the real number).

And then of course, calculating a volatility on a tick-by-tick basis is really unstable... and you have to be wondering whether volatility is actually even a sensible measure to characterize the behavior and whether or not a different kind of model is called for.... but that's another story. Wink

Give my regards to the lemonade crowd. Party

The Figs Protocol.

jaiman


Total Posts: 248
Joined: Oct 2004
 
Posted: 2006-10-13 21:32

thanks FDAX. 

my thought would be, if i were building a really short term model, i'd want an idea of short term volatility and drift and start from there.

but just so you know, since i found this place, my post count on the other spot is negligible and mainly resticted to asking simple math questions that i don't want to bother this forum with. If i offended, there or here, all i can offer is an apology and offer a beer if you're ever in Toronto.

Previous Thread :: Next Thread 
Page 1 of 1