As I have explored the topic of support and resistance I have taken interest in a specific kind of S & R levels, pivot points. I have read a couple different articles that show a statistically derived probability that a certain pivot point will be touched or some behavior will occur in relation to the various s & r levels. Take for example Jamie Saettele’s article in Investopedia. (Ref. 1)
The result: there have been 2,026 trading days since the inception of the euro as of October 12, 2006.
• The actual low has been lower than S1 892 times, or 44% of the time
• The actual high has been higher than R1 853 times, or 42% of the time
• The actual low has been lower than S2 342 times, or 17% of the time
• The actual high has been higher than R2 354 times, or 17% of the time
• The actual low has been lower than S3 63 times, or 3% of the time
• The actual high has been higher than R3 52 times, or 3% of the time
I have degrees in engineering and physics:22:, so I prefer empirical proof over “take my word for it”. The issue I’m having is that my statistics are a little rusty! How are these probabilities calculated? I mean how is the data statistically analyzed to come up with 44% of the time the low has been lower than s1 (see the first bullet above)?
It seems reasonable, given the history of pivot points, there should be examples on how to do a hypothesis test or make a probability distribution of price action as related to these levels. Or is this just esoteric knowledge?
CAN ANYONE HELP???
Ref. 1: Using Pivot Points In Forex Trading; (Link Removed-Just google it!)
you could download the price data, calculate the pivot points, count how many times the daily high was above R1… divide that total by the total number of days. Seems like simple stuff. If you get a demo metatrader account you can download the price data to an Excel spread sheet and set up formulas there.
I agree with you, seems simple. I do tend to make things more difficult than they need to be. I’ve done as you suggested, about a month ago. My Excel spreadsheet has over 8,000 data point and after coding algorithms for pivot points etc it tends to crash excel! I’ve had to move analysis between multiple spreadsheets and I am now doing cals in R using R Commander.
It seems you and I have come up with modeling a binomial distribution. Going forward from say r1 to r2 I’ve repeated the process. So if I find that for a given period (say 3 months) I have a 39% chance of reaching r1 and using the same method I find that I have 20% chance of reaching r2, what is the probability that I will reach r2 if price action is at r1? Would I just multiply the two percentages to made this assessment or maybe divide? So, .39 * .20 = .078 (Seems Low)?
If you’re looking at multiple probabilities you would multiply for example the odds a coin will come up heads twice in a row is .5 times .5 = .25 but you already know that. To find the probability empirically of reaching R2 for any that reach R1 I would just take a subset of all those that reach R1 and then see how many of those reach R2.
To reduce the chance of Excel crashing, I would break it into smaller tasks. Just one year of data on one currency pair and all you need is daily data because that’s what the pivots are calculated on is the daily high, low, open and close, so really you would only have 260 rows of price data. If a full year is not statistically significant (but I think it would be) then you could do another separately and combine the totals later. And rather than calculate all pivots I would only calculate for R1. Then clear those formulas and do it again for R2, then R3 etc.
I’m a computer programmer so I’m used to breaking things down into small steps. Actually I’m almost tempted to download all the data and put in an IBM iSeries(AS400). You can really do some number crunching with a workhorse like that.
Nice to see you still around my mathamatical friend! So… whats the statistical probability of a winning trade if I short swing trade EU or GU next week? Just asking.
Hey RC haven’t talked to you in a while, good to see you!
Actually my bias is generally up especially EU. GU is kind of choppy. But you can always make some pips shorting on the old daily high low!
I’ve been getting up early for London Open lately ala Matt strat. As well as looking at smart money triggers ala ICT and supply / demand ala Eremarket.
And hopefully not getting information overload in the process!
edit:
I looked into my crystal ball… sometime next week price will be lower than it is right now. Stay tuned and lets see how that prediction turns out.
Sorry for the delayed reply I drove from South Carolina to Alabama and back (500 miles) and supported my wife in a 50K run in the last 24 hours. Yesterday after visiting with some folks I was able to determine that by Bayes’ Theorem that I would divide P(R2)/P(R1) and get the probability of hitting R2 after getting to R1. So I’ve downloaded historical data (2004-Present) and have determined historically the probability of touching the different S & R levels. I also played with calculating using subsets like you suggested, it works just fine. I’ll try to post a file or screen shot with my findings. The problem I face now is being able to explain and demonstrate what I’ve learned (Maybe someone else will find this useful someday).
Talon when I talked about writing an algorithm I am a complete novice. I have to write my ideas up on a white board and the go step by step when trying to figure out the coding in excel or vba in excel. No “real” programming or algorithms for me. I saved an old computer and cleaned everything off of it except the operating system, Matlab, and fxcm’s trade station. I use the old dog to optimize strategies with trade station. I was thinking that if I got really serious someday I could use the old machine to do the number crunching for me. But first that is beyond my capabilities and I would have to do some serious studying (who has time for than when your studying other aspects of the forex market), second the old machine is just that OLD and I’m sure there are better machines out there that can do what I want and better, like your suggested IBM iSeries.
With all of that said, lateley I’ve been wondering how in the world am I ever going to become a successful trader, thus the portion of the title “Esoteric Knowledge” of this thread. I have seriously been studying and experimenting with strategies for the last 6 months. I’m only $300+ in my demo account than where I started. Can you tell me where to get a crystal ball?
I keep my crystal ball on a shelf right next to my box of scratch. silly buggers on the weekend. Anyway lacking a real crystal ball, I made my little prediction based on the fact that on a weekly chart the price is at a high and since most candles overlap, next weeks low should be lower than this week’s high. Who knows by how much. On the other hand after a strong move up like last week there may not be much overlap if any. So we’ll see.
Back some time ago I did some probability analysis of price moving a certain distance beyond the daily high or low using Excel. The results were interesting but I didn’t pursue it to form a trading strategy. Another statistical analysis I did one time was to see which way price would go based on the previous candle. For example if the previous hourly candle closed up would the next one close up also? A year’s worth of data showed random 50:50 results with the exception of a slight bias due to a longterm trend that was clearly visible on the chart. So anyway, I kind of gave up on statistical analysis.
Hi,
i’ve worked with dukascopy’s platform for a while, and i think it shouldnt be too hard to write a little EA and backtest it. The calculations of the different levels shouldnt be hard at all… then we could have the EA calculate the number of candles closed above/below certain levels on any timeframe, and give us a summary in the end.
haven’t talked to you in a while, good to see you!