Monday, May 16, 2016

How to build your own Set Files

I wanted to give a quick run down of the process that I go through to give us set files so everyone isn’t wasting their precious time “throwing darts” but making real steps forward with black & white hard data. I would like to help advance our EA’s (Blessing & Evolution) as well as help other developers/designers with this very efficient process of “tuning” our EA so more people can help in the process as there are just too many combo's to discover (which is nice!). So below is a ROUGH draft of the method I developed testing many-many EA’s over time. It will allow you to rather quickly learn, test and tweak ANY EA out there to the fullest of it’s ability. No mystery, its hardcore gathering data. I spend all day, everyday testing and writing down what the test showed, but this means almost nothing if not gathered in the right logical order.
Allow me to explain my common sense logical approach to tuning EA’s.

My “CHAIN OF COMMAND” METHOD

In the natural world, successful businesses, militaries or even just families have in place a “Chain of Command.” There MUST be order or there is confusion and a computer program is no different. It is this TRUTH that the whole method of EA code breaking with set files comes into play. We M-U-S-T find the Chain of Command, -the “hottest” settings to the least hottest and adjust them accordingly IN THAT ORDER, or it will throw off the “balance” of the EA and it will run “choppy” if at all just like things in the natural world without a tight Chain of Command do.
Example:
Take an “ON/OFF” switch of ANY EA, because the on/off switch IS THE HOTTEST setting if it is in the “off” position every other setting under it in the Chain of Command doesn’t matter, tweak on them all you want the EA is OFF and it does no good. See, how the Chain of Command works and the importance of it? Likewise EVERY setting has lesser settings that follow it’s orders. Our job is to find these “Alpha Male” settings and note them from Alpha to Omega, beginning to end. If we do not follow this order there will be chaos and this will show in the profit graph of a back test as it will be unsmooth. You’ll make no/none/ZERO “solid” progress forward day after day as there is no “logic” behind tuning your set file, and no logic means no “balance,” and no balance means it’s a gamble or like every other EA out there.
BALANCE is defined as the ability of the setting to “read” any market condition correctly with any amount of capital as it is self adjusting to the markets condition as well as handling the money management of the account.
The difference between indicators and an EA is that an indicator just "records" what it is programmed to do, it is never "wrong" as it can only record. An EA using indicators to look for recorded events that point to historical market conditions that can be seen recorded before they happened therefore we can profit from them. I simple example would be a Stochastic indicator and how it can be oversold or overbought and the outcome result of each...
Quick HINTS to keep in mind, then I’ll explain the method step-by-step:
-Never try to find a NEW set file with a back-test with a low amount of money, instead use at least $100,000 with Money Management "OFF."
Why?
A back-tests report “can” show MANY things. For instance, if there is only 50 trades and then an EP hit, what does that tell you? Well, not much, it MIGHT tell you that the setting you are trying is very out of balance, but if you used a $100,000 with Money Management OFF and a fixed lot size of only .01 and start your back-test this should be enough to get you over any DD area which is important because we NEED TO “SEE” what & how our setting is running so we know WHAT TO FIX. You may see that this was the only spot where your setting had trouble and that it ran very well the rest of the test and in fact ISN’T as out-of-balance as you thought. You then can zero in with a VISUAL back-test of that problem area and maybe you see 2-3 long drippy candles that caused this DD area like the USD/JPY or EUR/CHF pairs like to do. If that’s the case, then adjusting the “Delays” in our EA’s should take care of that problem and now you have a great set file to continue working on that you would have missed with a lesser starting amount.
In the above example, if we started with a low back test amount (or MM "on") we would have skipped right over the setting and thought it was “junk” simply because it crashed right away in the back-test. See now why the higher starting amount is NEEDED so we can “SEE” just what we’re dealing with so we can then absorb that into our setting and make it safer when trying to find new set files?
So never back-test with Money Management “on” when finding new set files. With ANY set file you can adjust it so that it’s so aggressive it’ll crash, so how can you know what is too much or too little aggression? We test and note where it starts to get "hot" so then we know for sure where the "comfy" spot of the setting is. What we want are set files that make it through the test period, we ARE NOT looking for profit when finding new set files, not at first, but then we will be looking to “tune” our EA in so that it can “read” the markets conditions correctly and then first see what it can yield in profit.
Profit settings come after we find set files, never before, -how can they???
  • Never use a STOPLOSS when finding a new set file.
