Arbitrage back-testing charting package

Discussion in 'Trading Software' started by tonyf, Sep 6, 2019.

  1. tonyf


    I am looking at securities A, B and C (with C being an exchange listed derivative).

    Is there a charting package that would plot 2 lines: A and (B + 0.5*C) as in Half a C for every B and summed in line 2.

    In other word, some chart where I can apply a basic calculation formula. Free, if possible. Purpose is to back test an arb opportunity.
  2. ZBZB


    tonyf likes this.
  3. ZBZB


    Cqg or esignal do a ratio spread chart. Esignal classic delayed is $55 a month.
  4. bone

    bone ET Sponsor

    All of these mentioned tools are only good for screening potential arbitrage candidates. With the possible exception of the incorporation of Bloomberg Time & Sales data - your back test results will be contaminated by artifact tic prints.

    From my experience, you will need to have the actual DOM real time order books up side-by-side in order to gauge arbitrage viability.

    And, you'll require a very expensive ($$$$) ECN and execution platform (RTS) because you're competing with DRW, Geneva, Jump, Belvidere, Peak 6, and other really really big boys.

    Your 100x100 fiber line won't cut it.
  5. tonyf


    A friend has access to Bloom - are you referring to intraday price and volume data?
  6. bone

    bone ET Sponsor

    No, I am referring specifically to Bloomberg Time & Sales. Exactly what I typed.

    This hunt for a legitimate arbitrage that you can exploit on a retail level will be fruitless, btw. Your back testing will show promise because all back testing shows promise; but your data is flawed and your assumptions about execution efficiency are wildly optimistic. What you chart synthetically overlaid is vastly different than what the real time side-by-side order book bid/offer is.

    I wish you good fortune !
  7. tonyf


    Let's put it this way: some guys trade, and others sell trading tutorials for $7,500 and earn referral fees to prop shops. Guess which box I belong to?
    GRULSTMRNN likes this.
  8. bone

    bone ET Sponsor

    Let's put it this way: as I type this I can pull up two order books side-by-side in TT. Real time. Because I trade. Anything I (or you) can plot or backtest using CQG or CME or Bloomberg data will not replicate the on-the-run bid/ask differential between two products.
  9. Total nonsense. Bloomberg Times and Sales is not any more special than some other Times and Sales data. Not even Bloomberg incorporates all special cases such as specific off exchange transactions on, for example hedges of otc derivatives trades.

    Then now you say the opposite, that Bloomberg also does not provide sufficient data to spot arbitrage opportunities. Sounds like you don't really know what you are talking about. Let me explain:

    Replicating the entire order book is extremely difficult and complex but it can be done. What is even easier to do is displaying the data once one has it. I can display ticks with nanosecond precision if I wanted to. The real challenging part is not the data source or visualizations, it's not the testing and design of a strategy but it's the execution part and latency requirements. That is the truly expensive and challenging part.

    Are you really selling your crap for 7.5k? Holy cow, you are fucking genius

    tonyf likes this.
  10. gaussian


    While I agree with you that if they're trying to make an actual risk free profit they're boned (no pun intended) - there's also the possibility of a pairs trading arbitrage strategy which can be executed for a profit at a significantly slower timeframe (vs something like index arbitrage). Though this doesn't disagree with your statement that they will need a better ECN than a normal retail one to get good pricing. I'm not an expert in this field (you are) but the books I've read don't seem to imply HFT for something like ETFs going uncorrelated due to market madness. be fair if his system didn't work he wouldn't make referral money to props because they'd stop taking his references. Yet here you are, a guy who can't even figure out how to plot the linear regression of a response variable and its predictors in Excel criticizing a guy who makes money in exactly this field.
    #10     Sep 6, 2019