Experience in using a group to do back testing

Discussion in 'Strategy Building' started by Georgii, Nov 7, 2013.

  1. Georgii


    Hello all,

    I'm thinking of getting one or two or more people together to do some back testing and market study. My question is has anyone done this before, has it been helpful, and what hints do you have to do this effectively?

    I would obviously assume you'd want to find people who think similarly to you and are serious. More tips?

    Thanks in advance!
  2. Programmers?
    Low paid monkeys who can change a few parameters in a retail program, and click "Start"?

    Why can't you do that yourself?

    Or are you thinking more of the lines of creating your own brainstorm group, and then testing the group's ideas/strategies?

    Or are you thinking of starting a trading group where you're the head honcho, and the people are testing and trading your ideas?
  3. Bob111


    ---- My question is has anyone done this before,--

    --- has it been helpful--


    --- and what hints do you have to do this effectively? --

    i have no hints for you. because you all have to be roughly on same page in terms of experience and knowledge. if you are not,then you are here:

    ---Or are you thinking of starting a trading group where you're the head honcho, and the people are testing and trading your ideas?---

    i can come up with trading ideas\patterns daily. who would backtest it for me? without any pay. you have to be very optimistic\altruistic person to do that :)
  4. It's not hard to find a few unmotivated blobs who are content with the security of a paycheck, but... you aren't going to get the highest caliber folks from this pool...

    However, if you have access to a few hundred million, you could form your own fund and gain access to the motivated blobs....

  5. Georgii


    The idea is to find people similar to me who are committed to trading and will be using these patterns in the live market together with me. In other words, I'm not looking for minions but actual traders, people who will co-brainstorm, offer feedback, etc.

    I'm thinking of back-testing the old fashioned way, I'm not a programmer and I don't use indicators in my approach.

  6. gmst


    how much money do you have in your account? what do you trade - equities or futures? with whom do you trade - retail or prop firm?
  7. Bob111


    imo-it's pointless in current times..

    the NEW old fashioned way is this: you program pattern or whatever you have in your head-then you manually review each signal\trigger(and it's outcome). and run your tests over and over again. if you can express what you see into a code-this ability allow you a) remove any subjectivity(which is imo NOT your friend). in other works-what you see is not necessarily visible to others. b) you can backtest (and make adjustments) and test it again really quick.
    c) it will give you far more data points,than you can achieve by doing it manually. more data points-more statistical stability or more 'guarantee' that it might last.
    d) ability to test over various market conditions(what working today,might not work at all in markets similar to 2007-09)
    e). once you have solid system-you can spread it across endless list of markets\individual stocks etc.

    you just can't achieve this by doing all calculations manually. you will be dead,before you can find any solid statistical advantage.
  8. Georgii


    Bob, I would argue that there are definitely downsides to using too large a sample when testing. The main concepts I intend to use are simple things like breakouts and pullbacks, and the idea is to be iterative about it, constantly reviewing them as market conditions evolve.

    Whatsmore, not all patterns can be easily programmed. From what I've seen programming works best with things like indicators. It is much harder to describe technical patterns themselves, not merely from a mechanical standpoint but because they require understanding a broader context.
  9. The problem with any kind of backtesting is that people usually don't have real money applied to that data. You are guessing and might as well just turn it on and see realtime.

    Mass up a ton of hard drives' worth of data... then trade your system and every day backtest and optimize all over again. Come up with a correction factor for your real executions vs. your simulated... then next time you run an idea you will have a MUCH better idea of how it will actually behave. Pair 3-4 strategies together like this and you will get a REALLY good idea of what works or doesn't.

    Working that way in a group only makes for competition to see which one of your friend's can steal your strategy first. :D