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)
I am a bit confused. Inventory (Time and Sales) Analysis is concerned ONLY with Transactions; not with outstanding unfilled orders.
So it is "Trade Flow". Whereas analysis of Market Depth, as these are Unfilled Market Maker "orders", would be more precisely called "Order Flow". Just sayin'... So a "stop order" is not an executed transaction which would appear on Time and Sales; and so I don't know why you say "if you factor stop orders in as well". That makes no sense to me.
hyperscalper
Can you help answer these questions from other members on NexusFi?
Stop orders are orders that trigger market orders under certain circumstances. My understanding is that rithmic's MBO data allows you to detect if a market order was created by an exchange side stop. Just as we make assumptions about how limit orders are opening positions and taking profits we can make assumptions about the market orders generated from stops being market makers taking losses.
Just ran an "alpha" demonstration on ES Futures this morning; takes so long to get everything in place !!
I was blown away by how predictable the market is; especially since I've been talking about how difficult
it is to deal with "The Beast" which is the Depth of Market. So we're using Rithmic's feed into NinjaTrader.
Had to go with Rithmic's routing also, as I explained elsewhere; but that's good because of the speed of
execution, even though higher margins.
Anyway... I can give myself a slap on the back, so we'll be ready next week to trade this stuff.
So I feel like I've come a long way in "understanding the problem"...
I'm just saying, in general, that "Trade Flow" analysis, which I call "Inventory" analysis, is certainly an interesting
area; where we estimate MM's "position" against the Retail Market. But the real determinant of market
direction is Analysis of "The Book" and that's an extremely difficult beast, as I said, to grapple with.
The Devil's always in the details when you attempt something like that.
I'm not doing the analysis inside of NinjaTrader, but in a separate process on the same system through a local
socket; and that's Java, where the real logic resides in this current implementation. Very focused on just
a couple of contracts, ES and what I like to call "the real Prize" which is NQ are the targets.
[edit] so this is a complex indicator; which could be done "in process" by spawning a Java VM,
but we just extended it into a separate process on the local machine to get an implementation
sooner.
One thing that's emerged as a great trend continuation indicator is to examine the variability of Quotes
as they appear at each specific Price. I have mentionned that in seeking "the true size' one key aspect
is to observe the Minimum Size which appears over a suitable rolling time window in your Price-specific
analyzers. The key here is for your "data capture" to be absolutely Price-specific. And, as I've already
mentioned; when we evaluate a "bias" between Bid-Side and Ask/Offer-Side of the Market in an attempt
to predict direction; it is only then that we take the Price-specific information, and convert it to a
"Tier-specific" representation. That is, we compare like Tiers, such as Bid minus 10 ticks against the
corresponding Ask/Offer plus 10 ticks; of course we do a range of these when developing a bias
estimate.
But back to Spoofing. Spoofing is intimately tied to the concept of Variability. In Spoofing, sizes are
advertised on the DOM which give "the visual impression" of strong quotes; but the micro-structure
shows that larger so-called "spoofing" size may be shown 95% of the time; but a smaller but more
Persistent size is shown during a brief interval maybe 5% of the time; or even for only a few
milliseconds. This smaller size; or the Minimum Observed Quote Size, is closer to "the true size"
which you should take as the basis of your comparisons.
So if we measure Variability; which is a necessary feature of this "eyeball spoofing technique" then we
will see that Variability is significantly increased, but On WHICH side of the DOM? If course, it is on the
OPPOSITE side of the DOM than Price will be moving. Much lower variabilities are seen in the direction
where Price will be moving.
Implementations of these basic ideas are difficult to achieve; given the dynamic nature of "the Beast"
which is the Depth of Market. BUT, this is TRUE ORDER FLOW analysis, because it looks at the Quote
behavior of Market Makers; and does not consider Time and Sales or "Transaction or Trade Flow" which
is another topic entirely.
[EDIT] Remember that "Price moves to Size on the DOM" so when MM is "more eagerly" placing Quote Size near the market; then the market inside Price will be "pulled or attracted" to that Size, since Market Makers "want to interact" with the Retail players on that side of the Market. Specifically how this is measured, does have some latitude, and some challenges for algorithms.
UPDATE. Coming up for air after a lot of coding... with regard to
"Order Flow" which I break up into Trade Flow (Time and Sales analysis)
or Inventory Analysis, and Depth of Market (unfilled orders on the Book),
both of which contain predictive information to be extracted.
As I mentionned earlier, for me, the NQ futures contract is the prize, but there
are some significant challenges processing the data.
INVENTORY. With an overwhelming number of Time and Sales entries being
SINGLE CONTRACT trades, there is a huge stream of events to integrate
into an Inventory Analysis. The entire "chain of custody" of the data in
the pipeline needs to be properly double-buffered, with the ability to discard
data as necessary during "bursts".
DEPTH OF MARKET. I'm using the Rithmic incoming feed, which allows for
at least 40 Tiers above the Offer/Ask and 40 Tiers below the market. Data
rates averaged over a minute show as much as 1200 average events PER SECOND,
with burst rates probably much higher.
This is with a C# implementation in NinjaTrader using the Rithmic data feed,
so I'd just say that it's a significant implementation challenge both to just
capture the relevant data, and then turn it into meaningful predictions.
I have to say that NinjaTrader is definitely "the bomb" for this type of
very fast and complex analysis, so that's my enthusiastic recommendation !
