Caution: Math Geek at Work

No prob's.

The hardest part with this strategy is the first couple of units in, like at the moment I still have 2 in as per previous post but I have a very wide stop (as per your blog entry ) which is a bit scary with the current pull back. I'm still in with a chance though and I have another buy stop at another 50 pips, if this gets hit I'll move the previous stops to trail by 15 or to the last buy in price if it moves past quickly enough, but I'll leave the last unit with a wide stop to stay in for the correction but move it up with the others if my next buy stop gets hit....I hope

Btw, I'm trading the 4H for that trade so it will take a while.

In regards to box theory (I haven't quite got to read your paper yet) what would you recomend as an indicator that would be a close match, CCI perhaps?

(what a coincidence there's a bit of buzz on the forums today about them!, I was truly just studying them this morning)

Now I'm lost. I was talking about a way of looking at the distribution of a variable in statistics called box plots. Not sure what you mean by "an indicator that would be a close match."

Been looking at a spreadsheet and typing up new notes all night. Going to skip tonight's scalping session because I know I'm too tired to trade right. I'll check back in tomorrow. Once I finish this project I'll post my next installment on how I'm organizing my trading process.

I guess I'm after something that shows the price is away from the median and therefore show some potential for a return to it, doesn't matter as we know all indicators are after the fact but they can help support a theory or confirm one.

This is what I'm talking about, see below images, I ended up buying in again at a lower price as I was confident that my assumption was correct, I set my trailing stop to 85 pips and let it go, if I had of watched this and stuck to the buy in plan the figures would be very impressive for this move!

Cheers,
Mick.

Note: the profit in the last column is in pips with total being in $.

Great thread. Keep it going! You always have something interesting to say. What is more, you used my chart posting practise thread, so you get a gold star for that.

Quite like MickD's profit column. Looks like you got the whole move, Mick.

I'm still in and just bought another unit @ 229.12, this one has the stop at 2xATR with the others trailing at 55pips.
I missed some buy-ins last night as I set and forget and went to bed, sometimes that works better than watching the markets as it's too easy to pull the trigger when it's better to let it run

I still have to work the trailing stops out, if I set them all but the last buy in to 10-15pips (which is close to my last buy in price) I would have been taken out at least twice but could have rebuilt my column again as price continued to rise. I will have to do some figures to work out which was more profitable or made more sense if the trade did reverse.

Ah well, my four got stopped out but I still have one in in case the trend continues. If I had real money on this one I would have been up all night to stick with the buy-in strategy and could have done a lot better but it did pretty well I think - 1288 pips from a 400 pip move is nothing to sneeze at!

Oh ok, you must be using MetaTrader or something where you can custom design your own indicators. I don't know of any standard indicators that use the median price at all. All MA's use the mean. I've done some studies using the median and 1st & 3rd quartiles but it isn't anything I could plug into a chart (I use very few indicators as it is).

I basically set up the information in an Excel table and use it that way. For instance I might have a table that shows that half of all daily bars on EUR/USD that break the previous high go up by at least X pips but only 25% of them go up by at least Y pips, etc. I use stats like that only as guidelines in addition to standard TA & price action though. Never as standalone indicators because there's too much randomness involved.

As a matter of fact the project I'm working on this week is something like that. It's a statistical tool that I can use to help choose between several valid opportunities that I've found with other standard methods. I keep saying I'll post more about it when it's done, but I'm still only about 2/3 of the way through it.

Note to self: Less time browsing FF and more time working!

Wow! Nice trade on GBP/JPY. In another forum I use we refer to that pair as "the beast."

I managed to increase my account by around 3% today I think thanks to a short in gold and a long USD/NZD position I took earlier in the week. Grabbed the profit on both today. Also, I had the best scalping session yet; +26 pips in a single trade on EUR/JPY. I may be starting to get the hang of that. I won't increase my lot size until I have 10 straight winning sessions in a row though.

Welcome Smikester! Good to see your smiling electrons over here.
A gold star?? w00t!!

Don't worry, I'll keep the journal going. It's a good way to stay focused on what trading issues I still need to work out. So far all I've talked about:

1. Some musings on money management, scaling in, trailing stops, etc.
2. My choice of timeframes and the reasons (very long and very short term)
3. How I set up my weekly trading schedule.

