Topic: Analysis - page 263. (Read 941563 times)
Log scale chart of complete bitstamp USD history.
Why might the latest pennant be such a different beast to the previous ones?
Why might the latest pennant be such a different beast to the previous ones?
Because you can have independent, i.e. fundamental reasons to believe the long term log uptrend for BTC will continue to hold, but that assumption is should probably not be based on observing a pattern 4 times and concluding it will hold a 5th time. That's simply not enough of a sample size.
...
Just to note, I'm not concluding anything.
I would think that a sample size of 4 is statistically very significant when the full set is 4 or 5.
Log scale chart of complete bitstamp USD history.
...
Why might the latest pennant be such a different beast to the previous ones?
...
Why might the latest pennant be such a different beast to the previous ones?
well, we're currently resting on the bottom support in your figure. wouldn't this model be falsified if we broke below?
I guess so, falsification being proportional to the break-below amount.
It would also be falsified if it broke-above too much.
falsification tends to be a binary option. the pennant shape relies on a moving or flat support which cannot be violated except by known outliers (e.g. 135 BTCUSD data from MtGox last month). if this support is broken, then the pennant model is falsified.
in other words, the model you've presented suggests that this is a market bottom. however, if we move below the last low, this model will be falsified.
So far as I understand, pennants are pairs of trendlines allowing one to form a binary decision.
Admittedly the fact they are drawn with a hard boundary does not suggest this.
The call as to when the price breaks outside of the pennant is left to the person analysing.
The majority of pennants that get charted tend to have a few excursions in them as the chartist creates them:
If one were to sample the opinion of many technical analysts, their call on "falsification" will be a fuzzy spread around the pennant.
Log scale chart of complete bitstamp USD history.
Why might the latest pennant be such a different beast to the previous ones?
Why might the latest pennant be such a different beast to the previous ones?
Because you can have independent, i.e. fundamental reasons to believe the long term log uptrend for BTC will continue to hold, but that assumption is should probably not be based on observing a pattern 4 times and concluding it will hold a 5th time. That's simply not enough of a sample size.
@arepo: yes, that's the book. The results are due to some author named Bulkowski, but the way it looks it isn't really an academic study. Still, the Kirkpatrick/Dahlquist is a great book, in case you're wondering.
Log scale chart of complete bitstamp USD history.
...
Why might the latest pennant be such a different beast to the previous ones?
...
Why might the latest pennant be such a different beast to the previous ones?
well, we're currently resting on the bottom support in your figure. wouldn't this model be falsified if we broke below?
I guess so, falsification being proportional to the break-below amount.
It would also be falsified if it broke-above too much.
falsification tends to be a binary option. the pennant shape relies on a moving or flat support which cannot be violated except by known outliers (e.g. 135 BTCUSD data from MtGox last month). if this support is broken, then the pennant model is falsified.
in other words, the model you've presented suggests that this is a market bottom. however, if we move below the last low, this model will be falsified.
Technically, I disagree. You can see the April triangle spent some time below the lower line, but this did not invalidate the data. I think it would need to break down and then the triangle would have to become resistance before this was invalidated.
Log scale chart of complete bitstamp USD history.
...
Why might the latest pennant be such a different beast to the previous ones?
...
Why might the latest pennant be such a different beast to the previous ones?
well, we're currently resting on the bottom support in your figure. wouldn't this model be falsified if we broke below?
I guess so, falsification being proportional to the break-below amount.
It would also be falsified if it broke-above too much.
falsification tends to be a binary option. the pennant shape relies on a moving or flat support which cannot be violated except by known outliers (e.g. 135 BTCUSD data from MtGox last month). if this support is broken, then the pennant model is falsified.
in other words, the model you've presented suggests that this is a market bottom. however, if we move below the last low, this model will be falsified.
Log scale chart of complete bitstamp USD history.
...
Why might the latest pennant be such a different beast to the previous ones?
...
Why might the latest pennant be such a different beast to the previous ones?
well, we're currently resting on the bottom support in your figure. wouldn't this model be falsified if we broke below?
I guess so, falsification being proportional to the break-below amount.
It would also be falsified if it broke-above too much.
it's interesting to see trading sites and others say that "pattern X is generally bullish/bearish" as if they were citing a statistical study that showed this result >50% of the time. triangle continuation patterns are only continuation patterns until they break counter-trend 
from my own work in Bitcoin, it seems that these patterns are very sensitive to their price environments (that is if they occurred within an uptrend or dowtrend; near, below, or above the previous high/low; and if the previous movement pushed the price into overbought or oversold territory).
once the environment has been factored in, i think that there is something to these patterns having a consistent "context-sensitive" bias. but it is ridiculous to make a blanket statement about them, not only because of context-sensitivity but also because i imagine the behavior is different in different markets (stocks vs Bitcoin).
--arepo
from my own work in Bitcoin, it seems that these patterns are very sensitive to their price environments (that is if they occurred within an uptrend or dowtrend; near, below, or above the previous high/low; and if the previous movement pushed the price into overbought or oversold territory).
once the environment has been factored in, i think that there is something to these patterns having a consistent "context-sensitive" bias. but it is ridiculous to make a blanket statement about them, not only because of context-sensitivity but also because i imagine the behavior is different in different markets (stocks vs Bitcoin).
--arepo
Interestingly enough, the Kirkpatrick/Dahlquist TA book has a (smallish) section on empirical studies that were done about patterns. I'll have to look up the details, but the continuation patterns flag and pennant seemed by far the most reliable ones *however* they were (IIRC) all short to midterm patterns, i.e. way shorter duration than what was proposed above (spanning from December to now).
thanks for the reference. is this the book you're talking about?
Log scale chart of complete bitstamp USD history.
Why might the latest pennant be such a different beast to the previous ones?
