NFP System - Backtest data provided

If I look at last two bars on all 6 pairs' charts in FXDD, the time would be 2008.09.22 18:30 bar and 19:00 bar I'm looking at on all 6 pairs:
EURUSD 18:30 close = 1.4695, 19:00 close = 1.4688, so gain of 7 pips if selling.
GBPUSD 18:30 close = 1.8482, 19:00 close = 1.8462, so gain of 20 pips if selling.
AUDUSD 18:30 close = 0.8439, 19:00 close = 0.8424, so gain of 15 pips if selling.
USDCHF 18:30 close = 1.0834, 19:00 close = 1.0827, so loss of 7 pips if buying.
USDCAD 18:30 close = 1.0381, 19:00 close = 1.0376, so loss of 5 pips if buying.
USDJPY 18:30 close = 106.35, 19:00 close = 106.40, so gain of 5 pips if buying.

So add these gains/losses together, and you get:
7+20+15+(-7)+(-5)+5 = 35

But if I look at:
original Sentinal Index (set to MODE 1 for pips), at 18:30 close is 701.2857, 19:00 close is 701.4286, for gain of 0.1429.
Sentinal Revised, at 18:30 close is 17.3850, 19:00 close is 17.3938, for gain of 0.0088.
Sentinal Revised Extended 2 (attached), at 18:30 close is -52196, 19:00 close is -52137, for gain of 59.

So we gained 35 pips during that half hour if we would have been "buying USD", but none of the indicators are showing this correctly. Even if I take averaging of original Sentinal into account, and instead don't average, that would be .1429 * 6 pairs, to equal .8574 gain.

Well, I would think it would be simple to fix this, but like most things, it does not appear to be so simple. If I look on the chart, Sentinal Revised Extended 2 looks more like original Sentinal Index indicator's lines, where Sentinal Revised doesn't look like either of the other two. This leads me to believe that the decimal is making a problem in Sentinal Revised indicator, which I've (hopefully) fixed in Sentinal Revised Extended 2 by separating USDJPY from the others and making PipsAsWholeNumber set to True.

When Sentinal Revised Extended 2 has PipsAsWholeNumbers set to False, it looks like Sentinal Revised when on chart. When Sentinal Revised Extended 2 has PipsAsWholeNumbers set to True, it looks like original Sentinal Index (MODE 1) when on chart. I'm thinking Sentinal Revised hasn't taken the decimal problem into account, where the other two are.

Don

crossover for usd/chf and u/j

Kevin,

I have attached my original indicator I built, as yet another confirmation to what I'm seeing. This indicator looks like Sentinal Index and Sentinal Revised Extended 2 indicators when put on the chart. But my original indicator does not update correctly, just like Sentinal Index doesn't.

If you put Sentinal Index, Sentinal Revised, Sentinal Revised Extended 2, and Dons 6Pair BUYUSD PipRange092298 on the chart, which one looks different than the others? Sentinal Revised. Needs checked.

Kevin, I do thank you for updating/making a new indicator, I just want and need to make sure it's right before using in trading.

Thanks again,

Don

I would recommend the Extended 2 version to use at this point, to test with. I would not use any of the indicators in trading until they can be verified to be correct in calculations (which we can do by manually taking a 30 minute period, and checking if the combined pip gain or loss is the same in the indicator as it is in reality).

Don

If you place the Sentinal Revised indicator onto a M30 chart (not USDJPY, just in case it causes a problem doing so, I don't know) using the defaults that came with it. Now place it again on the same chart with the first one, except make every pair (EXCEPT USDJPY) to be 0, leaving JPY at 1.0. Now these should look different, but they don't.

This is just more proof that USDJPY value is leading (2 decimal places), while all the other pairs aren't contributing (4 decimal places taken for two would make whole numbers 1/100th their actual value, so would explain why USDJPY is leading and the others look non-existent even when made 1 in inputs, meaning used). Each pair's (EXCEPT USDJPY) 100 pip gain would be seen as a 1 pip gain, so basically not contributing. The fix would be to make the USDJPY pair's value be divided by 100, to make it 4 decimals, and therefore the same as the other pairs involved, before adding the total gains together to make the Sentinal. Then, all I wouldhave to do is multiply by 1000 to make 4 decimals to be no decimals (make pips as whole numbers).

Kevin, I hope you know that you did a great job building the indicator. I have just found this problem that has made this indicator needing more work. My day job is to find problems and solve them (in commercial printing as a prepress technician of 13 years, 15 counting college). When I see multiple tests/ways that tell me the same thing, then I must take it seriously. And since I consider I'm building a trading business, I must take this seriously.