When back-testing you never use a stoploss, as this will affect the profit graph and you won’t be able to “SEE” how the underlying EA is running. Back-tests where stop-losses are used are choppy, you simply cannot get a read on how your EA is doing if a stoploss is getting in the way. A SL was NEVER meant to produce profit, if a EA has a stoploss being used to “help it” make profit that is simply a underdeveloped EA. A SL is only used as a LAST DEFENSE should the market throw something at the EA that it wasn’t programmed to deal with (underdeveloped).. Notice our EA’s are very advanced, even the free B3, and we don't have a "stoploss" as they just aren’t needed as the set file is developed and the settings matched to the base setting. Not that we run blind with no safety, no way. We have Max DD% protection levels among other things.
  • And any type of stoploss in an EA is ALWAYS THE VERY L-A-S-T thing to finish your set file as it has NOTHING to do with how your EA runs but is there as a LAST safety device. EA’s on the net that attempt to use a SL as a active part of their set files are giving away tons of profits and their profit graphs in back-tests are choppy because they are way under developed and should be avoided. The more a EA is developed (logically, -Chain of Command) the “smoother” it will trade as its program is more advanced/developed so it deals with market conditions better.
Let’s start, ---Here's what you do.
  • Use 1 notebook for each pair. You will find that the better you keep your notes the more use you’ll get out of them later as you read and go over your notes perhaps looking where there might be a “fork-in-the-road” that you can return to, to perhaps find a better setting. I do this all the time as there are “forks” at each major “step” within the set file. I’ll define these as we go on.
*I highly recommend getting several different colored pens and highlighters, one color for WHAT is being tested, another color for the RESULTS and another highlighter for BEST result. This will allow you to quickly flip back and watch each pairs development and clearly see the “forks” in the roads as you went. You'll be able to quickly SEE "where & what" a set file was born out of which may give a clue to another set files tweaking... This spot will also be a pivotal "fork in the road" where you can return to and try a different path and see a different end outcome (there are MILLIONS of these pathways mixed in the settings to find and explore).
If you’re totally new to the EA, start with 1 setting at a time, notice how "touchy" (hot) it is and rate it, 10 being VERY HOT meaning a small change can EP hit the whole setting. After you have ALL this data on how touchy each setting is, look at which ones are the "hottest" as those are the ones you want to adjust on FIRSTit M-U-S-T be done this way. Do not adjust a not-so-hot setting and then expect to "draw" the higher settings to it, -it CANNOT be "balanced" if we do that. Remember the “on/off switch” example and how it didn’t matter what setting is used if the higher, hotter setting wasn’t “on” first. Map out the Chain of Command, that is your road map to success, without it you’re throwing darts in the dark as the set file will not be logical…
This sounds complex but this is very simple, just do things in the right order is all.
Now, the HOTTEST setting, adjust it up in logical “steps” until it busts and note at every step the:
  1. Profit
  2. DD%'s (Max & Rel)
  3. Win/Loss ratio
  4. Total trades
Let’s talk about Steps as they can be tricky:
[This is where a optimizer can be used with "stepping" different settings looking for a door opener, but ALWAYS double check your findings!]
The Steps are the spacing between numbers that will be tested to give you the data to show you where you need to be looking in DETAIL and at every step. Remember to start with the “hottest” setting… For this example let’s use the GAF setting. If we have a setting of “1” for this we need to know "where" this setting is within the setting, -is it "hot" or cold for example? First run a backtest so you have your "benchmark" to beat. We’ll then test in “steps” with .1 steps and write down the data (1-4) as we did with the benchmark setting. Then test from .1 all the way to 1 (10 tests) the benchmark setting, and note what happens to PROFIT, DD% max/rel, Win/Loss ratio and Total trades at EACH STEP.
Profit of course tells us how well the setting is taking profit and keeping it.
DD%’s will tell us how well the settings are entering trades.
Win/Loss will show the sharpness of entry/exit.
Total Trades will show us how “tight” the settings are, -or how slopped out it is, or if it's "sleepy."
Then adjust the GAF up and do the same, "stepping" the settings upward until bust and filling up pages of data if need be, -data that "shows" where the profits are and where they are not in black & white in your notebook. Now we do this with every setting, taking data by "steps" as we go. Some settings (grid) might be tested in steps of 10, while a GAF setting will be tested in steps of 1, or even smaller, 0.1 or .01 or .05.
You may find that the best setting is a 1.2 GAF for now….
If a setting is set at 3, test at 2 and 1 and even 0.1 to see if it works, -if it does.... -does 0.2 work also??? –if yes keep on even checking to see if .09 works, explore and learn what this setting does by doing these sometime countless tests and NOTE what they do (1-4). You WILL be re-reading these notes later as you “fine tune.” Sometimes there are wide ranges of settings so take “steps” of 5 or 10 as mentioned.
Example: 110, 120, 130,…
Now work your way to the LEAST HOT setting and as you go leave behind you the "trail" of the best settings behind you. Don’t “cheat” and over step settings when testing, you may “miss” a doorway. Note also that with every change in the setting that these are the MAJOR “forks-in-the-road” I was talking about.
So, the “trail behind you” in other words, are the settings that had what you wanted, be it low DD% or Profit or? You KEEP PICKING OUT what you want as you go and that is how you line up a “BASELINE.” The baseline is what all the other indicators are tweaked toYou can baseline low DD settings or baseline the best Profit settings or even a blend like DD% & Profit (my favorite) for a baseline.
So “Baselining” is finding every settings best setting by using the steps data to find what setting each one gives the best result with according to what your baseline is being used for. A simple example would be “Profit.” If you wanted to “baseline” Profit you would work through all the settings and use the settings that produced the best profit for each setting. You would have then “baselined” all the settings to “Profit.”
I’ll repeat, you can baseline anything, DD%, Profit, Number of trades (for you rebate cashback hounds!) or combinations can be baselined. But something MUST be baselined because all the settings have to be pulled to it. That gets rid of the chaos and by doing this in a Chain of Command way you work in the Balance (smoothness) which will be seen in your profit graph of your backtest.
All settings in order of the Chain of Command remember!!!!, TO THE LEAST of the settings. -Or remember they won't be "balanced" and will spin out and a crash will happen. It has nothing to do with a EA being “good or bad” but how you told it to trade.
Let me say it Again, adjusting the "hottest setting first, to least MUST happen." Or you’ll end up with crappy fluky EA settings like all the other guys on the net. Take a look at FapTurbo, this monster coded EA trades only the slow moving Asian market period, and yet they still have yet to produce CONSISTANST profits and try “band-aid” after band-aid fixes but don’t know about “balancing” settings. Instead their approach is to give their followers “new” set files monthly, ---great marketing, but lame. What a joke IMO. Their massive code can’t “read” a slow Asian market trading session, so THEY have to do it by looking (I’m guessing) at higher charts like weekly/monthly and adjust the settings MONTHLY to try to slop out a “lucky” profit. –unreal ~
Would YOU trade YOUR real money like that? Gambling month to month with what someone "thinks" will do good and that changes month by month?...
Could you imagine B3 or the highly advanced Evolution having to be “adjusted” monthly as time goes on?... LMAO ~~
Wouldn’t that be admitting it doesn’t read the market for crap??? Well wouldn't it????......
I want people around me to KNOW what I do so they can start helping, because EV’s combinations for "tuning it in" are limitless now, even B3’s abilities can clearly be seen with honest study. But with all these endless combinations of settings and different pairs (as well as Metals and even stocks like Google as I’ve demonstrated) I simply "can't" do every test that shows me every combination, so an ARMY of "me's" are needed, pure & simple and we hope that is what our "Club EV" does.
I have no reason to keep any secrets as far as what or how I do my job here as Chief of Development and am looking forward to working for years to come with these EA’s that we have and the people who run them. You’re time spent learning them won’t be a waste as the theory of their trading basics can be used in Forex or the Stock market, -and as further developments take form and different EA’s come out you’ll already know how to drive them basically.
There are no “Glass Ceilings” here, and this is one of the reasons why I love these EA’s and why I would love to have a group of people testing, developing and breaking the codes to solid safe profits for years to come.
Now, let’s talk about some troubles that you will run into with the Chain of Command Method: Some settings are pretty much as touchy as another so which do you do first? Also there are some (like grid) that have multiple settings within itself. This is like a whole tuning area within a tuning area, -so how do we tune it?…
If there isn’t a CLEAR defined Chain of Command (ranking) then settings that are out of “balance” will happen. This can usually be cleaned up by adjusting down the MM system. The “cleaner” you can align the Chain of Command the more the setting will want to scream with profit. How do we make them run cleaner???
By revisiting our notes in our notebooks, we look back at the forks in the road and see if there might be a place where we could back up to and start down a different path.