Why might the latest pennant be such a different beast to the previous ones?
well, we're currently resting on the bottom support in your figure. wouldn't this model be falsified if we broke below?
Log scale chart of complete bitstamp USD history.
Why might the latest pennant be such a different beast to the previous ones?
Why might the latest pennant be such a different beast to the previous ones?
it's interesting to see trading sites and others say that "pattern X is generally bullish/bearish" as if they were citing a statistical study that showed this result >50% of the time. triangle continuation patterns are only continuation patterns until they break counter-trend
from my own work in Bitcoin, it seems that these patterns are very sensitive to their price environments (that is if they occurred within an uptrend or dowtrend; near, below, or above the previous high/low; and if the previous movement pushed the price into overbought or oversold territory).
once the environment has been factored in, i think that there is something to these patterns having a consistent "context-sensitive" bias. but it is ridiculous to make a blanket statement about them, not only because of context-sensitivity but also because i imagine the behavior is different in different markets (stocks vs Bitcoin).
--arepo
from my own work in Bitcoin, it seems that these patterns are very sensitive to their price environments (that is if they occurred within an uptrend or dowtrend; near, below, or above the previous high/low; and if the previous movement pushed the price into overbought or oversold territory).
once the environment has been factored in, i think that there is something to these patterns having a consistent "context-sensitive" bias. but it is ridiculous to make a blanket statement about them, not only because of context-sensitivity but also because i imagine the behavior is different in different markets (stocks vs Bitcoin).
--arepo
Interestingly enough, the Kirkpatrick/Dahlquist TA book has a (smallish) section on empirical studies that were done about patterns. I'll have to look up the details, but the continuation patterns flag and pennant seemed by far the most reliable ones *however* they were (IIRC) all short to midterm patterns, i.e. way shorter duration than what was proposed above (spanning from December to now).
it's interesting to see trading sites and others say that "pattern X is generally bullish/bearish" as if they were citing a statistical study that showed this result >50% of the time. triangle continuation patterns are only continuation patterns until they break counter-trend
from my own work in Bitcoin, it seems that these patterns are very sensitive to their price environments (that is if they occurred within an uptrend or dowtrend; near, below, or above the previous high/low; and if the previous movement pushed the price into overbought or oversold territory).
once the environment has been factored in, i think that there is something to these patterns having a consistent "context-sensitive" bias. but it is ridiculous to make a blanket statement about them, not only because of context-sensitivity but also because i imagine the behavior is different in different markets (stocks vs Bitcoin).
--arepo
from my own work in Bitcoin, it seems that these patterns are very sensitive to their price environments (that is if they occurred within an uptrend or dowtrend; near, below, or above the previous high/low; and if the previous movement pushed the price into overbought or oversold territory).
once the environment has been factored in, i think that there is something to these patterns having a consistent "context-sensitive" bias. but it is ridiculous to make a blanket statement about them, not only because of context-sensitivity but also because i imagine the behavior is different in different markets (stocks vs Bitcoin).
--arepo
Guys. price is in negative weekly BB territory, in negative daily BB territory and additionaly bellow daily sma 200.
It could be a bullish pennant (with horizontal triangle?!) - I don't exclude. But I stay very bearish.
However this question will be resolved very soon.
PS. My pic with inverted cup and handle was invalidated of course.
It could be a bullish pennant (with horizontal triangle?!) - I don't exclude. But I stay very bearish.
However this question will be resolved very soon.
PS. My pic with inverted cup and handle was invalidated of course.
If a descending triangle is bearish, stocks' ascending triangle above is bullish? 
No! That's called an ending diagonal or a wedge. A triangle has either flat top/ascending bottom, flat bottom/descending top, or a more symmetrical descending top/ascending bottom.
If both lines are ascending or descending and not parallel, then it's a wedge.
What do others think of Luc's inverted cup and handle suggestion?
Its a dream. No fucking way is that going to happen.
Worst case scenarios is another dip into the low 500s, but we will not be seeing sub 500 again.
Should have known that China would be China. Nothing is for sure in bitcoin. If it does come true that on the 15th its banned in China then 400 probably wont hold. Otherwise I would very surprised if we broke the bottom of the pennant (at 400).
Triangles are continuation patterns. They break in the direction they are entered. If the original drop to 380 was an A, then it's bearish. If it's a 4 to the 1200 tops 3, then it will break up.
Again, quoting the StockCharts definition of a descending triangle (which has been my main source of info when it comes to learning TA)Quote
The descending triangle is a bearish formation that usually forms during a downtrend as a continuation pattern. There are instances when descending triangles form as reversal patterns at the end of an uptrend, but they are typically continuation patterns. Regardless of where they form, descending triangles are bearish patterns that indicate distribution.
Triangles are continuation patterns. They break in the direction they are entered. If the original drop to 380 was an A, then it's bearish. If it's a 4 to the 1200 tops 3, then it will break up.
It can, however, also be counted as a simple ABC correction where we are in the beginnings of a 5th wave down for the C
It can, however, also be counted as a simple ABC correction where we are in the beginnings of a 5th wave down for the C
If a descending triangle is bearish, stocks' ascending triangle above is bullish?
this is a bullish pennant.
Really? I was under the impression that a pennant had to have far less bias on either side, and according to the stockchart definition of a descending triangle; "The lows do not have to be exact, but should be within reasonable proximity of each other." A <5% difference in the lows would seem to be within the margin of what is "reasonable proximity" (someone correct me if I'm wrong).
Looks like your inverted cup n' handle is becoming a likelier possibility, luc
That, combined with an (approximate) descending triangle, seem to be some pretty damn strong bearish signals if validated
That, combined with an (approximate) descending triangle, seem to be some pretty damn strong bearish signals if validated
this is a bullish pennant.
Jump to: