You may want to check with the various exchanges as IB is just reporting.? The exchanges have pools that don't print during the day and only hit the tape after the close.? I'm not sure what time frame each of the exchanges have for final posting if trades.

Mel?

I am sorry to be a bit stubborn but I prefer to start in Purebasic. Might change later eventually.

?

However I am struggling using the TWSlib.dll. I can connect to this lib but I am having problems to calling a function.

What I would like to know is if TWSlib is the best option or is it better to use the windows socket solution ?

I am not sure but it seems that the source code suppplied with TWS API of the other languages use the socket version ?

?

And I also have the impression that the socket version is more generic and makes the program independent of changes made by IB to TWSlib.

What do you think ?

?

Please help me, I really would like to start making TWSAPI calls and start working on the interesting part of the program.

?

1- Not every stocks are traded after RTH even if traded premarket (unusual though)

2- Even if market is still open doesn't mean there are any trade done.

3- Look at your market data subscriptions

?

1- do a reqContractDetails and look for TradingHours

2- use TWS and look at Booktrader

?

isn't this exactly what useRTH=False supposed to mean?

Hi,

?

I am trying to retrieve historical price data of a few FOREX currency pairs: mainly AUDCAD, EURUSD and EURGBP.

?

Since November 25th, I am unable to retrieve historical TICK data via API. Before that date, I was able to download any price even going back to 2022.

?

When trying to download TICK data, I am getting the following error:
Historical Market Data Service error message:HMDS query returned no data: <SYMBOL@TYPE OF PRICE>

?

I get the error regardles if it's:
CASH or CFD
IDEAPRO or SMART
ASK BID BID_ASK_MIDPOINT price

?

What I tried:

1. I tried and I CAN download 30 SEC candles meaning that it is not a problem with the code, connection or ports.

2. I used another account and another IP.

3. I tried different DATE formats, e.g. yyyyMMdd-HH:mm:ss

?

I am running out of options.

?

Can you please help me understand what might be the problem?

I experimented some time ago with downloading daily data for 200 US stocks to see what worked best for me. I tried sequential versus parallel.

With sequential I mean I submit a reqHistoricalData() for one ticker and wait until I have received the data and stored it in a csv file. Then I do the same for the next ticker on the list.

I repeated the experiment with parallel downloads. With parallel I mean that I assign a request ID to each ticker. Then I send all reqHistoricalData() requests for all of these, taking the pacing limit of 50 messages per second into account. All received data was stored with the particular request ID and then matched with the ticker symbol. Again, all data was stored in a csv file, one file per ticker.

This parallel approach was a lot lot faster. Since then I don't do sequential data requests any more.

Be aware though that if you request too much data you might encounter a "soft limit" imposed by IBKR. As it is soft, there are no real criteria provided by them. When exceeding these soft limits you will encounter connectivity issues (IBKR will disconnect your app). So what I do is, after I once downloaded the initial historical data, to download daily only the recent historical data and merge the new data with the existing data. Getting one month of historical data for around 200 tickers takes around 15 seconds with this parallel approach. I don't know whether this is the most efficient solution, but it is good enough for me.

There are at least three recent posts that you should look at, but you should also spend a little effort and search the archive yourselves:

• "How to download 100 tickers data of 1 minute bar for 1 day" describes an approach how you can download full day 1min bars for all 500 stocks in the S&P500 in 26 seconds. The approach presented there can be applied to other bar sizes and time frames.
• How do folks here front load their databases with TWS?
• How should one request 6-months worth of historical 1-min stock data from Interactive Brokers API?

Batch request
Never heard of it on IBKR,

- Look to me against the design pattern used by the API. Data callbacks do refer to a reqid, not to a symbol, so I don't see a way to resolve OHLC Bar for multiple symbol

- as you can asynchronously launch multiple requests, batch mode is in batching requests.

- IMHO as it's a also a Trading API, provision for batch would also induce some dramatic coding errors.

?

yes there are, read:

optionally study what IBKR call "quote boost" (you shouldn't need it if I understood what you look for)

?

Reality is that it depends also on the popularity of the instrument, like if IBKR does not cache every data and you may need to wake-up old storage

Try to stay under 10 request/second if not intra-day, (you say "includes" so I don't know what else than Daily you may be looking for.)

?

None I heard of, and again I doubt anything exist you can manage yourself, this would be against API pattern (how to manage the cache size ? Purge ? Cache synchronization is a huge burden on IBKR shoulders, etc ...)

