Here's a thread of some guy posting regular 24h videos of the MtGox spread!
Some guy here. It's been interesting making these vids so if you like 'em or have any suggestions it's always good to hear from viewers. 
Topic: The Form of a "perfect" bid/ask Curve (Read 5255 times)
Here's a thread of some guy posting regular 24h videos of the MtGox spread!
I think the position of walls is often related to Mt Gox fees. Max 1% difference to not lose money, min .5%.
but in my experience no one ever really knows what they are doing anyway so lets go for it.
+1
I think the next step should be including fat tail risk, otherwise known as kurtosis risk. The normal curve does not describe the bitcoin market well due to the small market cap and obvious presence of "dragon players", as someone described our manipulators. I would say a large proportion of the variance is due to the actions of these traders. They use FUD to incite emotional reasoning amongst traders with smaller positions, thus increasing the proportion of variance found in the tails.
I am imagining a three dimensional scatter plot of skew, kurtosis, and variance... with the variance coming out of the page (since it cannot be negative) and also color coded to facilitate comprehension. This will allow us to plot values that describe state of an ideal market in which all information regarding sentiment is present in the bid/ask walls, and all traders have perfect understanding of this info. It should look something like this:
Then it will be time for maths and gathering empirical evidence to figure out why the model does not fit reality (e.g. spekulatius' theory about the effect of trends, fake walls, etc). I foresee Fourier and Monte Carlo in our future. The data will probably end up being too noisy and this type of analysis has probably been published somewhere already, but in my experience no one ever really knows what they are doing anyway so lets go for it.
I am imagining a three dimensional scatter plot of skew, kurtosis, and variance... with the variance coming out of the page (since it cannot be negative) and also color coded to facilitate comprehension. This will allow us to plot values that describe state of an ideal market in which all information regarding sentiment is present in the bid/ask walls, and all traders have perfect understanding of this info. It should look something like this:
Then it will be time for maths and gathering empirical evidence to figure out why the model does not fit reality (e.g. spekulatius' theory about the effect of trends, fake walls, etc). I foresee Fourier and Monte Carlo in our future. The data will probably end up being too noisy and this type of analysis has probably been published somewhere already, but in my experience no one ever really knows what they are doing anyway so lets go for it.
Ok, so I think that, rather than volume, a better way of thinking about spekulatius' V- and U-shaped neutral curves is as "high certainty" and "low certainty" sentiment regarding the current price. This is seen if we change the variance of the neutral normal curve seen in my original figure. Changing the variance of the neutral curve then drags the bull and bear curves along, as seen below:
When there is high certainty around the current price, traders will place bids and asks very near the price. When there is low certainty about the current price traders are more likely to place bids farther out. In either case, which direction they place them depends on if they are neutral, bulls, or bears leading to the black, red-blue, and blue-red U-shape and V-shapes. Low neutral (or sideways, as spekulatius puts it) volume is associated with high certainty, while vice versa for high volume... it leads to uncertainty and unwillingness to place large bids/asks near the current price.
Once again, the form of my charts is due to sentiment, not movement. They are predictive, not reactive. The result of movement is as spekulatius described, but I think including movement so soon is complicating the model before it is ready. During a movement, uptrends naturally generate bearish sentiment, while downtrends generate a bullish sentiment. So the system is self-correcting until other factors come into play (external info, new money, manipulators, etc), which eventually result in uptrends and downtrends.
Make sense?
When there is high certainty around the current price, traders will place bids and asks very near the price. When there is low certainty about the current price traders are more likely to place bids farther out. In either case, which direction they place them depends on if they are neutral, bulls, or bears leading to the black, red-blue, and blue-red U-shape and V-shapes. Low neutral (or sideways, as spekulatius puts it) volume is associated with high certainty, while vice versa for high volume... it leads to uncertainty and unwillingness to place large bids/asks near the current price.
Once again, the form of my charts is due to sentiment, not movement. They are predictive, not reactive. The result of movement is as spekulatius described, but I think including movement so soon is complicating the model before it is ready. During a movement, uptrends naturally generate bearish sentiment, while downtrends generate a bullish sentiment. So the system is self-correcting until other factors come into play (external info, new money, manipulators, etc), which eventually result in uptrends and downtrends.