The topic for my next main post will be on how I choose what to trade considering that I follow 18 pairs and it's easy to get confused. The tool I'm working on is meant to help with that so I'll discuss that in the same post.

OK, I'm off for another scalping session and then back to my project...

Good to see you turn a profit!

I can't wait to see your study, sounds good and similar to what I've been pondering.

Like you said though, the markets are far too random to take any indicators on face value and I wouldn't place a trade purely on a signal but a study like yours could help calculate potential targets and swings. Should be good.

Here's a screenshot of a nice scalping session. Second winning session today. And not only that... Madonna's singing "Like a Prayer" live in my earphones right now...

Traded this one like a sniper. Like a panther stalking the kill... A lot of waiting and trying not to give into the temptation to anticipate what the price was going to do; bounce off that trendline or break down through it? Finally the market told me. Gotcha!!!! I hope to get more and more sessions like this. My scalping trades are still in the red overall.




I'm out until tomorrow.

So here was the scalping session I had this morning:



After it was over, I took a walk in the Garden of FOREX and there I encountered Master Market...


Master Market: Ahh, hello grasshopper.

Grasshopper Cap: Hello master.

Master Market: You look disheartened.

Grasshopper Cap: Yes.

Master Market: Where are the 41 pips you gained yesterday?

Grasshopper Cap: I have lost them master.

Master Market: How can this be so? You picked the tops and bottoms perfectly.

Grasshopper Cap: Very funny master. I sold at the bottom and bought at two tops. I wasn't trying to find tops and bottoms. All the wizards say that it is wrong to do so. I was trying to enter with the trend. And in so doing, I found the tops and bottoms for which I wasn't looking. But I was facing the wrong way every time.

Master Market: Do you wish you had acted differently grasshopper?

Grasshopper Cap: I wish I had bought each time the price moved below the spine of the rainbow.

Master Market: I see. Because the spine is rising?

Grasshopper Cap: Yes master.

Master Market: But then why were you selling? What trend were you trying to enter grasshopper?

Grasshopper Cap: This one master...on the 5 minute chart. I was trying to enter on a pullback.



Master Market: Ahhhh...grasshopper. So the main trend was down. With this in mind, do you still wish you would have bought when price broke below the spine?

Grasshopper Cap: Well...no. No. To do so would be to fight the main trend. But master, if I should not have bought because the main trend was down, and I should not have sold because the spine on the detailed chart was up, then....

Master Market: Then?

Grasshopper Cap: Then I should have done nothing.

Master Market: And would it not be better grasshopper to have saved 41 pips by doing nothing than to lose them by doing something without reason?

Grasshopper Cap: ahhhh....

As I've mentioned, I like to trade in either a very short or a very long time frame, so I am working on two separate strategies. One is for scalping and the other is for long term trades on the daily/3-hour charts.

My next step in developing my long term strategy is (as I've been saying all week) to finish the development of a statistical tool which uses weekly bar pattens. I'll be done with that this week and will use it for the first time this weekend in preparation for next week. When I do that I'll post more about it.

This post is about the short term scalping strategy though. I haven't defined my technique yet because I wanted to get some "screen time" with the WMA rainbow that LinuxTroll uses. By experimenting with different methods I expected to lose for a while, but also to gain enough experience to devise a method that I can trade consistently. After thinking about this morning's losing trading session (see the "Conversation with Master Market" post), I came up with a technique that I'm going to try consistently.

There are 5 steps:

1. Review the longer term charts to get the big picture of what's happening

2. Identify the main trend in the 5-min / 1-min charts

3. Wait for a pullback on the main trend. Price should pull into the blue.

4. Go down to the detail chart (10-sec) and wait for the spine to turn back in the direction of the main trend. This indicates that the pullback is over.

5. Now that the spine is moving with the main trend, enter when the price spikes to the wrong side of the spine. So in a downtrend, sell when the price breaks above the declining spine. In an uptrend buy when the price breaks below the rising spine. This increases the reward to risk ratio.


--------------------------------------------------
A CAPITALIST RANT WARNING
--------------------------------------------------

By the way, here's a pet peeve. Why do people say "risk reward ratio" when they really mean the opposite? When someone says they want a "high risk reward ratio" they're really saying that the want this ratio:

RISK / REWARD

to be high. Really? Is that what you want? Assuming a constant win rate and a potential reward of 50 pips, do you REALLY prefer to risk 40 pips rather than 20 pips on that trade? No. If you must say "risk reward ratio" then at least understand that what you want is a LOW risk reward ratio (all else being constant of course).

