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)
Attached is a nice illustration of "Support" indication,
which involves Inventory Imbalance and Risk. And
let's not forget Resistance at the Tops...
Let's just focus on these images. TWO instances of
TradeFlowRisk are operating simultaneously, and
are "drag overlain" on the main chart.
(Sometimes you need to specify that the Indicators
use the LEFT scale, since dragging doesn't always
put the scale on the correct desired side)
I like to Manually Scale the left axis with Zero Net
Inventory in the middle of the scale, and then
"drag" the scale up and down to get a reasonable
Net Inventory Deflection.
If one of the traces is "too small" in deflection, then
use a larger Multiplier value.
Also, if the indicator shows "too many" Risk Dots,
raise the RISK_THRESHOLD value.
The Numbers being shown on the Risk Dots should
be proportional to "Total Risk"; the more, the better
is the Support (for Long Risk) or Resistance (for
Short Risk)
This TradeFlowRisk Indicator can be used to confirm
our intuition that "Market Maker trades Against the
Retail population".
You'll notice over time, in general, that Major Trends
will continue until the Retail population is Strongly
shifted into the Long Risk scenario for "Support"
and into the Short Risk situation for "Resistance".
Params are:
# Faster Blue Net Inv line
# alternate color scheme
RISK_THRESHOLD=20
RETENTION_SECONDS=60
MULTIPLIER=4
BIGLOT_MINIMUM=2
TAPER_SIZE=False
SUPER_SPIKE_THRESHOLD_RATIO=5
As I've said before, this goes to the Heart of Trading,
since it reveals how Market Maker tends to drive the
Retail population to one side or the other side of
the Inventory imbalance; before significantly
changing trend.
Yes, Market Maker does trade AGAINST you,
and this is a way to explore the dynamics of
that by approximating or estimating what the
Net Inventory and Risk situation is.
[edit] I have a sophisticated Triggering system
which is able to trigger on several criteria which
occur together. In fact, I use Net Inventory
maxima/minima plus Big Lot single trades
and measures of Price deflection to trigger
on Tops and Bottoms. But I need to factor into
my own system, the actual Total Risk values
for even more accuracy.
I should say, that then you have identified
the "theoretical pivot level" but then you have the
problem of tolerating "Run Through" or Price
running beyond your theoretical turning point.
To handle this, you must be able to distribute
your trades across potentially a 20 Point
Price Adversity, thus improving your Cost Basis
without having to Stop Out; while you are
waiting for "the turn" to take place, but that's
a whole other area of trading technology...
[edit2] Realizing it's not fair for me to mention
something without showing what it looks like;
this is my rendering of this stuff. Circled are
Big Lot Retail Sales; and the arrow points to
the Purple line which is Long Inventory; and that
is exactly where my system distributes a Range
of Price-staggered small MNQ contracts to BUY.
So, despite more than a decade of using these concepts of
Net Inventory and Total Risk as critical Trade Signals...
...I find that I neglected to put Risk calculation into my
own Triggering system !! Duh... LOL
So I implanted the guts of the integrator into my custom
Analytics and Order Entry system, and am now wiring in
the controls to use Risk Levels as one of several Trigger
criteria.
But even without Risk, I found today that Triggering on a
Short Term (like 10-15 seconds) Retention Interval
typical Max value of Net Inventory, AND requiring a
Big Lot Trade at the same time; plus some Price deflection;
makes a fabulous Trigger condition for Nasdaq Scalping...
(These are "counter-trend" entries, anticipating the
exact turning point of the market, in theory... and often
using a range of Micro contracts staggered in Price
levels, to tolerate any "run through" price adversity.)
Sometimes one can surprise oneself, even after all these
years
When comparing Multiple "Simultaneous" Conditions
which are required for Triggering, it's essential to allow
a "fudge factor" to facilitate the Overlap in Time of the
multiple conditions.
So, when a Condition is True, we take the Timestamp and
we hold it True for a Minimum period of time. The same
for all the other conditions. This allows for Conditions to
be evaluated in an overlapping fashion; without being
missed due to a few milliseconds of non-overlap.
bool isTriggered = C1 && C2 && C3 && C4; /// ... && Cn
// if isTriggered, then execute a Trade action
Now, if NinjaTrader's typically used standard ATM Strategy
had a convenient plug-in API, then the TradeFlowRisk
Indicator could trigger a resettable "one shot" Trade
fairly easily.
Unfortunately, even if a Trader has clear idea what needs
to be done, the technical nature of the coding to achieve
the goal is beyond a non-programmer.
I suppose that reinforces my view that Traders should
jump into coding, using Indicators as a starting point; with
an eye toward partially automating some decisions so that
a Trade can be triggered precisely where a Signal suggests.....
So, only you can acquire the skills to be able to do that;
and for short term scalping, working with the concepts of
the Trade Flow Risk analyzer can provide some good signals...
SO DIVE IN, THE WATER'S FINE, and learn how to wire these
Indicators into Trading actions...
[edit] I see that NinjaTrader does have an external
ATI (automatic trade interface) but I've never tried to
use it. Maybe that's a possibility but probably fairly
complex...? Maybe somebody has simple examples
where an Indicator could trigger? Not sure if that
would be the appropriate solution unless you were
generating signals completely outside the NinjaTrader
execution environment...?
From within a ninjatrader strategy I would probably just use the CrossAbove/CrossBelow functions. It will return true if the value crosses your trigger within a specified number of lookback bars. So you can combine multiple such triggers together quite easily. I also often find myself creating additional data series to track trigger conditions. IE set the value for that bar to 1 if a large order was detected in that bar and 0 otherwise.
I'm thinking that MarketMaker indicator aside, the Inventory information is probably the some of the best
that this provides..
Its also so obvious and logical that in a First Hour High push (like the standard friday long above that high value today) the long
should be a safe bet until SOME short inventory risk comes into play.. When there's both sides with inventory risk is when it
gets interesting... There's so few things that are logical in the indicator world - It stands to reason that the best information
isn't an indicator ....(yet)
Hi. Thanks for the good words. Several things to keep in mind:
0) We are, in fact, measuring trades by Retail players; against
Market Maker... trivial cases aside, this is how Trading takes place.
So we're saying that the majority of Retail players Buy the Ask,
and Sell the Bid, and you are NOT a Market Maker unless you
NEVER do that... by definition.
1) persistent trending due to external factors; will "swamp"
any technical indicator; which includes TradeFlowRisk.
(NOTE: THE REASONS FOR THIS ARE MANY, among which are
"deeper dynamics" which this indicator doesn't process.
Think about "spread revenue" offsetting "open risk" and that
is basically what's going on. And also, the "fractal" nature
of a market means that Market Maker has "long term"
positions which are quite huge, and those positions need
to be liquidated eventually)
2) as a "rule of thumb" you've pointed out that we can expect
a Trend to move from Long Risk to Short Risk, which will
reflect the "sawtooth" nature of markets.
3) A predominance of "Big Lots" will also be seen at the
tops and bottoms of micro trend. This can be evaluated by
using an "extreme" setting of a fairly short Retention interval,
and, unfortunately; the current implementation does not
include a "MINIMUM_VOLUME" which I really should
implement. In other words, not running sizes into the
Integrator unless they are, e.g. 4 contracts in size; and
ignoring smaller contract trades. This will produce a SPIKE
at tops and bottoms as the BIGGER LOTS come in near
the tops and bottoms of trend pivot. CLEAR AS MUDDY WATERS?
[edit] To see such a spike; you might need to increase the
MULTIPLIER. If you hack it slightly yourself, to exclude
reporting any volumes below a size threshold, then you
can see the effect that will have, where Inventory consists
only of "bigger lots". Note: I use the Nasdaq
NQ/MNQ; and you may be applying this to a radically
different futures contract, so your mileage would vary
You can see the concept is useful, but I've often described
it as providing "situational awareness" of the probability of
a turning of Trend; but NOT pinpointing exactly where the
Trend will turn, and that will depend on your required precision.
I keep "teasing" that I have a much more comprehensive
Inventory Analysis, and I am planning to put those legacy
algorithms into my own custom Order and Analysis console.
[edit2] Here's the problem. The current approach is a simplified
Summation (i.e. Integration) of signed volume over the
Retention Interval. However, within that collection of volumes,
there are "profit taking" pairs, where a Buy volume (Retail Sell)
by MM matches a Sell volume (Retail Buy); and that Paired
Volume (only when profitable) should be taken OUT of the
inventory collection under evaluation. The current algo doesn't
even try to do that; and the thing I'm teasing; does just that.
(This yields Market Maker's "running spread profits" which
offset MM's "open profit/loss" situation dynamically.....)
If I do that; then I may go "the extra mile" and provide a
NinjaScript which uses this more comprehensive analysis.....
No promises right at the moment; I'm fighting other fires !!
Hi @hyperscalper!
I want to make sure that my understanding of "Market Maker inventory" is correct. I thought that the calculation of inventory was the same as for Order Flow delta in a moving window. Is this the way inventory is calculated?
My reason for asking is that when I compare the plots of the TradeFlowRisk indicator and the NT OrderFlowCumalativeDelta indicator on a 1 min chart they differ significantly. The Close of a 60 sec moving window and a 1 min bar should be the same. (I have used the TradeFlowRisk indicator in post # 501 and run it in on MarketReplay.)
I can see a pretty clear correlation between the two "calculation methods".
However, I'd suggest trying something like a 5 minute retention interval on the
same chart with a 1 minute bar.
Also, I know how my code calculates; but I can't vouch for that other indicator
you're using; so I couldn't say how the calculations differ !
And if you are using Seth's modification of my code; I couldn't vouch for
its performance in replay or e.g. under backtesting either; since my code
specifically was not designed for that.... Just sayin'... I've made that
very clear several times, I think For one thing, the Retention Window
duration won't be calculated properly; so.... obviously the calculations
will be thrown off...
[edit2] To answer your question, the two are roughly equivalent, except
that my version looks at a concept of "out of the money risk" and tries
to estimate "how much total risk" is being generated by the trading.
In general, the simplified algo in both NT's version and mine, simply
assign a "signed volume" different for Buy than for Sell, by comparison
of Last traded price with Bid/Ask; and then Sums them up over a
moving window of time. This is a rough estimate of Net Inventory
Imbalance, which can be quite useful in indicating the "situation" and
an elevated probability of a Price Trend reversal.
I THINK you're using it on the ES contract; and for that behavior,
you should really be adjusting the retention interval to be fairly long,
since ES movements are much slower than the targeted Instrument
I use, which is the NQ. They are polar opposites in the degree to
which they will generate "Risk" and especially "Total Risk", since
ES generates way more (huge) Total Risks as compared with
the NQ contract, most likely. ?
One more thing, for TradeFlowRisk you should probably calculate
OnBarClose; since it will throw a DOT multiple times when you are
using OnEachTick as the method. (Internally, it won't miss any data,
but will just draw a single DOT every time OnBarUpdate is called,
and you probably don't want multiple dots, so OnBarClose is best.)
[edit] Devil's in the details, but I use a RealTime clock calculation; so do not
support the Historical phase. Seth has enabled the code to accept data from
the Historical phase, and has written Parameters instead of using my
File-based parameter file. All that would be OK, but I said that a "Virtual
Timebase" would need to be implemented *everywhere* and that has
not been done. For that reason, the results will NOT be correct. Without
a "virtual timebase" (i.e. taking the timestamp from the historical data)
and then NOT using real time timestamps; that simply will not work
properly... You can't "wish" this into existence for that purpose; since it
is designed for RealTime calculations only. Again, just sayin' that if
you're going to re-purpose an Indicator, then you have to take responsibility
for the entire code base, and have it behave properly...
The only thing I can guarantee is the code I wrote, and that code will not
process historical data; but kicks in only in the RealTime phase... It
also uses the parameters file approach, enabling you to adjust the
parameters with Notepad++ or similar editor, as it is running in
real time. It does not, therefore, work in a backfill or backtesting context, since it is quite difficult to get accuracy in that environment,
so I concluded it wouldn't be worth it to push it into that application.
So with my code, you're stuck with RealTime, and I'm sure it does
what it should do with regard to Risk and Total Risk calculations.
ALWAY USE THE LAST POSTED COPY.
Thanks for the prompt and extensive reply. And thanks so much for all your interesting posts on this topic.