Make sense?
It isn't merely that existing bids were were moved closer to center. Rather, money that wasn't on the books has appeared on the books.
Quote
3 trends: down, neutral, up (psychological)
2 signals: bull, bear (technically suggested using probabillity analysis)
6 data: Bidvolume, Askvolume, Bidplacement(price), Askplacement(price), historic volume, historic volatility (raw information)
2 signals: bull, bear (technically suggested using probabillity analysis)
6 data: Bidvolume, Askvolume, Bidplacement(price), Askplacement(price), historic volume, historic volatility (raw information)
I would suggest to switch the terms "psychological" and "technical", as "trends" are usually referred to in chart technical contexts and psychological sounds more like bullish or bearish sentiments (--> moods). Also a "signal" is to me an indicator to predict future market movements, rather then a proof of past or present movements. I think bullish or bearish sentiment meets it pretty good and we should stick to that term.
In the provided picture below are some interesting developments visible in comparison to the prior posted one. I think what we see here is not the norm and rather exceptional (although market manipulation to that degree can be witnessed more often with bitcoins).
!About 1 hr ago the mighty BIDwalls were moved closer to the center!
To speak in your terms: Due to recent (about 1hr ago) BIDPLACEMENTS close to the center in contrast to more off-center ASKPLACEMENTS, the prior bearish sentiment (totally agree with you, that it was bearish) visible in UV-shapes, -->see older picture, was not confirmed and a more VU-like shape formed while volume dropped. In consequence the price is driven upwards and we now have an uptrend --> VU shape transists to UV shape and continues with the trend.
In other words:
Because the manipulator
placed its unsurmountable BIDWALLS so close to the current price, that it is driven up; although the overall sentiment is bearish. We have an uptrend and will have in consequence a consistent UV shape, which I state is typical for uptrends (!not bullish sentiments, but up TRENDS, so to speak up-movements!). Because the manipulation to the market is fresh, the current shape (picture above) resembles more the one of V (due to low volume), but when the trend continues, I am shure, the nature of the UV shape will show.
Oh yea, definitely not practical for predicting future bitcoin price. I don't trust the depth chart at all.
i find the depth data should be considered in a market where a large player can consume a wall (for instance, the one that WAS at $7.20 not so long ago) and cause an avalanche/roller coaster all by themself. these kinds of movements often belie other indicators.
what exactly is "depth"? .. i mean out of what is it calculated? historic volatility and historic volume?
what effects do you think have a high or low depth for my last example?
.. that was:
neutral trend, low volume, a UV-Wave (which i suggest is a bearish signal) slowly appears
@ bitcoinbitcoin123:
- I think you confused bearish with bullish sentiments in your picture
- The alteration with using parabolas instead of triangular functions is IMO a good idea and a more realistic approach
- The accumulated volumes likewise (good job!)
@ gewure:
Ahh, O.K., makes much more sense this way.
I hope I can contribute to even further explore the matter, although I have to admit not to have any emperical data at hand as well.
Therefore the following is only product of my own observations and reflections. Any comment or support by emperical data would be
welcome:
- In times of bullish or bearish sentiments (clear trends indicated in either direction, visible to the mayority of traders)
the two "price movement spread shapes" show, because of their shape Ill call them "UV"- and "VU"-shapes:
- In times of uncertainty or contrasting estimates by equal amounts of traders wheter the price will go up or downwards, the other two
remaining shapes "U"- and "V" shapes appear. The U-shape may be typical for times of sideways trends or neutral triangles, when
traders wait for clear signals before they feel save to place their bets, in consequence there will be low volume on both sides of the
spread and little price movement. The U-Shape can therefore be seen neutral
- In times of contrasting estimations caused by numerous signals pointing in different directions, e.g. during a large doji
(http://en.wikipedia.org/wiki/Doji) or anomalities during trends I would expect a V-shape to form, because high volume on both sides
of the spread keep the spread small and price movement little. As long as a V-shape consists, it can be seen to be neutral as well.
Actually, someone must have asked those questions before. Im sure the answers are out there already. Maybe someone can provide some helpful links?
Please comment, thx
- I think you confused bearish with bullish sentiments in your picture
- The alteration with using parabolas instead of triangular functions is IMO a good idea and a more realistic approach
- The accumulated volumes likewise (good job!)