Risk is in the numerator. Reward is in the denominator. Lower risk and higher reward makes a LOWER ratio. Yeesh.

Ok, in this post I'll give an example of how I often use basic statistical analysis to help out with strategy development.

Going back to my scalping strategy that I discussed in the last post, I talked about how I want to enter the market when the price drops below a rising rainbow spine (WMA144 in my case) or when it penetrates above a falling spine.

The natural question I had at that point is what happens if the price keeps going once it penetrates the spine? So I decided to check to see how far price penetrates the spine once such a penetration occurs.

From Dukascopy I took about 16,000 10-second bars for EUR/JPY (I don't know how I got so many; I thought they capped it at 10,000).

Anyway, I then set up my Excel spreadsheet to calculate the WMA(144) for the close of each bar. This was a little tricky because I had to check the formula on Wikipedia to see how successive values are calculated, had to check that whoever wrote the article had the math right, etc.

Next I defined a "rising" WMA(144) to exist when the last value is greater than or equal to all of the previous 30 values. A declining average was one that was less than or equal to the past 30 values. On the chart this translates to a spine that has been going in the same direction for at least 5 minutes.

Next I set up a column to identify bars in which the high had penetrated above a declining WMA(144) or the low had penetrated below a risng WMA(144). These are the places where the price is on the "wrong" side of the spine and represent entry opportunities. I set this column up to return the extent of the penetration in pips.

Next I did a quartile analysis of the penetrations. I found that the median penetration was only 2 pips and the maximum was only 11 pips. So I decided to set up a detailed distribution of the results by penetration distance. From that I could calculate the frequency with which price reaches each penetration level. Here's a chart of that:



Finally, I did a reality check by looking back at this morning's scalping session. Sure enough, the price low never gets past 10 pips below the WMA(144) when the average is rising. So now I know that even if the spine is penetrated just a little bit, by a couple of pips or so, I can safely grab it at that point without having to worry too much about it spiking down by another 20 pips or something.

Having the stats in hand helps create more confidence in a method.

Hello again!

Last week was much better than the previous week due to a short position I took in gold and a long position in NZD/USD. I like to trade these together because of the carry revenue, but I don't hold them very long. I entered because it looked like gold was bouncing down off of a double top. I exited a couple of days later when the momentum seemed to be going away. At any rate, my account value increased by 4.6% so here's the updated chart:



So far I've talked about what timeframes I like to trade (very long and very short) and how I set up my trading schedule (10 sessions throughout the week during the London/NY overlaps and the Sydney/Tokyo overlaps). So now it's time to figure out WHAT to trade, or at least what to keep my eye on during the coming week.

I follow 18 pairs plus silver and gold, but the problem with that is that I tend to bounce around aimlessly from pair to pair without concentrating on any of them enough. In order to eliminate this confusion, I set up three similar tools on Excel, each of which looks at all of these pairs through a different "lens." There is a fundamental tool based on economic releases, a technical analysis tool based on trends, and a statistical tool based on weekly bar patterns.

Here's what the output of these tools looks like for this week:



Explanation in the next post...

In each of the three tables, the base currencies are listed down the side and the counter currencies are the ones across the top. So the upper right hand pair in each table is AUD/USD for example. Pairs which I don't follow (or which don't exist) are greyed out.

1. Technical Analysis: Trends

The top table shows the trends as of last week's close. These are based on the relative positions of price, the SMA34 and the SMA144 on the daily chart. If P > SMA34 > SMA144 then we have an uptrend. These are in dark green. If the price has pulled back to between the two SMAs then we have an uptrend with a pullback which is shown in light green. The dark red and light red are similarly downtrends and downtrends which have pulled back. White pairs have no trend by this definition.

2. Fundamental Analysis: Last week's news

The bottom table shows an integer value for each pair which I calculate based on the previous week's news releases. At the beginning of the week I start off each currency with a value of zero and then add or subtract a number from that according to each news release on that country's economy.

The value for each release is the product of two numbers; the first being the "importance" of the release as color-coded here on the Forex Factory calendar. That part can either have a value of 1,2, or 3.

The other part of the product is a function of the release value vs. the forecast and last month's number. Based on some research (available in my site's archives), I found that the most common occurrence is for the forecast change in the economic release to be in the correct direction from the previous month, but with a lower magnitude. In other words, the forecasters usually lowball their estimate for how much the number will change from month to month. For example if last month's consumer confidence level was 88% and the forecast is for it to be 85% this month, what will most probably happen is for the actual number to come in lower than 85%. My research showed that this also has the least effect on the market.

When the change from one month to the next is in the right direction but SMALLER than what the forecasters think, that causes more of a reaction. This is also the second most likely event. Finally, when the number goes in the opposite direction than the forecasters think, that causes the greatest reaction. This is also the least likely event.

So the second part of the product is based on this behavior. If the number comes in in the right direction and the change from last month is greater than what the analysts forecast, then the value is +1 or -1 (the sign is based on the generally accepted effect of the news). If the magnitude of the change doesn't meet the forecasters expectations then the value is +2 or -2, and if the number goes in the opposite direction, then the value is +3 or -3.

Combining this with the "importance" of the release gives me a value for each release that ranges anywhere from -9 to +9. Here are a couple of examples:

Just tonight, Japan released it's Capital Spending report. Last month's number was -4.9% and the forecast was -2.5%. So the forecast was for the number to increase by 2.4%. In fact the actual number was -1.2%. So it increased by MORE than the forecast. This is positive for Japan, but is the most common situation so we assign it a value of +1. But the report itself has an importance of 2, so I actually add +2 to Japan for this report.

Australia released its Trade balance tonight also. The previous month's value was -1.9B and the forecast was -1.8B. So the forecast is that the number will improve by 0.1B, and normally what will happen is that the actual improvement will be a little more because the forecast is usually lowballed. The actual number came out at -3.0B, which wasn't even in the forecast direction! This has the most effect on the market, so we assign it a value of -3. But the report has an importance of 3, so this release has a total effect of a big red -9 for Australia.

So far we have +2 for Japan and -9 for Australia. To find the combined news value for any pair we just add the value for the base currency and subtract the counter currency. So the value for AUD/JPY right now would be -11.

3. Statistical Analysis: Bar patterns

The middle table is a result of the tool I was working on all of last week that I kept talking about. Again, there is a 16-page set of research notes in my site's archives on this tool so I won't go into too much detail. Basically I created a simple set of 16 possible bar patterns that can occur (there are countless ways to create such sets) and checked each one to see what the average price behavior was during the following week. I cross checked this with 14 pairs to see if the behavior was mostly the same across pairs. This allowed me to assign bullish or bearish "scores" to each pattern ranging from -5 to +5.

Now this by itself is not a tradable concept because there's too much randomness in it. Just because there is a "statistically bullish bar pattern" one week doesn't mean there is a significant probability that price will rise during the following week. However, by finding the bar pattern scores for each of the pairs that I follow, I can then comine them into average scores for each currency. For example if most or all of the bar patterns in the XXX/JPY pairs are bearish, JPY will have a positive average "strength." Finally, I recombine the currencies back into pairs, using the strength of each currency to come up with strength for each pair. That's what the table shows. All the specific calculations are in my research notes.

Next post: So what do I do with this??

So now that I have 3 different windows on the market, how do I use them to decide which pairs on which to concentrate?

The most important thing to me are the trends, so I start off with the top table. If there's no existing daily trend in the pair then I don't want to bother with it. Looking at the table tonight I can see that all XXX/USD pairs are in uptrends and all USD/XXX pairs are trending down. Whatever happens, I'm pretty sure I want be short the USD sometime this week!

What does the middle table tell us? We're seeing a lot of red in the JPY column, indicating that there are bearish weekly bar patterns in all of the XXX/JPY pairs. So long Japan seems like a good idea too. We also see bearish patterns in all GBP/XXX pairs and a bullish pattern in EUR/GBP. Short that pound! When we look for the highest magnitude indication in this table, we indeed see that it is -3.42 for the GBP/JPY. Looking at our trend table we see that GBP/JPY is in a downtrend, and our news table shows us a negative news score for this pair based on all of last week's news so we're good there too. We'll look for short opportunities in GBP/JPY this week.