Tradition is that you handle the local storage is a DB of any form, including CSV file

?

1.0- Unless you are dealing with crypto (where precision is paramount), consider another specialized data feed for your data, one that support batch request

like Polygone or FinancialModelingPrep or ActiveTicks,

IBKR primary purpose is to be a broker, not a data feed.

?

1.1- Other feed benefit of alternate feed is that they report more popularly used data.

IBKR seems to tradeoff with some data they fudge themselves, which reflect another kind of reality, somewhere closer to the efficient reality needed when you go manage orders.

I won't discuss who is right or wrong, what could be important if doing decision is frequently to use same data used by other algo

If you just look for OHLC, discard this advice IBKR one's are same, IBKR VWAP is different, probably more accurate, as other feed seems to base their VWAP on 1min hlc3 as AP.

?

2- Loop with 10 request in parallel, don't limit yourself to 1.

You will need a "rendez vous" method to synchronize all historicalDataEnd.? More advice involve knowing that language you expect to use.

?

?

Hi everyone,

I’m working on a project that requires pulling historical data on hundreds of stocks using the Interactive Brokers TWS API. The data includes daily OHLC (Open, High, Low, Close) values and trading volume. My current method involves sequential API calls in a loop, which is slow and not efficient for a large dataset.

Here’s my current setup:

1. Sending requests for individual stocks sequentially using the TWS API.
2. Saving the returned data locally after each request for further analysis.

I’d like to optimize this process to handle large batches of stocks more efficiently.I’m interested in advice on:

• Batch Requests: Does the TWS API support batch requests, and if so, how can I implement them?
• API Limitations: Are there known rate limits or best practices with the TWS API to avoid throttling or data loss?
• Caching Strategies: Are there effective ways to cache or save data to avoid redundant requests during retries or updates?

If anyone has experience optimizing the Interactive Brokers API or similar APIs for high-volume data retrieval, I’d appreciate your insights. Also, if this topic has been covered before, I’d be grateful if you could point me to the relevant discussions in the archives.

Thanks in advance!

Margin by itself is not additive as it is calculated under the maximum norm. The closest thing that works in real-time is CME's SPAN mechanism. There's some documentation on the client systems wiki if you want to implement something similar, although it would require simulating all scenarios (up, down, sideways, high vol, low vol).

?

Margin(A) + Margin(B) != Margin (A+B)

?

With the SPAN methodology, the risk vectors are additive:

Risk(A+B)[k] = Risk(A)[k] + Risk(B)[k]

?

So margin can always be deduced:

Margin(A+B) = max ( Risk(A+B)[k] for k in [1..16] ) = max ( Risk(A)[k]+Risk(B)[k] for k=1..16 )

?

IB is doing something similar on their side.

Interesting! I didn't know that. I am going to check that out!

Richard wrote almost exactly what I wanted to reply.

In a nutshell: go with python. Lots of support available, lots of sample code.?
If you have any kind of programming experience, the learning curve to python will be a pleasant and a short one.

Hello all,

?

I would like to set an upper and lower limit for selling stocks I hold. For example, let's say that I hold 120 shares of SMCI and the current price is 32.50. I would like to SELL those 120 shares if it reaches 34.71 OR if it went down and reached 31.20:

?

As 闯ü谤驳别苍 mentioned, you need to keep track of PnL yourself based on the executions.

Reconciliating against reqPnl should be a governance / secondary check function and not be based on for trading.

?

As such it's not directly a TWS related question. For example, more advanced PnL engines will keep track of the following also:

- arrival time

- arrival price (what is current market price when you decide to go long)

- fill times

- fill prices

- average of fill prices

- slippage (fill price - arrival price)

- commisions

- market cost (fill price - mid price on each fill)

- price impact (mid price 3 minutes later vs fill price)

- MAE (in-trade)

- MFE (in-trade)

- draw-down across trades

- profit

- %win

- avg. win

etc...

?

You would need to pay alot of attention to the number handling (especially where you might have zeroes and a division - for example with 0 trades you would return something for %win even if the division by zero number of trades is undefined).

Depending on what you're doing you might want to optimize secondarily also for different parameters (like minimize price impact) or just use one of the existing TWS algos.

I am going to have a look on Python, but I am affraid that mastering all the subtleties of a new language could take a lot of time? (a bit lazy too...).

?