@ gewure:
Ahh, O.K., makes much more sense this way.
I hope I can contribute to even further explore the matter, although I have to admit not to have any emperical data at hand as well.
Therefore the following is only product of my own observations and reflections. Any comment or support by emperical data would be
welcome:
- In times of bullish or bearish sentiments (clear trends indicated in either direction, visible to the mayority of traders)
the two "price movement spread shapes" show, because of their shape Ill call them "UV"- and "VU"-shapes:
- In times of uncertainty or contrasting estimates by equal amounts of traders wheter the price will go up or downwards, the other two
remaining shapes "U"- and "V" shapes appear. The U-shape may be typical for times of sideways trends or neutral triangles, when
traders wait for clear signals before they feel save to place their bets, in consequence there will be low volume on both sides of the
spread and little price movement. The U-Shape can therefore be seen neutral
- In times of contrasting estimations caused by numerous signals pointing in different directions, e.g. during a large doji
(http://en.wikipedia.org/wiki/Doji) or anomalities during trends I would expect a V-shape to form, because high volume on both sides
of the spread keep the spread small and price movement little. As long as a V-shape consists, it can be seen to be neutral as well.
Actually, someone must have asked those questions before. Im sure the answers are out there already. Maybe someone can provide some helpful links?
Please comment, thx
What do you think is the difference between your U and V neutral curves and the bear-bull, bull-bear balanced (or mixed) curves?
thanks for all this observations!
i still think your interpretation of the UV-wave as bullish and the VU-wave as bearish is partly wrong. it may be observed that the waves form such a look while a bullish/bearish trend is going on, but that does not necessarily mean that the bullish/bearish trends always form such a shape. in fact they dont have to at all.
i would recommend that we first focus on the research of the future-related implications the various wave-figures give us rather than studying the details of trends..(which i think are more complex).
e.g. UV-wave forming out of a U- or V-formed market suggests selling pressure and uncertainty on buying and therefore implies a psychological mood on the market regarded as bearish not bullish.
so to take your trend model:
we now have:
3 trends: down, neutral, up (psychological)
2 signals: bull, bear (technically suggested using probabillity analysis)
6 data: Bidvolume, Askvolume, Bidplacement(price), Askplacement(price), historic volume, historic volatility (raw information)
i would now suggest that we take a neutral trend and given data to form our pychological explanation model on marketmoves.
we have already done many interpretations.. great progress inside here
Quote
My recent observations from the footage recorded by me (see 3 posts up) suggest, that my prior estimation was wrong and V shapes actually form with LOW volume, while U shapes require high volumes as it lays in their nature, so to speak, to bumb all minor walls out of their way while spiking in all directions. In that way they rounden every V- to a U shape.
Yeah, you can see it in the vid a bit (not much movement that day). Watched last 2 days but records got fucked up. So apart from the one vid, I can assure that with high vol (like the last 2 days) U shpes form.
Today with low volume its taking a more V-like shape actually. More than that, even a (possibly) bearish sentiment induced UV-shape is forming (watch below or on mtgoxlive.com):
Cheers
So did you record that? I was at work so I didnt get to see the walls get eaten.
Alrigth, gotcha.
To answer your question: I think the difference is, that the neutral (U and V)-curves only occur during sideways trends, while VU,UV-curves ideally form during up or down trends.
As gewure pointed out, UV and VU curves may also form during sideways trends - and Id like to add, that it probably requieres a period of relative quietness for them to form. A period with little price movement before a clear trend is visible, otherwise the steep flank couldnt form. (I use the term "probably", because I lack empirical data to be sure here)
The combination of U and U flanks or V and V flanks of UV and VU shapes as it comes to mind in my picture is only a hypothetical game and has no relevance for reality, because there will only be UV or VU shapes at a time and if their flanks approximate, it will either result in V or U shapes. (I hope everyone can follow and picture the dance of shapes here;)
My recent observations from the footage recorded by me (see 3 posts up) suggest, that my prior estimation was wrong and V shapes actually form with LOW volume, while U shapes require high volumes as it lays in their nature, so to speak, to bumb all minor walls out of their way while spiking in all directions. In that way they rounden every V- to a U shape.
Summary:
U-shapes: High volume, sideways trend
V-shapes: Low volume, sideways trend
UV-shapes: bullish sentiment, [high or medium volume, UP TREND] or [extended low volume, extended SIDEWAYS trend]
VU-shapes: bearish sentiment, [high or medium volume, DOWN TREND] or [extended low volume, extended SIDEWAYS trend]
To answer your question: I think the difference is, that the neutral (U and V)-curves only occur during sideways trends, while VU,UV-curves ideally form during up or down trends.
As gewure pointed out, UV and VU curves may also form during sideways trends - and Id like to add, that it probably requieres a period of relative quietness for them to form. A period with little price movement before a clear trend is visible, otherwise the steep flank couldnt form. (I use the term "probably", because I lack empirical data to be sure here)
The combination of U and U flanks or V and V flanks of UV and VU shapes as it comes to mind in my picture is only a hypothetical game and has no relevance for reality, because there will only be UV or VU shapes at a time and if their flanks approximate, it will either result in V or U shapes. (I hope everyone can follow and picture the dance of shapes here;)
My recent observations from the footage recorded by me (see 3 posts up) suggest, that my prior estimation was wrong and V shapes actually form with LOW volume, while U shapes require high volumes as it lays in their nature, so to speak, to bumb all minor walls out of their way while spiking in all directions. In that way they rounden every V- to a U shape.
Summary:
U-shapes: High volume, sideways trend
V-shapes: Low volume, sideways trend
UV-shapes: bullish sentiment, [high or medium volume, UP TREND] or [extended low volume, extended SIDEWAYS trend]
VU-shapes: bearish sentiment, [high or medium volume, DOWN TREND] or [extended low volume, extended SIDEWAYS trend]
Oh yea, definitely not practical for predicting future bitcoin price. I don't trust the depth chart at all.
i find the depth data should be considered in a market where a large player can consume a wall (for instance, the one that WAS at $7.20 not so long ago) and cause an avalanche/roller coaster all by themself. these kinds of movements often belie other indicators.
I think you may have misunderstood. By "mixed" curve I mean to combine the blue and red into one curve. For example, your V(neutral) curve appears very similar to the V-side of the bearish sentiment combined with the V side of bullish sentiment.
- In times of bullish or bearish sentiments (clear trends indicated in either direction, visible to the mayority of traders)
the two "price movement spread shapes" show, because of their shape Ill call them "UV"- and "VU"-shapes:
- In times of uncertainty or contrasting estimates by equal amounts of traders wheter the price will go up or downwards, the other two
remaining shapes "U"- and "V" shapes appear. The U-shape may be typical for times of sideways trends or neutral triangles, when
traders wait for clear signals before they feel save to place their bets, in consequence there will be low volume on both sides of the
spread and little price movement. The U-Shape can therefore be seen neutral
- In times of contrasting estimations caused by numerous signals pointing in different directions, e.g. during a large doji
(http://en.wikipedia.org/wiki/Doji) or anomalities during trends I would expect a V-shape to form, because [glow=yellow]high volume on both sides
of the spread keep the spread small and price movement little[/glow=yellow]. As long as a V-shape consists, it can be seen to be neutral as well.
Actually, someone must have asked those questions before. Im sure the answers are out there already. Maybe someone can provide some helpful links?
Please comment, thx
remaining shapes "U"- and "V" shapes appear. The U-shape may be typical for times of sideways trends or neutral triangles, when
traders wait for clear signals before they feel save to place their bets, in consequence there will be low volume on both sides of the
spread and little price movement. The U-Shape can therefore be seen neutral
- In times of contrasting estimations caused by numerous signals pointing in different directions, e.g. during a large doji
(http://en.wikipedia.org/wiki/Doji) or anomalities during trends I would expect a V-shape to form, because [glow=yellow]high volume on both sides
of the spread keep the spread small and price movement little[/glow=yellow]. As long as a V-shape consists, it can be seen to be neutral as well.
Actually, someone must have asked those questions before. Im sure the answers are out there already. Maybe someone can provide some helpful links?
