Sorry to beat a dead horse. I'm still experimenting with this, to find the best (or most correct) setting for my (apparently nano) IBFX account:

I am currently swapping back and forth between 2 different Assets.dta files. One is directly from the broker via Download script. This one lists the assets apparently in nano lots, which we know is confusing to Zorro due to the issues mentioned above.

One workaround I like is to force in the Zorro.mq4 script:

which (I think) forces Zorro to essentially take 10x the trade lot size (ie 10 nano lots = 1 micro lot I believe... so 10 cents instead of 1 cent).

The other workaround discussed above is to modify Zorro scripts by forcing more lots, such as "Lots = 10".

At first glance, you would think these 2 workarounds should have the same effect. But in trying to figure out why the test results were sometimes very different, I came across this:

Here is a snippet from Assets.dta when I let it pull the ACTUAL nano account info from broker:


And now look at the same snippet when I've forced the Lots x10 in the Zorro.mq4 script:


The difference, I believe, are caused from the rounding in PIPCost. Too much value is lost if you start with a rounded value (nano) and multiply to increase the size. Seems more accurate to do it the other way.

In thinking about this, it seems to make sense. In fact, philosophically, it seems like a bad idea in general to test with bare-minimum lot size if rounding would be an issue. For example on spreads and rollover, etc, broker is always going to round up to the next penny. If you're only trading pennies to begin with, then your results could be way off. Half a penny could represent 50% difference lol.

So my (long winded) argument here is that... for testing (and trading), that nano lot accounts should not be sliced-and-diced all the way to the penny level. It is not too optimistic to increase lots flatly by a factor such as I have done. Rather, it is helpful because it allows more accuracy in testing with less rounding. (Please argue otherwise if you don't agree)

THANKS

It is better to increase the number of lots in the script.

If you increase it in the account parameters, you end up with wrong parameters. If you then trade the system live, a given margin setting will open ten times more lots than it should, and thus trade ten times more risky.

I agree, with regard to actual trading... but for the purpose of having accurate testing... how can I resolve the issue of the PIPCost being rounded?

In the example above the PIPCost is 0.0961 --> 0.0096 but I have seen others that are much worse, for example what if it was 0.0964 --> 0.0096 ? That seems too inaccurate doesn't it?

Do you know if I can just manually tweak Asset.dta to have a 5-digit precision? That way I could get the best of both worlds. So in other words, keep the correct Asset.dta nano lot settings, but provide the 5th decimal like this: 0.00961

If you're not sure, I can just try it and report back. THANKS

You can manually tweak Assets.dta by just editing it and entering the higher precision number. Nano accounts need 5 or 6 digits precision for PIPCost. We'll change that in the next version.

Awesome... thanks

I noticed that the precision doesn't make much difference for major pairs like those included by default with Zorro. However, for some exotic crosses it does seem to make a difference.

Also, I noticed that Zorro (or perhaps the broker?) seems to not be rounding correctly when producing the PIPCost in Assets.dta. It seems to be just truncating the decimal places.

An example is this:
When I download IBFX assets by forcing LotFactor = 0.1:


But when I download IBFX with its correct LotFactor of 0.01:


So I would prefer to see 0.00789 (or even more precision), or at the very least it should be rounded to 0.0079

Please check other critical figures also for rounding issues, in case you find that this is a Zorro issue.
THANKS

I'm attaching here the Assets.dta file that I've tweaked for US-based IBFX mini* accounts. I carried the precision to 2 more decimal places in both the PIPCost and Margin columns, based on information from the actual broker (via Download.c script)

I believe this has improved my testing accuracy slightly. If you agree this is all setup correctly, feel free to include it with a future Zorro release to help other US-based IBFX traders.

*NOTE: IBFX only offers "Standard" and "Mini" accounts. The account called "mini" is actually a micro account, allowing nano lots of 0.01 (ie, 1 cent)

The next Zorro version will store PIPCost with higher precision.

Computer languages don't round numbers when they store them in text files. Although this could theoretically be implemented with special functions, increasing the precision is normally the better way.

Hello, once again I must apologize for digging up this dead horse thread... but I've only just realized today that something has changed regarding Assets.dta (I'm currently on v1.14)

The Assets.dta that I prefer came from my IBFX broker and is 2 posts above this one.

However, in the current Zorro version it seems that as soon as I run a backtest, Zorro modifies Assets.dta. It seems to reduce precision and also round-up both the PIPcost and Margin column.

I'm ok with this, as long as you can assure me that rounding-up is always the preferred pessimistic choice. If it could introduce too much optimism then that would be a problem I think. Here is an example:

original Asset.dta


modified by Zorro after first backtest


THANKS

Hmm, PIPCost and Margin are neither optimistic or pessimistic. So the rounding does not introduce pessimism. I'll look into this for the next version, it should not round but keep the original precision.