So my understanding that the MM inventory calculation is similar to the order flow delta calculation in a moving window (= up volume - down volume) doesn't seem completely wrong.
I have had a look at the code of TradeFlowRisk and I have difficulty to follow all the details. But the VWAP used for calculating "delta ticks" in the moving window myIntegrator.getVwapPrice(VolumePriceExpiryIntegrator.VWAP_POSITIVE_VOLUMES) and myIntegrator.getVwapPrice(VolumePriceExpiryIntegrator.VWAP_NEGATIVE_VOLUMES) does
correspond to the delta calculation in an advanced VWAP indicator almost to the tick. What do you mean by "out of the money risk"? Is there some trick that may explain why the myIntegrator.getSumVolume() doesn't correspond just as well to the Order Flow Delta?
My answer may not be fully satisfying to you, but here's how it
goes; along with the rationale for doing in that way...
I've said before, this is not a Full Inventory Analysis. (In fact, I just
started converting a Full Analysis from some Java code, and there are
not a few adaptations necessary from that code... to implant it within
my own C# stuff...)
To get the VWAP when Net Inventory is "MM Short", as you pointed out,
we use NOT ALL VOLUMES, but selectively choose only the Retail Buy
volumes, when we form the Volume Weighted Average Price, representing
the "Price of MM's Short Inventory". Then if the Current Market Price is
ABOVE that VWAP, the Market Maker is in "Short Risk" since Market Price
is higher than the VWAP against Retail Buyers (the Retail Buy volumes).
Same, but reverse logic, applies when Net Inventory is "MM Long".
Only the Retail Sell volumes are used, whose associated Prices form
the VWAP. If the current Market Price is BELOW this VWAP Price,
then we say that Market Maker is in "Long Risk" since MM's BUYING
PRICE (VWAP aggregate) reflects a (temporary) UN-PROFITABLE
situation.
If someone wants to jump on that as Invalidating the whole approach,
I'll just say that this approach is a simplification; and that selection of
calculations is a good approximation of what a Full Inventory Analysis
would yield. Sorry, I keep saying that; and most of you could reasonably
assume i don't know what I'm talking about but a Full Analysis would
use a "Matching Process" which separates the Inventory Collection into
Profitable Transactions (Closed Inventory), leaving only not-yet-profitable
Inventory that is OPEN and not yet matched (Buy with Sell; Sell with Buy)
so is waiting for a Buy to match a Profitable Sell price; and a Sell to match
a Profitable Buy price; in Long versus Short inventory, respectively. LOL
Such an analysis is "Real Time Only" and Cumulative, and requires a
great deal of memory which holds ALL Trades in memory; and then
periodically performs a Full Matching Process, as poorly described above
Such a full Inventory Analysis will ALSO reflect the fact that Inventory
Analysis is Dynamic and..... wait for it ..... FRACTAL.... meaning that
(For us non-mathematicians) that we can perform the Matching Process
on any timeframe we choose; and EACH such timeframe will yield
a DIFFERENT result, as would be expected. That's because "all timeframes"
are simultaneously being used, during the trading process (within reason)
so the Inventory Situation you wish to evaluate, should roughly match
a timeframe you wish to use as a Trader.
These are "cycles" with different "retention intervals" which are "mixed"
together; in the same way that audio waves are combined into a
multi-frequency signal. Well, that wasn't a great analogy; but maybe
it helps. Net Inventory swings from Long to Short; but over what
period of time should be calculate that ?? That's the issue.
I've got off into the weeds here, and down the nearest Rabbit Hole; but
that's what Full Inventory Analysis will yield; and you can imagine that
it's a compute-intensive Evaluation of the Collection of Buys and Sells
which originate on the "Tape" or "Time and Sales".
Obviously, we're not doing that in the current simplified example; but I
can assure you it is a decent approximation of the effect which the
Real Thing (tm) LOL would yield...
Clear as Mud, probably; but it's how to understand the Dynamic relationship
between Market Maker, and the Retail Population of Traders. Insights into
quite a few Price movements can be better understood when the
Fundamental process of Buyers Versus Sellers is better understood,
since that's the Heart of Trading, after all.