Thank you,

Don

OK. I figured out I needed to divide USDJPY by 100 to make it 4 decimal places like the others, then multiply USDJPY by 100 to get whole number. These two moves strike each other out, so I removed the multiply by 100 on USDJPY in indicator Sentinal_Revised_Extended2. I then left the others multiplied by 1000 to make whole numbers. This should work fine now. The appearance on-screen should look like the original Sentinal_Index, except this indicator looks to update correctly on the charts (thanks to Kevin. woohoo!). I have looked at the results of this indicator's numbers when looking at the times I mentioned in earlier posts today, and between 2008.09.22 18:30 bar and 19:00 bar on EURUSD M30 chart the result is +35 pips, just like reality.

Just as before, I have set moving average Period to 336 (although the 100 SMA Kevin set as default in Sentinal_Revised looks like my 48 SMMA I use, so I would suggest using it on M30 chart instead of 336). I have set inputs Average to True, and PipsAsWholeNumbers to False, so you get a "relative" USD Index value. I will use Average to False, and PipsAsWholeNumbers to True, so I get an "absolute" pip gained from "buying USD" (buying USDCHF,USDCAD,USDJPY and selling EURUSD,GBPUSD,AUDUSD). If the line goes under the moving average, I will "sell USD"(selling USDCHF,USDCAD,USDJPY and buying EURUSD,GBPUSD,AUDUSD).

The indicator to use is attached:
Sentinal_Revised_Extended3

Sorry if I've posted too much, but it took going through it to get to the right results.

Don

I was laying in bed last night and figured out I had made a mistake with the last indicator.

OK. Let's logically work through this:
USDJPY has two numbers after the decimal. One pip is 0.01. We want to make it 0.0001 to be like the others, so we can add all gains together. If we didn't do this, I'm thinking that one pip gain on each would look something like this:
EURUSD +0.0001 (will be positive if we are selling and it goes down)
GBPUSD +0.0001 (will be positive if we are selling and it goes down)
AUDUSD +0.0001 (will be positive if we are selling and it goes down)
USDCHF +0.0001 (will be positive if we are buying and it goes up)
USDCAD +0.0001 (will be positive if we are buying and it goes up)
USDJPY +0.01 (will be positive if we are buying and it goes up)
equals 0.0105 on Sentinel Revised instead of .0006 like would be correct. This needs to be verified by someone else.

Note: The selling vs buying may be confusing, but we "buyUSD" (buy USDJPY,USDCHF,USDCAD and simultaneously sell EURUSD,GBPUSD,AUDUSD) when the Sentinel crosses above the moving average, and "sellUSD" (sell USDJPY,USDCHF,USDCAD and simultaneously buy EURUSD,GBPUSD,AUDUSD) when the Sentinel crosses below the moving average.

Since we need to make USDJPY to be four decimal places, 0.01 / 100 = .0001. So we divide USDJPY by 100, and don't have to do this for the rest of the pairs, since they are already four decimal places, so one pip is equal to 0.0001.

Now if I want to make whole numbers, I simply multiply all pairs by 10000. 0.0001 x 10000 = 1

I have made these logical changes, and tested the indicator a little. I'm not sure if I need to upload it or not. Something still doesn't look right to me. I would rather Kevin or someone else take a look. I just know that to get whole numbers, I should be able to multiply numbers with 4 decimals by 10000 to get a whole number, and it should numerically match the Sentinel without averaging, except for the decimal place.

Don

Alright. I decided to go back to Sentinel_Revised that Kevin posted, and just take out everything that might be causing trouble. I removed Averaging, so it is no longer a user input. I made PipsAsWholeNumbers to be the norm, so it is no longer a user input. I removed Input for AppliedPrice, and just used the close. I separated all pairs so I could treat USDJPY separate. I have tested a little, and it's looking good. Less user inputs, but the way I see it, if it's not right, then it doesn't matter how many inputs it has. If I can make it right with less user inputs, I'll do it, and think I have done it successfully. Only testing will tell for sure. I have also attached a pic for those that are interested, so that anyone can tell the difference between Sentinal_index, Sentinel_Revised, and the new Sentinel_Revised_Don. If the lines didn't look SO off, maybe I wouldn't have noticed. For everybody's sake who uses it, I had to voice the problems I saw, and come up with a fix if I could. Seeing the difference between them on-screen (see the pic) should explain why I made a big deal out of the correctness of the indicator. The top indicator is Sentinal_index. The middle indicator is Sentinel_Revised. The bottom indicator is Sentinel_Revised_Don.

Don

Kevin,

As you've hopefully seen, this was not the problem. The problem mainly was that JPY's two decimal places was causing problems. I should have been able to take your indicator and just multiply by 10000 to get the correct whole number, and not use Point or anything. But that wasn't the case, so I followed the problem to where it lead, and found a fix by dividing ONLY USDJPY by 100 to make it four decimal places, then multiply all pairs by 10000.

I would rather you update it and provide a working Sentinel_Revised that looks like Sentinal_index when put on the charts (because it looked corrrect). I have provided a simple 6pairs indicator that also proves the point visually and numerically.

I don't want to argue right or wrong. That is not productive. Just looking to get it fixed by the original author, because it would be best if you did it, since you gave the NFP strategy and also the revised indicator that's not correct. I do thank you for posting the revised indicator, which does update the same on all pairs' charts, which Sentinel index didn't and doesn't do. If it wasn't for you posting the revised indicator, I couldn't have fixed the one issue, and we wouldn't even have an indicator that is reliable to be used for any strategy. So thank you again for posting the revised indicator. It would just be good if you, the original author, fixed it.

Don

Hello everyone,

I decided to make the indicator to have Averaged Gains by default (so set to True), and to not use Pips As Whole Numbers (so set to False), to be like the original Sentinal. The appearance is no different than when putting the first Sentinal on the chart, except this Sentinel_Revised2 updates properly because Kevin was kind enough to fix that. This works where Sentinel_Revised didn't because that indicator didn't take the difference in decimal places between JPY and others, so rendered the lines in the indicator incorrectly. This problem is fixed.

So that indicator Sentinel_Revised2 is the latest, and trustworthy that I know of. I did make it so that if put on a non-USD based chart of one of the six pairs, it will be "upside down" from what it looks like on usd-based charts. This was done so that the signals from Kevin's original post in this thread could be implemented into an Expert Advisor (EA). So when this indicator is under the moving average on one of the six pairs charts, sell the pair. On another chart at the same time it may be over the moving average, so buy that pair. This is named Sentinel_Revised2, and goes into the experts' indicators folder.

I have also added an indicator that is my attempt at helping automate the process of putting horizontal lines on the chart at beginning of NFP Friday and next Monday (Sunday 11:00PM Central). I also added shading to the non-trading area (red = no trades during NFP cooling period, orange = no trade zone after NFP cooing period).

There is something wrong in the highest high and lowest low that it's trying to get to form the horizontal lines. The horizontal lines look correct for this time (so far), but I also added the lines for last NFP/last month, so we could have a reference and also to help me check my math. The horizontal lines that are supposed to be for the highest high and lowest low during NFP cooling off period, these lines are wrong. This can be clearly seen from seeing price action on the chart during last NFP and seeing that the red and orange rectangles (no trades zones) are nowhere near the actual highs and lows. If this gets fixed, then this strategy can be backtested in an EA. This file is called NFP_Sentinel_Don and goes into the experts' indicators folder.

Speaking of EA, I also included an EA with the rules as best I can understand them for the buy and sell, although the closes need worked on. This goes into the experts folder.

Remember folks, this is a work in progress. But a work that is totally worthwhile. How would you like to make hundreds of pips in an automated way? If so, and you can help fix the indicator that is named NFP_Sentinel_Don, then please do help, and we can get any last things changed in the EA we need to.

Thank you everybody, especially Kevin. I hope you're doing good Kevin, and I hope I haven't offended you. It's not my intention.

Don

OK. I posted the updated indicator, and another indicator, and an EA. The other indicator and EA are a work in progress.

I see that NFP doesn't really affect us when using this indicator (Sentinel_Revised2), but it's still a good method to use the lines, etc. As I said before, it looked like the no-trade zone was working correctly for this month on NFP_Sentinel_Don indicator. Today, I watched as both EURUSD and GBPUSD both breached the NFP cooling off period's low. Both are headed down right now as expected after the breach. On a couple usd-based pairs, we see the Sentinel heading up, Sentinel is above 336 SMA, price of each of the pairs is above 336 SMA. USDJPY is moving up strong. USDCHF is moving down right now in the no-trade zone, but is going to go up along with the others (it's just taking it longer to "catch up"). Pics attached. We would be trading 3 of these 4 pairs right now with the money management in the EA, which maximizes potential.

The only thing that needs fixed is the highest high and lowest low of the NFP cooling off period, and then I could finish the EA (unless someone else wanted to contribute).

