Calculating Scalabilty

Discussion in 'Professional Trading' started by HFStartup, Nov 14, 2011.

  1. Investopedia defines Scalability as:

    "A characteristic of a system, model or function that describes its capability to cope and perform under an increased or expanding workload. A system that scales well will be able to maintain or even increase its level of performance or efficiency when tested by larger operational demands."

    This may be a basic question to some, but is there an industry standard for calculating scalability for a trading system? If so, how is it done and does the methodology change across financial instruments (equities, derivatives, bonds, futures, currencies, etc.)?

    Thanks.
     
  2. bone

    bone ET Sponsor

    Please look at a live order book several different times during the course of a trading day for the markets in question. Everything else is a futile academic exercise with no basis in reality.
     
  3. The industry operates with the asumptions at the standard Bone mentioned.

    For this reason the industry perfoms relatively poorly as compared to the market's offer.

    You can easily get the correct answer by doing a running tabulation that gives you the market's "capacity" (its coping ability) at ant particular time.

    What happens most in markets is adding and pulling orders so the book alone is just humor.

    T&S, on the other hand is limited to the orders that do get filled where one side is a Market type trade and the other is a book type trade.

    To narrow the opportunity to a place really close to reality, there is one other consideration that the financial industry screws up. The book holds additional info that is really helpful but indirect in terms of availability.

    What the financial industry does mostly is get left at the alter. Orders place go unfilled and the market moves "against" the unfilled orders. worse still is the nature of a thing called "whipsaw", a term used to describe failing the accomplishment of a required trading goal.

    To prevent this form of loss, another kind of mistake is made and it is called the "early exit". What happens in terms of "coping" is that a lot of player money is always sidelined and not available to face.

    So as Bone says, the formulas used have very little value to him and his ilk.

    To include all these variables is easy; the capacity of the market (its coping ability) can be measured by the trading velocity divided by a number that assures you no slippage in the partial fills of orders. Stops cannot be part of this since they do not work for any account of any size. Discontinuing the use of stops is part of the trader's growth process. You can read almost any post and determine if the person is still depending on stops instead of skills and knowledge.

    The no slippage calculation also depends on the market's harmonic orientation (odd or even).

    To include the harmonics you use a spacing strategy that still yields no slippage and completing correcting the side of the market to be on.

    A strategy for making money can be deployed that is a multiple of the Wikipedia level of thinking. (It is also very superior to the current flavor of the month HFT) A good comparison is one shot one kill (sniper's do this) to the use of an AK47 or a mortar. Heavier packs are required in the field.

    The ultimate measure of success is money velocity for capital available.

    You run at bursts to have no slippage AND you do partial fills using a stacatto of shots that occupy the whole window for the turn.
     
  4. MGJ

    MGJ

    I think you may have better success if you attempt to define (or to calculate) the asymptotic upper bound to scalability, for a given ensemble of trading methodologies applied to a given set of tradeable instruments.

    How much money is too much? Can you trade the ensemble with 10^6 dollars? with 10^9 dollars? with 10^12 dollars? with (10X the annual GDP of all countries on earth combined)?

    At what dollar level does performance degrade noticeably? (What is its -3dB point?) At what dollar level does performance fall apart completely?
     
  5. Jack, thank you for the detailed response. It raised a few questions:

    You wrote: "You can easily get the correct answer by doing a running tabulation that gives you the market's "capacity" (its coping ability) at ant particular time."

    Once the "capacity" is determined, how does one calculate the point at which markets are adversely impacted by a trading system's activity (volume)?

    You wrote: "To include all these variables is easy; the capacity of the market (its coping ability) can be measured by the trading velocity divided by a number that assures you no slippage in the partial fills of orders."

    How is trading velocity measured? Can you share the mechanics of how this is actually calculated? And what do you suggest using as a 'number that assures you no slippage in the partial fills of orders'?

    You wrote: "The no slippage calculation also depends on the market's harmonic orientation (odd or even)."

    What is harmonic orientation? Will I be at odds if I try to get even? :D

    You wrote: "To include the harmonics you use a spacing strategy that still yields no slippage and completing correcting the side of the market to be on."

    Huh?

    Thanks again.
     
  6. Thanks, GMST. I already make money consistently so that is not a concern for me, but I am always interested in gathering "nuggets of knowledge" when I find them. I appreciate you calling attention to the importance of what Jack is saying because I was just debating whether or not to invest the time to try and understand all the ramifications.

    At this point, my intention is simply to get a rough idea of the scalability of my trading system to present to investors.

    Best of luck to you as well.
     
  7. gmst

    gmst

    hello, it was a sarcasm thats why i deleted my post.

    I have a serious suggestion for you though. Since you are already making money consistently and also have clients, I am sure you can sit down and think about a good rationale approach to compute scalability. IMHO, no harm in seeking opinion however, thinking through the issue yourself will lead to a much better solution.
     
  8. GMST,

    Sorry, totally missed the sarcasm. I know there are a lot of strong opinions about Jack on ET, but that is typically the case for most active posters.

    It is funny that you mention developing a methodology and posting it because I did come up with one, although I didn't post it because I thought there must be some industry standard I am not aware of. I am probably missing some important details. I tend to be straight forward in how I do things, so this may be too simplistic...

    Let's say I manage a portfolio that trades five equity indices and their corresponding derivatives. In this case I'll refer to them as indices A,B,C,D,& E. In reality, allocation to each of these indices is dynamically established by a proprietary trading model based on market activity rather than establishing targets at the beginning of the year or dividing the AUM by 5. But for this calculation, I made the assumption that capital would be allocated 20% to each index. (Assume all are highly liquid such as DIA, TLT, QQQ, etc.) Positions are taken that vary in length from 4 to over 270 days.

    According to my trading methodology, I establish a position first with options by selling a put and setting aside the capital to buy the underlying if assigned. This means that I only acquire the underlying equities if assigned. Therefore, it seems to me that any limitations would be related to the option market's volume flow as opposed to the equity's. Further assume I can spread my orders over a range of strike prices and expiration months (let's just say 6 either way from the current ATM level.

    My guess would be to take the average daily contract volumes of these options and multiply times a percentage that I feel would not impact the market. As a ballpark figure I would start with 3% as a estimate. Hypothetically speaking, if 100,000 contracts on average are traded per day on the options I would use to establish a position, then I would multiply by 3% to give me a total of a maximum of 3,000 contracts per day that could be traded without impacting the market. Since I always set aside the capital in the event of being assigned, I would then need to calculate this amount and add it to the amounts calcuated for the other 4 indices. If that total came up to $500,000,000 that is what I would presume is the level of scalability for my system givent current market volumes. I suppose I could even take this one step further by calulating the prior growth in volume of the option markets referred to above, and use that a my growth factor to determine how scalability will increase assuming constant future market volume growth at prior rates.

    Obviously, the model above is based on a lot of assumptions and this is where my concern comes in. For example, is 3% as a level of market impact reasonable? Does this number vary considerably based on the financial instrument in question? How about the market in general?

    I invite all to shoot holes in it or offer suggestions, because I have to come up with something and it has to be defensible.

    Thanks.

     
  9. Why don't you just run a simulation using VWAP
     
  10. When you say simulation, do you mean there is software out there that does this? If so, what is it called?

    Thanks.
     
    #10     Nov 16, 2011