You see each change of a set file changes all the others. Even once you make it through an entire setting and have found all the best settings and have lined them up with your baseline in a Chain of Command order, -you’re not done……
You must now go through the whole order AGAIN… –and yet AGAIN if “any” setting is changed again. Because each setting will need to "settle" into place and what "was" the best setting may not be the best setting by the time you're done going through the whole set file.
But as you go through the settings each time you’ll be able to scan smaller “steps” a couple out both ways up & down and see if in fact the setting that was changed affects the other settings enough to open doors to allow tweaks to their setting. But careful, sometimes you’ll find that a setting can change another setting A LOT, so much that another setting might have to be changed a lot to allow the new doorway to be opened. So just keep in mind that when re-going over the settings after your first time through them it still is better to do a lot of testing of the steps in both directions, and not just a few out each way. You will find that you’ll get a “feel” for things the more you work with it. You should start to develope a feel of if it's running "comfy" or is stressed.
So these area’s are the “forks in the road” where we come back to, to find different roads to go down if we feel the path we are on isn’t where we want to be, and you’ll KNOW it because you’ll be fighting with a setting to much, like it won’t want to become smooth in the finer settings. Because of things like this the balance is affected, but can be offset by lowering LAF or the AGGRESSION of the setting but that’s kind of cheating. It’s ALWAYS better to get it running as “clean” (smooth) as it can.
Multiple Steps within Steps, how to test them:
First, let me say that the advanced settings allow EV to decide for itself where to place trades by the filters it uses. Usually a grid setting is used here to make sure trades don't go off to close together you'll see this more with our "1 Trader" set files.
Now back to multiple steps in steps... Example, our grid has different AMOUNTS of trades, LEVELS of trades and even different Take Profits amounts for each, so how can we set them to a logical interacting setting?
One way, is to Cross Hatch the data.
Test the first setting in steps and note the best, leave it at that best setting and go onto the next. Do this with the whole line. Then go back and retest the entire line, if no more tweaks are found move to the next line. Do the same here, step testing each as you go leaving the best setting behind. Do the same with the Take Profit line.
Then pretend these are just one line and retest and allow each to double check the other. That is, you may have to jump back and fourth up and down the line to try to get the settings to compliment each other. They will slowly weed out which settings are just “hot” and need to be replaced. You’ll know these if you change a setting and it crashes the whole. The more “comfy” the settings are together the move they can be moved around and not crash the back test. The “touchy” hot settings that crash your back test are the ones that are out of balance the most and where your attention should be.
Why $1,000 is used as a base for our backtests.
$1,000 is used because this seems to be the amount that passes when starting right before the biggest historical DD area's and gets through them with no problems. We wouldn't want to just start trading our EA and get wiped out because it turns out we started right on a trend change DD area so the $1000 is our insurance against the unknown. So from now on back tests won't be posted with a lower amount then $1000, although they will go lower and pass most times. Sometimes as low as $100, but should your account not have enough time to build equity to survive a DD you may bust, -take the “insurance” and start with $1000 per pair IMO.
If anyone gets a new setting to pass a back test with $1000 for a new pair please post. But remember for SAFETY’S sake we start with a balance of $1,000 as no DD area “should” be able to shut us down by using this amount, --- PROVIDED our starting Lot size is .01 and not extremely high because of a hot set MM setting…. But remember when trying to find totally NEW set files start with using a bigger starting amount when back-testing ($100,000 with MM "OFF") so you can SEE what is causing trouble and fix it. A $1000 start is for tweaking in a set file that already passes the test period.
Of course even if you’re just having fun and “throwing darts” and changing settings and feel you got lucky, post what you found. There are nights when I “throw darts” myself. I know this is worthless as an approach but there is always the factor of blind educated LUCK when doing this. But as you can hopefully see now from reading this, “luck” isn't the method of choice with our EA’s, -we'll leave that to FapTurbo’s monthly guessing game set files...
Our EA’s “read” the markets in every trading session, news or no news, gaps or no gaps because they are programmed that way, and we’re ever learning and knowledge never stops with the development team under Jeff Talon, from the best free EA on the net (Blessing3) to the sophisticated Professional “Evolution” EA.
We “read” the markets, not guess at them.
What drives me is that I can sit down every night and get one more step closer to having these EA’s run so steady and smoothly that more & more they become my source of income for my family. I welcome anyone who’s serious along for the ride for as we make new benchmark settings together upon new benchmarks we’ll together realize some of our dreams start to materialize because it’s a progressive method, and not “dart throwing.”
May our accounts always be GREEN my friends!
Ratz
Real Analysis of Technical indicatorZ.