Over thousands of API entered orders I've only had a problem once and I'm not even sure what caused that.? So it's not clear what the technical answer is, but the practical answer is that order updates arrive just fine.

But even with that I don't depend too much on the quality of information coming back from IB.? There are limits on the orders my software will enter to provide a margin of safety.

The IB TWS API documentation suggests that some of the orderStatus
states may not be transmitted, using the example of a market order in
a fast-moving market.

I am left unsure if I can depend on the "Filled" order status being
returned as a message. Given that I am only placing Limit and Trail
Limit orders, never Market orders, should I worry about checking up on
orders with execDetails via reqExecutions? Or am I at least safe in
expecting a "Filled" status update for these orders?

TIA.

Here is the IB statement on a "split-spread" order: "Split spread orders include any order that is priced within the spread between the bid price and ask price. When they fill, these orders yield significant price improvement."
?
IB should provide a detailed study explaining this statement because my own statistical study on this shows otherwise.? Once again, I realize how counterintuitive that is.? Markets tend to be that way.

By the way, it's not a bad strategy.? My own data shows it hasn't been harmful.? But significant price improvement?? No way.

The short answer is that this does not work well.? I realize that is counterintuitive but I've captured data on thousands of trades using my own algorithm for getting in between the bid-ask.? Your first thought is something like this:? If I can sit at NBBO until getting filled then OF COURSE I'm coming out ahead on average.? But then you'd have to ask why market makers don't do this if it were such a good idea.? The vast majority of liquidity from market makers is further into the order book.? So I won't review all the statistical data and explanations here but I can tell you that there is a deep rabbit hole you'll go down.?

If anyone is willing to share their own experiences and data about this I'd love to get more into it more.

As an alternative to programming it yourself you can use IB’s Adaptive Algo market/limit orders which, in the case of a long, start at the bid and slowly raise the order’s price until you are filled/hit the limit price:

Using API:

On Fri, Dec 24, 2021 at 10:52 PM, Erez Kaplan wrote:

“Split the Spread to Capture Better Prices”

I didn't get the email (so to speak), but this page sounds like the same thing:?

What they're talking about is quickly changing your order price to buy slightly above the highest bid price or sell below the lowest ask. You can easily do that with the API, but you have to program it for yourself. My experience has been that it is hard to beat institutional traders at that game. If you want to have a long or short position, you also risk having the price move against you until the trade becomes unprofitable.
[rwk]

IBKR sent this attractive email to users “?Split the Spread to Capture Better Prices”

The article explains usage via TWS only

how can this be achieved with API?
(sample code much appreciated)

cheers, Erez

Wow, 闯ü谤驳别苍 this is outstanding work. The internet is a beautiful place thank you so much for taking the time to reply and explain the steps you took.

I don't use 2fa or tws so your experience gives me motivation for the gateway. I had seen a similar suggestion on stack overflow regarding adding the openfx paths to the Java exec. I had some hunches where to trial and error in the shell script so your results are time-saving.

I also saw another suggestion to do something like a symbolic link between the default-jre (/usr/bin) Java and openjfk ( /lib something I forget).

Can't wait to get back to the office and try these ideas out. Cheers and happy holidays.

My apologies for the confusion, Todd. As I said in my post, I only have an old 32bit Raspberry Pi 2 and was unable to make TWS/IBGW work on it. My next step would have been to grab a modern Pi 4, install a 64bit Debian distribution (instead of Raspbian) and try again with a 64bit JVM. It looked like 64bit ARM Java offerings were much more rich than 32bit ones. But you are correct, a 64bit JVM will not work on a 32bit OS.

Having said that, I went back to the drawing board and did some experiments on the Pi2 with 32bit OS and and 32bit JVMs. I installed the 32bit Linux versions of TWS and IBGW 981.3g and got pretty far, but not all the way. Here some thoughts that may help you:

• IBKR installation packages contain a Java JVM 1.8.0_152 with embedded JavaFX libraries. So the IBKR code relies upon pre-Java 9 rules related to access between classes from different origins and jars.
• Java 8 versions of JavaFX do exist for ARM 32bit, but they are missing some of the SWING related classes IBKR uses (such as JFXPanel).
• After Java 8, JavaFX became a separate set of modules and more stringent module access rules apply

With that background I did these steps:

• Download the stable Linux 32bit versions of IBGW and TWS. That is 981.3g right now
  
• Install the IBKR package with an ARM 32bit Java8 JVM:
  
  • app_java_home="path-to-arm-java-jre" sh ibgateway-stable-standaline-linux-x86.sh
  • The installation uses your ARM JRE (instead of the packaged x86 JRE) and JavaFX is not required at this stage
• Install a full featured JavaFX version that includes all SWING classes plus the corresponding JVM. After some experimenting I went with
  • openjfx-17.0.1_linux-arm32_bin-sdk.zip
  • OpenJDK17U-jdk_arm_linux_hotspot_17.0.1_12.tar.gz
• Small modification to the ibgateway script (in my case located in ~/Jts/ibgateway/981)
  • At the bottom at the two exec java statements insert right after ...bin/java"
  • --module-path path-to-your-javafx-installation/lib --add-modules=ALL-MODULE-PATH
  • They have to be located at the front of the command line options and adding them to the vmoptions file will not work
• Start IBGW with
  • app_java_home="path-to-your-java-JVM"? sh "~/Jts/ibgateway/981/ibgateway"
  • I had to do that as the root user since JavaFX tried to access all kinds of restricted devices related to mice, video, audio. But that could be an issue with my old Pi 2 and OS installation.
• The login dialogue comes up but I cannot get past the 2FA challenge (long story, but the only account I can use for the test needs this). All input works fine but IBGW claims that I had entered an incorrect security code (which I did not)
• launcher.log and the start script show a few exceptions related to module access restrictions to certain JavaFX classes. Most likely because of the more stringent module security enforced by the JVM. That can probably be worked around with proper module-info files, but will need some work and probably requires repackaging of IBKR JARs.
• Starting TWS 981.3g works with the same recipe. Logging into a demo account eventually hangs in "Processing startup parameters ..."

Hope this helps,
闯ü谤驳别苍

PS. Contact me directly if you want to work on this some more, since this the nitty gritty is probably too far out of scope for the group.

开云体育

开云体育

Not sure what his problem is but I think you are doing an excellent job 闯ü谤驳别苍, thank you for taking up the reins on this on. We have all been handed a steaming pile of opportunity and the only way we get this sorted out is by shared experiences and POLITE discourse.

?That said I have some code I'd like the group to write, it's christmas and I feel like having a snooze. I'll send out the pseudocode later. Shall we say Dec 29th 2021 :-)

have an excellent christmas and a safe new year chaps.

Dent

We are absolutely here to help each other, and I think Nick did a great job giving you all the information, tools and hints you need to do what you have asked for. One of Nicks suggestions was

I think we all used IB's libraries as a reference. Pick whatever language you like best and look IB's code.

You probably want to use the newer "V100" protocol since it includes a length header for each message.

But apparently you did not do that.

We are not here to do the work for you and the 6,000 members in this group can expect that you read the documentation, make an attempt to understand the architecture, follow the advice you are given, and "use the IB code as a reference" before you ask more, frankly, trivial questions.

Let me know what you don't understand and I am happy to help you.

闯ü谤驳别苍


On Fri, Dec 24, 2021 at 03:53 AM, YSS wrote:

Thanks for the help Jurgen, we can close this forum guys, Jurgen descided we can find everything in the code!!? No need to help each other!! Great spirit for the time of the year, keep it up mate! Thanks for your awesome contributions here. You're such a godsent!

?

Thanks for the help Jurgen, we can close this forum guys, Jurgen descided we can find everything in the code!!? No need to help each other!! Great spirit for the time of the year, keep it up mate! Thanks for your awesome contributions here. You're such a godsent!

Likely depends on the API version, I've had issues like:

For those who upgrade and have a issues on the API level, an upgrade of the API is also required. I had yet to see issues like this any time before today (I was using 10.05).