I did have an issue getting Rithmic as the feed; since I was forced to take
Rithmic Order routing and broker (managed through NinjaTrader brokerage)
which has much higher margins than Ninja's normal data feeds and routing.
However, that being said, it's ultimately the right solution for me, so predicting
the volatile NQ market is still "the prize" for all the work involved. The next
challenge is to write an Order Entry strategy which is up to the challenge.
[edit] So, the approach is to analyze the NQ contract for all that information,
but then to execute on the micro contract MNQ using a multiple price level cost basis approach with partial profit-taking of an overall aggregate set
of trades. 10 MNQ contracts is approximately the same as 1 NQ...
Any interest in this discussion still and/or has it moved elsewhere?
I see that Sierra chart also offers MBO data through their Denali feed so there are multiple options now. I've been inactive but am thinking about doing some investigation into order flow across different markets, by which I mean filled/traded orders.
I'm sure there are some really useful things to be gleaned from the limit order book but (to some of the earlier points in this thread) that feels like a game for well capitalised companies with dedicated teams and the low latency infrastructure to execute on it. I don't think we can beat them at their own game ... but maybe that size causes other issues and also leaves a trace, meaning that we could win if we play a different game.
I may have been responsible for revitalizing the issues around Order Flow, and
in my work, I've certainly done a lot of it. But let me just re-clarify what I
understand Order Flow to mean.
WHAT IS "ORDER FLOW"?
Most people, including me, for a while; define Order Flow as "Trade Flow"
or Analysis of the Time and Sales for Buying versus Selling; and thus
calculating some concept reflecting an "imbalance" of the Market Makers'
"net buying or selling" behavior.
So, if the Time and Sales shows only Selling, which would mean that the
transactions occurred at or very near the BID price; then Market Makers
would be accumulating "long" or an excess of Buying, and be motivated
to move the Market Price upwards, to sell off what was purchased from
Retail Sellers, then to Retail Buyers at a Higher Price.
This can be useful, but I've described it more as a "situational awareness"
that "Yes, Market Makers on aggregate have absorbed a lot of Selling,
so they may be motivated to Support the Price, and lift the Price as
a result". It cannot tell you the instant that a down trend, becomes an
up trend, with any precision at all.
All of that is a very natural and logical way of approaching Time and Sales,
and many people consider that to be Order Flow Analysis. In a sense,
Orders on the Book are "Flowing" to Retail players as transactions take
place.
BUT THERE IS A SECOND AND MORE USEFUL MEANING, WHICH IS
ANALYSIS OF THE LEVEL 2 OR "THE ORDER BOOK" which exhibits Market
Maker Orders (and some Retail players limit orders) which exist at
varying Price distances from the Inside Market; BUT WHICH HAVE
NOT generated any Time and Sales transactions; they are merely
Orders.
The "Flow" of these Orders, or the "discoverable patterns" which do
exist in the various "Tiers" on the Order Book are an order of magnitude
MORE DIFFICULT to evaluate, in order to generate Trading Signals
which predict which way the Market may move in the near future.
Firstly, many cheaper brokerages offer only a limited number of Tiers
on the Level 2, and anyway, ordinary traders don't know what to
do with a fluctuating Depth of Market (aka "The Order Book"). If
you are given only 10 possible Tiers either side of the Market, then
you will be limited in perhaps what information you could extract
from that presentation.
Secondly, normal traders do not have the ability to apply algorithmic
real time analysis to the Depth of Market (the DOM) and, even if
they did, that analysis might not be useful in their trading, so
perhaps they are 1) unwilling to pay extra for Market Depth, and
2) the data volume or "bandwidth" to the local trading platform
is generally kept at a minimum for most traders, thus not
requiring much infrastructure to deliver.
Thirdly, Premium costs may be incurred for a Full Market Depth
which is "not aggregated" and which might offer 50 Tiers of
information on each side of the Market, and the data rates
of such feeds (such as Rithmic's Depth of Market or other
"premium" providers) may be prohibitive for retail platforms.
For example, the Rithmic Market Depth (not aggregated)
could generate up to 1000 or more UPDATES PER SECOND.
The potential to use this information is questionable; and the
retail demand is generally low; so why would it even be
interesting at all; given the high cost, and the inability to
UNDERSTAND the meaning of any patterns which exist in
the Depth of Market.
BUT HERE'S THE TRUTH. ANALYSIS OF THE MARKET DEPTH
is one of the strongest "edges" that a highly technical
trader can have. Notwithstanding those people who say
they can trade successfully with nothing but Level 1
pricing; and who claim that Level 2 never gave them an edge.
THE PROBLEM IS PROCESSING THE INFORMATION, and then
being able to predict where a market will move based on that
information, which is a highly specialized and relatively expensive
proposition.
You cannot just "eyeball" such information and expect to
extract anything useful. It is a highly technical algorithmic
problem to determine Market Trend from analysis of the
DOM (the Depth of Market) and you are likely to need a
"premium" non-aggregate feed (such as Rithmic's feed, and
probably some others) in order to do that.
SO I HOPE THIS CLARIFIES the differences between Time and
Sales Analysis, versus Depth of Market or "The Book" as
a source of Signals for trading. Both are perhaps confusingly
lumped into the category of Order Flow Analysis.
So a "
