Welcome to NexusFi: the best trading community on the planet, with over 150,000 members Sign Up Now for Free
Genuine reviews from real traders, not fake reviews from stealth vendors
Quality education from leading professional traders
We are a friendly, helpful, and positive community
We do not tolerate rude behavior, trolling, or vendors advertising in posts
We are here to help, just let us know what you need
You'll need to register in order to view the content of the threads and start contributing to our community. It's free for basic access, or support us by becoming an Elite Member -- see if you qualify for a discount below.
-- Big Mike, Site Administrator
(If you already have an account, login at the top of the page)
Use something like a 10 tick Chart, and observe the response of the
Trade Flow Indicator.
Remember that, as coded, it is REAL TIME only; and does not pick
up historical data. Converting to Historical requires MUCH more
coding than simply enabling things during the data loading phase;
since a virtual timebase would need to be established, etc., so just
be content with Real Time data analysis.
As has already been pointed out, this is designed for shorter
timeframes; whether Interval or Tick; but I haven't considered it
on other Chart configurations.
It could be a bit more interesting; just an idea...
It's probably best to stick to 5 second charts. When you use non timebased charts they have a tendency to optimize towards trades that you aren't likely to catch. Like by the time your system has finished the calculation and sends the order the opportunity is past. Time based bars backtesting is far more accurate in my own experience.
@hyperxtrading
By the way, while reviewing Friday's charts I noticed something that does not make sense to me.
In the attached picture I was running 3 instances on the same chart of the TradeFlowTutorial with the following "parameters'.
Top:
RETENTION_SECONDS=60
BIGLOT_MINIMUM=10
Mid:
RETENTION_SECONDS=30
BIGLOT_MINIMUM=25
Bottom:
RETENTION_SECONDS=60
BIGLOT_MINIMUM=25
I was expecting to see different absolute values on the same bar's "time stamp" on each of the 3 indicators. See left "Y" scale on charts.
However, the Top indicator and the Bottom indicator show identical values no matter which bar we are in. Of course they change value as we move to anther bar, but they remain identical.
My conclusion is that the BIGLOT_MINIMUM value is being ignored, (or there is a bug in the code).
I would expect larger swings on the 10 BIGLOT compared to the 25 BIGLOT.
JT TradeFlowTutorial_Params_delete
TWYS NWYT(Price Advertises Opportunity; Time Regulates it; Volume Measures its Success/Failure ---- Dalton)
THANK YOU for giving the indicator your attention, and for getting a "feel" for what
might be expected.
I'd say a couple of things to check:
1) Be certain that you do have UNIQUE files for each Indicator instance, which
I'm sure you do have.
2) I've run 2 instances with varying parameters, and I don't see that there's a
bug in the Big Lot Filtering myself.
3) I use Notepad++ and in your case, I would have 3 Tabs, each open on one
of the individual parameter files; make changes, but do not forget Cntl-S
for Save.
4) Also, if there happened to be an exception in Parsing of the parameters file,
I am not certain the error handling would always be perfect; so perhaps parameter
values you expect, for some unknown reason, are not actually applied within
the Indicator instance ??? Always a possibility you could keep in mind...
Sometimes, interactions between Retention interval and Big Lots, and dynamic
market data incoming; can produce results which as you are point out, might
be slightly counter-intuitive, but have a reasonable explanation on closer
examination... ?
CONSIDER THAT YOUR ASSUMPTION THAT SWINGS FOR SMALLER BIGLOT
THRESHOLDS should be larger than "swings" for larger thresholds. Think just
for a moment that this is ONLY a simple SUMMATION of all of the *SIGNED*
(Buy+ versus Sell-), so consider that a Lower threshold consists of more
data which may drive the integration (summation) back toward zero; whereas
larger threshold values give the Indicator an opportunity to stay at a more
deviant value, for longer ..... THIS COULD MODIFY YOUR INTUITION, ...
Also, the code SHOULD "taper" the final 25% of the Retention Interval Window,
linearly to ZERO as items exit the fixed window of analysis in real time.
BUT YOU HAVE ALL THE CODE, AND it should be a simple matter if I have
given you Bugs... sorry in advance if there is a problem with the code, but
I'm using it, and getting what seems to me to be correct results... ??
Of course, it goes without saying that the code is for Real Time data only; and will
not work for any attempts at Historical backtesting analysis, as written.