---------------------------------------------------------------------------------------------------------

Here is the code generated by chatGPT for the "TWSLib.dll" VERSION. This code is only for getting started.

---------------------------------------------------------------------------------------------------------

; Load the TWSLib.dll (make sure it's in the same folder as the executable or in the system PATH)

If OpenLibrary(0, "TWSLib.dll")

??? Debug "TWSLib.dll loaded successfully"

???

??? ; Declare the functions from the DLL

??? ; The function signatures are derived from TWSLib documentation and can be different

??? ; Depending on what you're doing, the function names and arguments might change

???

??? ; Connect to TWS

??? Define ConnectToTWS.i

??? ConnectToTWS = CallFunction(0, "Connect", 7496, "127.0.0.1", "clientId") ; The Connect function is for establishing a connection

???

??? If ConnectToTWS = 1 ; Check if connected successfully (depends on how TWSLib.dll works)

??????? Debug "Connected to TWS"

???????

??????? ; Now you can send a request. For example, let's request account data

??????? ; You would need to call the appropriate methods for account data or market data

???????

??????? ; Example of account request (you'd need to look up exact method and params in TWSLib)

??????? Define requestAccountData.i

??????? requestAccountData = CallFunction(0, "reqAccountSummary", 9001, "All", "NetLiquidation") ; Example parameters

?

??????? ; Process response (usually you need a callback handler to get updates,

??????? ; but you can poll the data periodically)

??????? ; Placeholder code for data reception

?

??????? Debug "Requested account summary data"

?

??? Else

??????? Debug "Failed to connect to TWS"

??? EndIf

?

??? ; Don't forget to disconnect when done

??? CallFunction(0, "Disconnect") ; Disconnect from TWS

???

??? ; Release the library when done

??? CloseLibrary(0)

Else

??? Debug "Failed to load TWSLib.dll"

EndIf

?

Here is the code generated by chatGPT for the SOCKET version.

; PureBasic code to connect to the TWS API using socket

?

; Define TWS API socket port and host

#TWS_PORT = 7496? ; Default TWS socket port

#TWS_HOST = "127.0.0.1"? ; Default local host

?

; Initialize the socket connection

If OpenNetworkConnection(#TWS_HOST, #TWS_PORT)

??? Debug "Connected to TWS API."

???

??? ; Send a request to TWS (example: request account summary)

??? ; Example request string to get account summary

??? ; The following message is the format for sending API requests to TWS

??? ; This particular message is an Account Summary request

???

??? Define requestString.s

??? requestString = "10|1|1|1|4|0|1"

???

??? ; Send the message to TWS

??? WriteStringN(requestString)

???

??? ; Wait for response from TWS (you can use a loop here to read the data)

??? Define response.s

??? response = ""

??? While AvailableNetworkData()

??????? response + ReadString(AvailableNetworkData())

??? Wend

???

??? ; Output the response

??? If Len(response) > 0

??????? Debug "Received response from TWS: " + response

??? Else

??????? Debug "No response received."

??? EndIf

???

??? ; Close the network connection

??? CloseNetworkConnection()

Else

??? Debug "Failed to connect to TWS API."

EndIf

As Richard suggest with reasons, Python is most probably the best choice for a programmer who need to switch to another language

You can find a lot of support here in Python, and Java (and C/C++)

Python is one of supported language from IBKR (you need a reference point)

?

While IBKR is Asynchronous (that's the beauty of it), Python users seems happy with it and there are some decent provision for multi-threading in Python.

Just fight using any "sleep", there are always a good trigger or cv to listen to.

I never heard of anybody complaining about python performance for algo trading, before you experience a difference you will have to deal with a lot of unrelated issues (coloc, venues selection)

As per portability, switching from one platform to another involve so many details that code refactoring is not the top one of the list (even in C++ if you use C20 and standard RTL call, porting from/to Windows/Linux is a matter of days on first iteration)

?

As per code generated by ChatGPT, able to deal with IBKR TWS API, I am presuming here, but it could be a layer that obstruct understanding of some detail and make your debugging life a misery.

Out of curiosity I wonder what it looks like, can you post it as an attachement?

I am really intrigued!?

Knowing the IBKR API,how can GPT generate a working lib out of nowhere but specs ?

?

Last though:
I am under the impression that IBKR have a decent REST API implementation.

This could be a simpler path for you if you don't want to use a lib. REST is now universally implemented you should not have issue in Basic

?