History repeats itself - a lie? - page 7
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
StatBars, it doesn't matter as much whether as a filter or as a strategy: this filter is not statistically sound. Or should our requirements for a filter not be as rigid as for a strategy?
Regarding joint analysis: quite possible.
Or should our demands on the filter not be as rigid as those on the strategy?
I have a question for youMathemat: look at the forecast table in the "stats" category, I mean the initial view which is used to make a forecast... I've noticed a mistake - we always have buy on Mondays, it shouldn't be like that...
And the reason is that GBP goes up all the time, I even see this trend statistically.
I cannot explain what I am asking, but I think you will understand.
Here is an example of what I want:
63
2,619543
61
2,854469
You see the number of cases is higher, but if you make a prediction on the second column it is exactly the opposite.
And I did a simple thing, I calculated the total number of baro B and divided all combinations with outcome B by this number, the same for H...
Do you think this is correct?
Can you elaborate on your approach to whether or not the event was successful?
1. Candle coding.
The old techniques were unambiguously oriented to small memory models, so a possibly denser one was invented - linguistic packing (when it was))))).
Now it is not relevant - there is a lot of memory, and most importantly linguistic coding interferes with analysis.
That's why coding candlestick is simplified: we start arrays int as much and as much as we want.
I.e. we look for patterns, ignoring the standard linguistic coding!!!!
2. timeframe.
On H4, the probability is pulled down to 0.5. On H1, there's nothing to see.
It turned out, and this is important, -Forexhas a cycle = 1 day, I emphasize this very and very important observation)))), which was long ago laid out y Larry Williams, but again, it had to be proved.
Candlestick analysis and proved it - candlesticks look on D1. Below D1 the patterns are blurred - above D1 there is no practical sense.
only on D1.
3.
And when it became clear to look only at D1 candlesticks - the need to determine the direction of the forecast disappeared - for D1 it's just the colour of the next candlestick.
I emphasize: the colour of only one next candle.(They say that candlesticks forecast by two bars - now I don't believe it!)
It's clear that for H; the colour of just one candle as a prediction would be both incorrect and insufficient.
Conclusion: candlesticks only for D1.
From the above 1,2,3 we get a quick algorithm.
If you agree with it, here is the result:
The running showed that the colour prediction of the next D1 candle is about 0.56 and weakly depends on the candlestick combination.
(The program is somewhere in the annals, I'll have to look for it).
This is not good. It is better to take a plot without a clear trend, i.e. with approximately zero result. Otherwise it will be the same - "stat advantage" B. But even in this case it won't be very significant. In general, the global trend cannot be removed from candlestick combinations, because its contribution is very small and practically does not affect the coding (B or H).
I don't understand something about the calculation. Numbers are too big. There are actually two numbers, or what - one is green, and the second is some kind of fraction?
This is not good. It is better to take a plot without a clear trend, i.e. with about zero result. Otherwise it will be the same - "statistical advantage" B. But even in this case it will not be too significant. In general, the global trend cannot be removed from candlestick combinations, because its contribution is very small and practically does not affect the coding (B or H).
I don't understand something about the calculation. Numbers are too big. There are actually two numbers there - one is green, and the second is some kind of fraction?
The first number is the number of such combinations just before that day. The 2nd number is the first column divided by the total number of B and H candles respectively. I have been counting combinations since 1990, as a result B=2405, H=2137.(it is divided by the first column).
And all the same, there is a trand contribution, and it's quite substantial...
1. candle coding.
The old techniques were unambiguously focused on small memory models, so a possibly denser - linguistic packaging was invented (when it was))))).
Now it is not relevant - there is a lot of memory, and most importantly linguistic coding interferes with analysis.
That's why coding candlestick is simplified: we start arrays int as much and as much as we want.
I.e. we look for patterns, ignoring the standard linguistic coding!!!!
2. timeframe.
On H4, the probability is pulled down to 0.5. On H1, there's nothing to see.
It turned out, and this is important, -Forexhas a cycle = 1 day, I emphasize this very and very important observation)))), which was long ago laid out y Larry Williams, but again, it had to be proved.
Candlestick analysis and proved it - candlesticks look on D1. Below D1 the patterns are blurred - above D1 there is no practical sense.
only on D1.
3.
And when it became clear to look only at D1 candlesticks - the need to determine the direction of the forecast disappeared - for D1 it's just the colour of the next candlestick.
I emphasize: the colour of only one next candle.(They say that candlesticks forecast by two bars - now I don't believe it!)
It's clear that for H; the colour of just one candle as a prediction would be both incorrect and insufficient.
Conclusion: candlesticks only for D1.
From the above 1,2,3 we get a quick algorithm.
If you agree with it, here is the result:
The running showed that the colour prediction of the next D1 candle is about 0.56 and weakly depends on the candlestick combination.
(The program is somewhere in the annals, look for it)
All is right as I thought, but there is one BUT, you're not guessing the colour of the candle in forex, and how many points - the spread, etc... it's not the correct assessment, and yet it's still not for combinations ...
For example the combination "spring = long candle with doji + reverse long candle" is only discernible huzzily but statistically indistinguishable.
91% of morning/evening stars end in nothing.
Tried to poke around this topic yesterday. Statistics for 3 bars (colour combination only) forecast for 1 bar. Watched GBP,EUR,JPY. Forecast accuracy fluctuates between 52-56%. I've been working on paternals since 1997. I've been testing it since 01.01.2007.
I think what to do with it and if I should do it at all, may be I should complicate a pattern with some data about colours of cumulative candlesticks (out of 3) for example.