[EDIT] One more thing, I'm familiar only with NQ but I think you're using ES,
which will have a different "Mix" of BigLot sizes, and of course time of day
will also influence the mix of Big Lots. (A histogram analysis plotting frequencies
of Big Lots would be fairly easy and interesting here) Opportunities to expand
on the underlying ideas behind the Indicator's purpose. It's a bit more "subtle"
perhaps than we would think, at first, this idea of Big Lot Analysis as signalling... ?
[EDIT] Another idea for easily enhancing for "Super Lots". Add a DOTTED or dashed series RED for any SELL BigLots which have an exceptional avg size multiple, and give that series a ZERO, or a -100 for detection of the Super Sized Sell. Add another series DOTTED in Blue maybe, representing Super Lot sizes for BUY transactions, and give that series Zero or a +100 in the OnBarUpdate,
to easily designate these Super Sized lots which have been detected.
I'd have to check, but the idea is the ability to draw one DOT per OnBarUpdate,
or similar concept, given NinjaScript's line styles, etc... there's always a way !!!
I see it's easy to draw a DOT, so I'd probably use that to indicate SuperSized lots.
Well, just spitballin' and brainstorming on some possible enhancements to the Indicator...
Maybe I'll make that enhancement and post the code for it... Who knows? LOL
Inventory Analysis, which I've done extensively; goes far beyond
the "simple summation" approach that we've been using in this
"cumulative delta" with thresholds for Big Lots so far........
This is a discussion as short as I can make it, of how Inventory
Analysis with "Risk Evaluation" is different from a simple
summation or integration of Trade Volumes...
First of all, the simple numeric summation of a Collection of Signed
Trades does provide us with quite a bit of useful information; and it
may be that more sophisticated analysis, adds only marginal value.
That's often the case, so just consider that.
WHAT ARE THE KEY RESULTS OF EXTENDED INVENTORY ANALYSIS?
The key value of a more extended Inventory analysis, would be an estimate
of the VWAP (Volume Weighted Average Price) of Inventory, and whether
Market Maker's "holdings" are "making money" or "losing money" as
measured against the VWAP, and just how much (temporary) loss
Market Maker might be experiencing.
Of course these are "estimates" expected to be correlated with Market
Maker's situation; and there's a limit to how much we can extract.
However, just because something has an irreducible "error" to be
expected, does NOT mean that it isn't "useful" and can yield "Tradable
Indications" or signals...
We don't live in the world of the perfect; but often "good enough" gives
us an edge...
WHAT WOULD BE THE RESULT, if we implemented the following algorithm:
When the simple Integration we already have is Positive, then let's determine
the VWAP Price of all of the BUY transactions ONLY? So just do a simple
VWAP calculation to determine the PRICE of the BUYing subset of the
transactions we hold? REMEMBER, that we are looking at Retail Buyers,
AGAINST WHICH Market Maker has SOLD the inventory...
VWAP is SUM( price * volume ) / SUM( volume ) where volume is always Positive,
or something like that... Volume is always a positive quantity.
We would then have a (crude perhaps) estimate of Market Maker's average
SELLING Price (against Retail Buyers) and could plot that against Actual
Market Pricing. I've called that Market Maker "Short Risk", again from
Market Maker's perspective.
When actual Market Price was significantly ABOVE this VWAP for Retail Buyers,
then we could suggest that Market Maker would be "losing money" and
we hypothesize some reluctance to continue to raise Price.
When our simple Integration is negative, then we would perform the calculation
but only on the Retail Sell transactions and form a VWAP (the volumes here
are negative, but we use the positive volume for the calculation) and so
we then arrive at a Price. If actual Market Price is BELOW this Price, again,
we suggest Market Maker is "losing some money" (on paper) against the
Price at which Inventory was Bought from Retail Sellers by Market Maker.
As market price fell below this Price, we'd say Market Maker experiences what
I've called "Long Risk", as we think of Market Maker Buying (accumulating)
inventory from Retail Sellers, and then lowering the Price.
THIS IS JUST A LITTLE GEDANKEN EXPERIMENT, but it might be a super
simplified way to get at this "Risk" that I've calculated many times myself.
NOW LET'S EXPAND OUR MINDS A BIT BY CONSIDERING THAT all Market Action
is "Fractal" in the sense that Market Maker is Trading against the Aggregate
of the Retail population; AT ALL TIMEFRAMES SIMULTANEOUSLY.
