Candlestick probability

I interpreted "Is there a probability..." as a statement looking for evidence or a refutation. This sentence is borderline meaningless.



To me, it's incredible people attribute some sort of actual probability to trading noise.Maybe if you're sitting on top of the exchange server it'd be possible. Even if you had an edge in the 5min or 10min I doubt your fills are good enough to capture significant edge.

To OP - You can test this relatively trivially. OHLC data is candlestick data. You can create a label "Up" or "Down" and run a logistic regression it with your predictors being N candle labels and the response being the label of the next candle. Be sure to use proper data processing and splits for out of sample testing. A confusion matrix is probably all you would need to at least sanity test your hypothesis.

You may find it works with some degree of edge (I'd consider anything better than buy and hold, even 0.5% better, an edge). I don't personally have the numbers for you but the system you are describing, "does a previous high predict a second high", has been debated ad nauseum with very little evidence either way. It comes up in TA of Stocks & Commodities probably twice a year. It's probably highly market dependent.

Then you would be wrong. 5 min is not noise in futures and never has been. Fills in ES are easy since hundreds of contracts can trade per tick so with limit orders there is NO slippage when one is trading 1 - 3 contracts. Maybe you are confused with high frequency stock trading which I don't do. No need for the server to sit on top of the exchange when my trades can last 30 min and sometime not even move for 1 min in either direction. Also, I am looking to make multiple points in profit not 1 tick in profit. In ES 4 ticks are 1 point.
 
Is there probability in predicting the formation of the next candlestick given the formation of previous candlesticks.
In order to become a BILLIONAIRE, you don't need to look deep into the future. You only need to know 1 bar or candle in advance. If you know that the next bar is going up, BUY; if it's going down, then SELL. Do that all day, every day with just 1,000 shares. You would be filthy rich in no time. I repeat, knowing 1 bar in advance can make all the difference. But the question is, is that possible?
 
OP, I would suggest you focus on being 40-60% accurate at predicting that you're taking a trade in a real-time uptrend or downtrend on a volatile trading instrument. When you can do that, having a long-term positive expectancy with your system is no big deal.

As far as guessing what price will do after the close of the current candle, I suggest you decide based on PRICE in relation to **something**. That something can be anything you want because your entry is a real-time position, highly relevant to where the market is vs. where you expect it to go to give you either a reward to risk ratio advantage or a winning pct advantage (you can't have both all the time highly in your favor as they are inversely proportional measures). If you base your entry on one indicator vs another then you are ignoring WHERE price is in real-time (aka lag) and that is going down the wrong road, imo.
 
Last edited:
I would suggest you focus on being 40-60% accurate at predicting that you're taking a trade in a real-time uptrend or downtrend on a volatile trading instrument. When you can do that, having a long-term positive expectancy with your system is no big deal.

"No big deal"?? :wtf:
You just described a coin flip, and associated that with positive expectancy. That's a pretty *big* deal. :D


The market has no awareness of or care about what chart type or period you are looking at.
^^ THIS ^^

:D
 
Say M1 charts, move the start and end time around by 15seconds and you'll get different candles making this useless as not everyones candles are timed the same.

Sadly, nothing in candle stick reading.
 
Certain types of candlestick analysis produce interesting statistical results.

Candlesticks are very small sub class of price patterns. There is program DLPAL S that can find many of those with required statistics but determining if they are fitted or robust is a challenge. Some turn out to be predictive.
 
Candlesticks are very small sub class of price patterns. There is program DLPAL S that can find many of those with required statistics but determining if they are fitted or robust is a challenge. Some turn out to be predictive.

Wise statement. As I always say, start with a hypothesis and see if the data match.
 
Back
Top