Don

Looking at the Sentinel_Revised2, if I want to see how many pips I could have made trading all six pairs (just using the Sentinel_Revised2 indicator, it's moving average and nothing else), I could make AverageGains to be False, and make PipsAsWholeNumbers to be True.

I now look at on 2008.09.12 17:00, where the Sentinal above the moving average (on a non-usd-based pair). It crossed again 2008.09.29 01:00. If I subtract what Sentinel_Revised2 says at one crossing from the other crossing, it's a difference of 1815 (9227-7412=1815). So how about making 1815 pips in less than a few weeks?

Then when it crossed that time, we would have gotten into a trade in the opposite direction. Between then and now, could have made another 3,648 pips in the last week trading six pairs (9227-5579=3648).

So a total of 5,463 pips in less than a month (1815+3648=5463)!

Now does THAT sound good to YOU!?

Hope so. Please help fix the indicator, and we'll be able to automate it with great money management. Actually, these numbers are not using the rules by Kevin. Like I said, this is the gains we would have gotten just using Sentinel_Revised2. So do we want to even fix the indicator NFP_Sentinel_Don? Why not? It does help to know where in the month we are, if just for curiosity.

Don

If anyone is interested I found the correct way in which they calculate the USDLFX.

USDLFX = ((USDCHF * USDJPY * USDCAD) / (EURUSD * GBPUSD * AUDUSD))^(1/7)
EURLFX = USDLFX * EURUSD
GBPLFX = USDLFX * GBPUSD

Now where is Kevin with that mixed basket of currency's candle chart script :-)

Although I did find this script to mix 2 pairs but the volume does not seem to merge together.

Dean.

Deanz,

Could you fix the indicator 'NFP_Sentinel_Don' that I posted in post #330? The horizontal lines that are to be for the highest high and lowest low during the NFP "cooling off period" (Friday when NFP released until Sunday 11PM Central, I believe, according to post#1 instructions) are not correct. I did the best I could. If they were correct, the EA could be fixed/finished to automatically trade Kevin's system.

I don't know what's happened to Kevin. Hope he's doing OK. I have just tried to move forward with the indicator/system since I found this thread, and do appreciate Kevin fixing the indicator to check that pairs histories are there, and setting the bars to use the lowest bar count of the individual pairs.

After I fixed Kevin's revised indicator to take difference into account of JPY being only .01 for pip instead of .0001 for a pip, then the indicator (Sentinel_Revised2 in post #330) works good now and can be used in conjuntion with an EA (provided also in that post, but not finished).

Don

Anyone following the rules and using the sentinel index as they were given on the first thread (as attachments)?

oops, I meant on the first *post* of this thread.

Sorry Don my programing skills are only very basic.

Maybe it is because you are using a Mac......

Have you tried asking in the Programming Discussion section:
http://www.forexfactory.com/forumdisplay.php?f=69

When I first started using metatrader I had people make me an indicator and they only charge $30-50. Maybe that is the way to go ?

Dean.

Mac has nothing to do with it

No, I haven't asked in the programming session. I didn't think I should have to pay for something I'm not gonna use. As I've said before, I use the indicator, but don't use KEvin's system. But if fixed, then the already fixed indicator (Sentinel_Revised2), the yet-to-be-fixed indicator (NFP_Sentinel_Don), and EA (also in that post) would be Kevin's original idea from post #1 made into an automatic system, with two indicators I last posted and the EA (Sentinel_Revised2 is correct, but the other indicator and EA are not finished, and won't get finished unless someone decides to do it, because I don't have enough skill to do it).

Didn't hurt to ask, since you are the only one besides Kevin and myself who has posted since I brought this thread back to life with my first post.

Cheers,

Don

There was an indicator made (Sentinal_index). It did not keep updated correctly on the charts. After the thread died, I asked if that could be fixed. I doubt anyone would be successful in using the original indicator unless they manually switched timeframes back and forth to have to manually refresh the indicator. The fixed indiactor Kevin provided did update correctly (Sentinal_Revised), but the line on the indicator appearance was wrong. I fixed the problem and posted a new indicator (Sentinel_Revised2), as well as a couple in-progress stuff - an indicator to show NFP "colling down period" start and stop, highest high and lowest low. It would be put on each individual chart for all 6 pairs. The highest high and lowest low are not correct, and would have to be fixed to be reliable. The EA has the basic rules in buy and sell, but closebuy and closesell are not finished. As I've said before, unless someone can fix the NFP_Sentinel_Don indicator, then the only thing we can reliably use is the Sentinel_Revised2 indicator, and the system must be traded manually.

Don

disbellj,

Thank you for your response.