Please comment, thx
I ran Camstudio last night and recorded roughly 7 hrs, 40 mins of the spread. Will record more in the days to come.
http://vimeo.com/35204706
@ bitcoinbitcoin123:
- I think you confused bearish with bullish sentiments in your picture
- The alteration with using parabolas instead of triangular functions is IMO a good idea and a more realistic approach
- The accumulated volumes likewise (good job!)
@ gewure:
Ahh, O.K., makes much more sense this way.
I hope I can contribute to even further explore the matter, although I have to admit not to have any emperical data at hand as well.
Therefore the following is only product of my own observations and reflections. Any comment or support by emperical data would be
welcome:
- In times of bullish or bearish sentiments (clear trends indicated in either direction, visible to the mayority of traders)
the two "price movement spread shapes" show, because of their shape Ill call them "UV"- and "VU"-shapes:
- In times of uncertainty or contrasting estimates by equal amounts of traders wheter the price will go up or downwards, the other two
remaining shapes "U"- and "V" shapes appear. The U-shape may be typical for times of sideways trends or neutral triangles, when
traders wait for clear signals before they feel save to place their bets, in consequence there will be low volume on both sides of the
spread and little price movement. The U-Shape can therefore be seen neutral
- In times of contrasting estimations caused by numerous signals pointing in different directions, e.g. during a large doji
(http://en.wikipedia.org/wiki/Doji) or anomalities during trends I would expect a V-shape to form, because high volume on both sides
of the spread keep the spread small and price movement little. As long as a V-shape consists, it can be seen to be neutral as well.
Actually, someone must have asked those questions before. Im sure the answers are out there already. Maybe someone can provide some helpful links?
Please comment, thx
- I think you confused bearish with bullish sentiments in your picture
- The alteration with using parabolas instead of triangular functions is IMO a good idea and a more realistic approach
- The accumulated volumes likewise (good job!)
@ gewure:
Ahh, O.K., makes much more sense this way.
I hope I can contribute to even further explore the matter, although I have to admit not to have any emperical data at hand as well.
Therefore the following is only product of my own observations and reflections. Any comment or support by emperical data would be
welcome:
- In times of bullish or bearish sentiments (clear trends indicated in either direction, visible to the mayority of traders)
the two "price movement spread shapes" show, because of their shape Ill call them "UV"- and "VU"-shapes:
- In times of uncertainty or contrasting estimates by equal amounts of traders wheter the price will go up or downwards, the other two
remaining shapes "U"- and "V" shapes appear. The U-shape may be typical for times of sideways trends or neutral triangles, when
traders wait for clear signals before they feel save to place their bets, in consequence there will be low volume on both sides of the
spread and little price movement. The U-Shape can therefore be seen neutral
- In times of contrasting estimations caused by numerous signals pointing in different directions, e.g. during a large doji
(http://en.wikipedia.org/wiki/Doji) or anomalities during trends I would expect a V-shape to form, because high volume on both sides
of the spread keep the spread small and price movement little. As long as a V-shape consists, it can be seen to be neutral as well.
Actually, someone must have asked those questions before. Im sure the answers are out there already. Maybe someone can provide some helpful links?
Please comment, thx
What do you think is the difference between your U and V neutral curves and the bear-bull, bull-bear balanced (or mixed) curves?
if you don't mind i will use your figures to discuss this matter with some fellow students at my university and look what they think about it.
Go for it. It is very theoretical right now, but if someone had enough money to map the true depth on both sides it may be useful, at least for a short time.
@bitcoinbitcoin113
i especially like your charts. makes more sense if you also include probabillity and volume into it, very good idea!
now the question is how to scale this model, to get usefull interpretations on future-price-movement-probabillities out of the mtgox bid/ask table.. (if at all)
there are a lot more systemvariables that need to be included into the modell, in order to have good interpretations.
if you have time, take close looks at the mtgox table, and write in here, if you observ any relations to our theory. i will do so.
thanks
i especially like your charts. makes more sense if you also include probabillity and volume into it, very good idea!
now the question is how to scale this model, to get usefull interpretations on future-price-movement-probabillities out of the mtgox bid/ask table.. (if at all)
there are a lot more systemvariables that need to be included into the modell, in order to have good interpretations.
if you have time, take close looks at the mtgox table, and write in here, if you observ any relations to our theory. i will do so.
thanks
Jump to: