Equity model stats - aim higher?

Discussion in 'Automated Trading' started by DGunz, Feb 11, 2011.

  1. DGunz


    This post is specifically for equity traders, utilizing backtesting/paper trading as a decision to go live or not on a new model.

    My prefered performance indicator is cents per share.

    I have back tests that yield about 3c/share (Gross- not including slip and comm) with little optimizing. The paper trade results are more or less in line with backtested results.

    I did go live but, there isnt enough significant days to draw any conclusions. But I am starting to suspect that 3c/share is not adequate.

    Slippage and commission combined I suspected would be in the range of 1 cents/share, leaving only 2cents/share net for me.

    Although the small sample size of live trades have out performed backtest and paper trades, my true slippage for this model was more in the line of 2cents/share, this model happens to be an aggressive fill strategy that needs to take liquidity therefore the slippage is higher than I first though. This leave me with about .5cents/share after commissions.

    Dont get me wrong, I'd take .5 cents/share all day everyday, However I feel as though I am not aiming high enough, as some pro traders on this site are routinely getting ACTUAL results of 3c/share (Lescor), while I only set my backtest to yield 3c/share...

    I cant possibly expect to outperform my paper and backtests, If I want realized gains of 3c/share I probably have to push my tests and paper trades up to 5c/share.

    Any thoughts would be appreciated.
  2. DGunz


    I want to clarify that these results and bench marks, are all intraday trading, with around 150 names per day. Just in case any of you, thought the results were weird and were comparing that to an long term strategy.
  3. EricP


    I think you are making a mistake in comparing your cps results to another trader. Different strategies will inherently offer different cps potential, with the length off trade being a key factor, as well as the price of the stock traded. => Easier to have larger cps with a 4 hour trade, versus a one hour trade, and easier to have a large cps with a $200 stock, versus a $20 stock.

    Your current cps results sound great to me. I would focus on how you can get your average slippage numbers down. Can you improve your order routing? Can you build positions with smaller incremental orders (instead of fewer, but larger orders that tend to increase slippage). Can you occasionally (even 5-20% of the time, under specific circumstances) place limit orders instead of marketable orders, to try to win the spread, instead of being so aggressive and losing the spread plus slippage. Also, is there anything you can do to reduce commission rates?

    In short, I think your cps numbers are not an issue whatsoever. Instead, I would focus on making incremental improvements in your current system to try to tweak out another 0.5 cps or more from smarter routing or cheaper commissions, or anything else you can come up with.

    For what it's worth, while I've done pretty well over the past two years, I've never had a single month in that time where I averaged over 2 cps on my trading. The average for the entire time frame is definitely below 1 cps (net of commissions/fees/slippage/etc).

    So, I wouldn't suggest going back to the drawing board... instead, keep your head up, and continue to improve every way you can with your current trading.

    Best of luck,
  4. DGunz


    Thanks for the bone. Thats some great advice and gives me a direction to go towards. For the next few weeks, I am going to focus on order execution and working orders to win spreads. I think those two things you mentioned could very well yield more cps than going back to the drawing board to tweak my model.
  5. Those numbers are enormous. 3 cents/share means you're making $3,000 for 100,000 shares traded. That's tremendous. Assuming 1c/share slippage and 3/10ths of a cent commission, that's still $1700 per 100,000 shares. It is rare for a daytrader to consistently make more than a penny per share. When I did back testing, my automation mentor told me that anything above 2 cents share would be good as it would still make a penny after deducting for costs.

    I make 1/10th of a cent per share. Today, for example, I traded 275,200 shares to gross $431.52 and net $370.12.
  6. DGunz


    Well, I am certainly glad to hear that. I wasnt too sure what was either too low or too high of a benchmark. Now I know I am not aiming too low, and will focus on execution to get as close as possible to backtest and paper trading results. Its nice to know that I am giving myself ample room for error.

    The cps seems high but I am running tests that trade only about 500k a month (the most premium of trades). I as I stretch out the the backtest to include a million share or more, it is in line with your observation that, cps is <2.

    Thanks for shaing NY.
  7. drp7804


    I guess the decision whether to continue with a strategy or rework some/all of it should be partly based on whether you believe it will meet your money-making goals. For example, is your goal is to have something you can live on week to week? Or maybe you're just looking to outperform some index over the course of the year (ie trying to find a better investment avenue). The value of the cents-per-share metric (in terms of helping you make a decision to continue or change direction) probably depends on what that overall goal is.

    Assuming the backtesting you've done says the strategy will meet (or beat) your goals, how did the live trading do? Seeing as how you've at least been profitable in the little live trading you've done, it probably doesn't hurt to continue to trade that strategy live until you have enough data on which to base a decent decision. If monetary risk is a concern (as opposed to wasted time, etc), maybe consider reducing trade size.

    Anyway, my two cents...