But what about shorting the dollar? We can't use the GBP or the JPY because we're already looking for a position in GBP/JPY. So what's the next best bet in the bar pattern table? I see gold and silver at a strength of 2.78 each, but those have spreads that are a bit too high. I also like to be short metals for the interest rate differential. So our next best score is a short position in USD/CHF with a score of -1.78. Looking up at our trend table we see that USD/CHF is in a downtrend so that criteria is met. Our news table shows us that USD/CHF has the third largest news score in the chart from last week and it's negative. Looks good.

Another possibility is the EUR/USD because that pair has the largest news score from last week. It's in an uptrend and the statistical bar pattern score is a respectable 1.61. I'll add that one to the list just to be complete.

So in summary, this week I'll be looking for opportunities to enter:

SHORT GBP/JPY
SHORT USD/CHF
LONG EUR/USD

This of course is for the long term daily chart portion of my trading. As always, I will be doing scalping sessions with EUR/JPY each day.

That's it for now!

A minor change to the pairs that I'm following this week for long term trades. I'm substituting a long EUR/AUD position for the long EUR/USD position that I mentioned in the last post in order to not have two positions in the USD at the same time. So my opportunity watch list this week is:

Short GBP/JPY
Long EUR/AUD
Short USD/CHF

I tried an entry on GBP/JPY last night and found that I was stopped out. The stop was way too close for daily bars and was actually well below the resistance up at 229.50. That got me thinking once again about not entering the entire position all at once, and the actual entry and exit techniques I want to use once I've identified my potential positions. So here's what I came up with on that:

1. Identify a place for the stop

The nearest 100 pip round number beyond the last S/R on the daily chart.

2. Look for a basic price action setup.

If I want to be short, I should be seeing daily bars that are either inside, outside or lower. If I see a higher bar I wait until the first lower bar before considering an entry. I only enter on a bar that has broken the previous bar's low.

So right now in GBP/JPY, we've been seeing higher bars since the low on 11/22. But then we saw an inside bar, a slightly higher bar, and finally today there's a lower bar. I'll place entry orders here.

On EUR/AUD we've seen three lower bars in a row, and now we've got a bullish higher bar. I have entry orders in here too.

On USD/CHF, we've had three higher bars in a row, and now we're seeing what looks like it will be an inside bar. I have no orders in here because there's no lower bar yet. This recent upswing isn't broken yet.

3. Place orders using a "set and forget" dollar cost averaging approach.

Basically I use my stop level as a zero point, and consider the distance from the current price to the stop to be the "cost" in pips of that order. Using a (relatively) constant dollar amount based on my account size, I determine my position size by dividing that dollar amount by the distance in pips.

To make the numbers come out evenly, I enter at round numbers. So what I do is simply place two orders, one below and one above the current price. So for GBP/JPY for example, my stop is at 230.00. The price is at about 227.70, so I place a buy order at 228.00 and another buy order at 227.00.

I'm using a risk factor of 600 unit-pips per entry, so this allows me to calculate my position size for each order. The first one is 200 pips from my stop so my position size on that one is 600/200 = 3 units. The second one is 300 pips away so that position size is 600/300 = 2 units.

Both of these may be hit, neither, or just one of them. In any case, I'll check the trade tomorrow and place additional orders based on wherever the price is then. If the pair is deep in profit I'll push the stop forward and use the new position of the stop to determine the new position sizes.

If the position is in losing territory, but not yet stopped out, I'll use the same basic price action considerations to decide if I want to add to the position. If the pair looks like it's going back in my direction, I'll add to it again. This time the price will be closer to the stop, so even though I'll be adding the same dollar amount of risk to the trade, I'll be adding a higher number of units. So if the price is only 100 pips from the stop for example, I would be adding 6 more units.

This isn't like a martingale or anything all crazy though. I have a stop price that doesn't get moved back, a top dollar amount that I will risk on the trade, and I only add to the trade when the price action is going back in my direction.

In other news, my scalping is still not up to snuff; lost another 11 pips on it today. Bleh.

We'll see how it goes this week!

Love the pairs charts Cap', good work!

Ahhh...the puzzle continues

errr..ahem. Ok, NOW the notes are actually there. Hopefully nobody went looking for them because I actually forgot to upload them to my site until just now. They're the 16 page ones marked November 30, 2007.