Estimating the Scalability of a Strategy

Discussion in 'Strategy Development' started by sf631, Sep 30, 2010.

  1. sf631

    sf631

    I've got a mechanical stock rotation strategy focused on a specific and somewhat illiquid set of stock tickers. It backtests well, it's performed equally well in the past 11 months of live trading, and I've looked at it from enough different angles to be confident that it contains a source of alpha that I can continue to exploit.

    An avenue I'd like to explore is raising outside money (fund, SMA, whatever) and significantly increasing my profit-making potential (on a 2/20 basis). One of the first questions a (smart) potential investor would ask after seeing the track record is how well the strategy would scale to 10x or 100x the asset base. I'm trying to figure out the most credible way to estimate that and would appreciate any help from the crowd.

    Without going into detail, here are some basic strategy parameters:
    * typically 30 equally sized positions at a time
    * screener refreshes daily, hold periods typically 5 to 15 days
    * average daily volume of between 40K and 400K shares at $5 to $25 each (so $200K to $10M daily $ volume)
    * universe of 150 possible targets (of which I'm taking the most attractive quintile)
    * spreads typically more than a penny but not much more (maybe 2-3 cents)

    Currently I make 98% of my trades as limit orders rather than market orders to get the exchange rebate. I'm turning the portfolio over 25x a year, so even 0.2% in extra slippage/commissions (that'd be 2 cents on a $10 share) would eat 5% of my annual return. I think if I relied on market orders it'd be much dicier since the market depth in these stocks is marginal at best.

    Some possibilities I've thought of to estimate scalability limits (or drag created at varying levels of scale):
    * max % of combined universe daily volume
    * max % of combined universe market cap
    * spread between my entry prices and daily VWAP (for all I know I may be doing worse than VWAP )
    * avg bid/ask spread (not sure how to consider this)

    Thanks in advance.
     
  2. Unless it's a family member, close friend, etc...
    Any remotely sophisticated investor would say:

    (1) "Show me your office"

    (2) "Show me your web site"

    (3) "Show me 3-5 years of Audited Financial Statements".

    Start working on this.
     
  3. sf631

    sf631

    Granted, fundraising is tough. Really tough. I know just believing it doesn't make it so.

    That said, me believing that a strategy scales 10x or 100x from where it is (currently about $500K in gross exposure) is a necessary but not sufficient condition to grow. Put another way, I'm not going to invest in all the things that one needs to do to be credible to outside investors if *I* don't believe that I can deliver the same kinds of returns at a larger scale. I wouldn't even pursue the money I could easily get (F&F) if it wasn't likely to be successful

    So, deferring the topic of all the other things that are necessary to take on money, I'm first interested in how to put an order of magnitude on a strategy's capacity
     
  4. you don't give enough details, but here are a few points to consider:

    1) the shorter the timeframe for execution the smaller the capacity. if these 5-15 day hold times are being scaled into/out for a couple of days, it's much different than putting limits out on a single day when price hit's there for a minute or so.

    2) given 1 (and i'm assuming you're not scaling in over days), the shorter your entry/exit time horizions the more you really need to look at intraday volume stats, ie when are you getting in, what percentage of daily volume, how much is going off at your target prices, etc.

    3) given 2, you're only going to be able to do a small fraction of daily volume before you start to impact things.

    4) taking 3 into account is probably the most important thing (market impact). this is where 2 becomes less important and the hypothetical 'what happens when you really need to get out' becomes key. at this point volume averages are much less important in determining capacity as is some worst-case scenario assessment. remember, your stocks are illiquid for a reason.

    a seat of my pants flat out guess based on your numbers and assuming you'll be using single days to get in/out and 5% max daily volume... < 10M. if you really want to push it to 10% (very difficult to do without impact), maybe < 20M. imo, not really worth aum until you have something that can scale to 9+ figures.

    so, i think more valuable to you would be some ways to increase capacity: 1) see if you can add a long/short element if your list is highly correlated (ie, etf's etc), this will allow significant more exposure if your corr's are tight. 2) see if you can scale out over longer time frames where dependency on intraday volume is less. 3) see if you can expand your list out to more than 150 issues.
     
  5. Prop is giving very good advice.

    Focus a bit on finding ways of increasing the liquidity of the strategy - increase the number of signals and / or see how it trades when things skew to the liquid names.. Most hedge funds won't bother with anything that doesn't scale into 10s of millions because its a matter of time and resources.

    That being said, any strategy in its own right that can increase or provision liquidity somehow is worth something in this environment regardless of the returns.... and I mean real liquidity.. not just HFT.
     
  6. sf631

    sf631

    Thanks for really helpful perspective.

    A few more details & responses:

    I don't usually scale in and out in any systematic way. Sometimes I'll take a full position, decide I REALLY like the signal so double the position a few days later, and occasionally I'll take half off the table,

    Generally, I'm less concerned about impact from getting in and out under normal conditions. I'd be fine with getting a daily VWAP in each side of the round trip trade, and I think I should be able to use iceberg orders and so forth to gradually build a block of 10% of the daily volume without paying too much of a premium. Totally agree that if I can bias this towards the $5M daily volume rather than the $200K shares that's a big difference.

    The larger concern of what you pointed out is the need-to-get-out-now situation where I'm a liquidity taker, and the lack of liquidity really becomes an issue. I do run a hedged book of relatively well correlated positions so the exposure to a market plunge is lessened, but still there. The flash crash on May 6 was by far my biggest daily drawdown, and that was about 1.4%. I don't know any way to really get around that risk.

    Psytrade, I'm curious about your comment about provisioning liquidity. Something I have thought about (since my trading to date has been limit order, liquidity provider oriented) is that one way that I could pick up a little extra alpha is to selectively market-make (but only take a significant position on the side that I want to be on anyway) thereby actually making a little more liquidity than currently exists. This doesn't solve the above concern though about what happens if I need to close out positions quickly if I'm the provider of the liquidity.

    Thanks for the input - any more thoughts welcomed
     
  7. provisioning liquidity is very important nowadays. You'll see it when you trade alot. Its freaky how little liquidity is out there.
     
  8. sf631

    sf631

    And sorry for being dense, but "provisioning" means providing liquidity (ie putting orders on the board)?

    Do you mean that a strategy that provides liquidity is valuable because so many strategies now rely on taking liquidity and those in a position to provide are collecting the spread + rebate? If so, I agree :)
     
  9. No I mean taking liquidity. But for purposes of picking up information, you need to do both.