I've done this stuff before, with Multi-Timeframe Inventory Analysis but, in
the end, there's just a fairly simple rule: CHOOSE THE TIMEFRAME which
most closely matches your desired Trading Period, which avoids the anxiety
of having to think about ALL (so-called Fractal) Multi Timeframes over
which all this crazy activity is occurring...
WHY IS ALL OF THIS IMPORTANT? OBVIOUSLY TRADING IS A BATTLE BETWEEN
BUYERS AND SELLERS, where everyone has the goal of Buy Low, Sell High(er)
or Sell High, and Buy Low(er), which is how Money is made.
Just some things I've considered over the years, and which can often give an
Edge in understanding the extremely complex nature of Market Dynamics
through Volume Based (Inventory) Analysis.
On Monday I will resume testing with live real time data using @TWDsje's suggestion, as well.
As further clarification from my part; the chart shown earlier was run with real time data on Friday ( I did not shot down NinjaTrader8, hence it is still running to this minute -- with no more new data of course). So, the charts have not changed.
I do have a unique file to run each instance of the indicator and do not touch it throughout the session.
I use Nija's Continuum data, which I believe is CQ and not the better Rhitmic data.
For now I am analyzing with ES since that is what I use most often with other software development. It seems that ES moves in a more orderly fashion than NQ. Also, for this analysis, I believe there might be more larger contract sizes than in NQ.
With more larger size contracts I also expect more frequent greater than 5 BIGLOTS than greater than 25 BIGLOTS. That is what meant to say when I mentioned "larger swings". Hence, I was expecting the numbers that appear on the left axis to be larger in absolute value on the 5 BIGLOTS indicator.
I have attached another chart, which was ran simultaneously to the previous chart. It is just to "beat a dead horse" -- no inhumane treatment of animals implied. This indicator, all had a different BIGLOG size with a constant RETENTION_SECONDS. Notice they all have the same left margin values.
LAST but not least. I will be very delighted if you go ahead and code the enhancement for "Super Lots". That would save me time, so I can concentrate on analysis.
JT
TWYS NWYT(Price Advertises Opportunity; Time Regulates it; Volume Measures its Success/Failure ---- Dalton)
I spent a couple of hours enhancing the code so that we could take
an (educated?) stab at estimating Inventory Risk on the part of
Market Maker, and Indicate that Risk on our NinjaTrader Chart.
I do a lot of these little enhancements; so it's pretty easy for me.
The question will be mostly how best to represent this information on
the Chart, and after I've been able to get some results, then I'll
post for code for you guys to study, and try out for yourselves.
NEW FEATURES
1) The Integrator still handles Volumes as before, but added to it is the
ability to Calculate VWAP Pricings for either or both Positive and Negative
Trade volumes, including "tapering" the Price weights as we've done before.
2) A SimpleRateLimiter object (for FREE !!! LOL) which limits the Rate
at which certain operations can occur. After all, sometimes incoming
data can cause performance issues, and it's very often a good idea not
to allow updates or redisplays that are TOO FREQUENT. Yes, yet another
little "innovation" many of you guyz may not have considered, and
most of which will say "Why all the Complexity?" C'est la vie, as the
French would say
Hey, I got nuthin' but time to code Indicators for all comers... LOL seriously, happy to do it.
I still think that maybe your intuitions are geared toward an UNSIGNED cumulation,
rather than a SIGNED cumulation (integration)...
That may be why you're not seeing what you're expecting... No criticism implied here; it's
just that sometimes things are not obvious until we get a better "feel" for the type of data
the algorithm is (faithfully?) rendering ?? LOL
I doubt that Time and Sales is significantly different between CQG and Rithmic sources,
but it's worth keeping in mind as a possibility. When it comes to Market Depth, there's
"no contest", of course with Rithmic.
Yeah, OBVIOUSLY, the lower the BigLot threshold, the more "hits" there will be, but
do remember these "hits" are a mixture of Positive (Retail Buy) and Negative (Retail Sell)
volumes which are being arithmetically summed up..... Just sayin'....
I've UNILATERALLY decided to focus on estimation of RISK in a quick 'n dirty fashion;
as opposed to prioritizing Super Lots; but yes that would the next thing to represent.
LET'S BE CLEAR HERE: Despite the amount of time I've spent on this sort of thing
over the years, it's amazing how revisiting things through offering up these examples,
can sometimes, even for me, give me new ideas, and get me to re-think. So I'm not
completely benevolent and charitable here; I'm also benefitting from giving it
a "second look" LOL
