Frosty's Autotrading Equities Journal

Discussion in 'Journals' started by frostengine, May 15, 2009.

  1. Over the years I have done several "unsuccessful" auto trading journals with futures. Those strategies turned out rather disappointing after back testing them on YEARS of intraday data only to watch them blow up in the real world. Now I have a rather bad taste in my mouth when it comes to the art of backtesting.

    This time is different. I am no longer trading futures, I am daytrading a basket of 12 carefully picked stocks. There are several benefits to this:

    #1 I can control my leverage which allows me to "develop as I go" (more on that in a moment)
    #2 I don't have to fit a square peg into a round hole like I did with futures
    #3 More "ideal" trading opportunities

    As I mentioned above, I am no longer very fond of backtesting after seeing "good" strategies not work when live with real money. There are a lot of reasons for this phenomena. Some of them are obvious but many not so obvious at first. For example there is the classic curve fitting.. everyone has done this. But a less obvious problem is what happens if you develop your strategy with trading starting at 8:30 central everyday.. but in live trading maybe you don't start until 8:31 one day? Think that doesn't make a difference? Download say NinjaTrader and run a strategy that is set to start at 8:30.. now run same strategy with start date at 8:31 over your history... VERY possible to have very different results.. That is a very simplified and stupid example.. Will provide better examples as the thread continues...

    For me backtesting is useless.. so what am I going to do? I call it "develop as you go". Basic premise is I took a "strategy idea" that I have used in the past... just the basic concept.. no optimized parameters... just the flat concept. Implemented the concept and then tested it over 1 week of "walk forward" trading. Did this over this past week. Not so much for the actual results. Instead this was done just to make sure the strategy was trading properly and properly identifying the setup I planned on trading. This process was completed over the past several days.

    Now I am moving into trading the strategy LIVE with real money. Over the coming weeks I will make tweaks to the strategy.. improve its identification of the pattern... better exits.. etc... But this will all be done in the context of it trading LIVE with real money. I am sure some of you will say this is a sure way to blowing up and won't work. Perhaps you are right, but in my mind it makes perfect sense. If what I plan to trade is LIVE with real money, then I should develop my strategy LIVE with real money on the line under REAL conditions as they happen not in a simulated backtesting environment where you don't watch day in and day out EXACTLY how your strategy is responding to the pattern and the market.

    This is one of the benefits stocks allow. I can controll my leverage to ensure I won't blow up my account or lose TOO much money while developing the strategy. To start with the system is limited to $3,000 per stock. Therefore since there is 12 stocks in the basket the MOST money that can be active at any given time is $36,000.. however in practice normally no more than $9,000 is active.

    Most days I will only post total number of shares traded and net P/L. Others I will discuss things I learned about the strategy that day and ideas for correcting or tweaking the strategy.

    As of this writing the strategy has initiated 2 trades so far this morning, So I will have some results to report by end of day.
  2. Good to see you back! Looking forward to your journal.
  3. A note about the strategy:

    Currently the strategy only takes LONG trades. I have not yet implemented my short setup. The reason I have held off on the short side is all too often when developing strategies I try to piece both sides together at the same time... This time I wanted to start with only 1 side, get it perfected then move on. A good long strategy will either not lose much or be relatively flat in down markets and do well in up markets. So, until the long only version can accomplish that I will not even look at the short side of my setup.

    Morning results:
    448 shares traded for -$6.66

    Was up nearly +$100 before market reversed course. Strategy was slow to realize and exit the open positions. I need to explore possible alternative exits as to me it was clear as day the market was turning... Will have to replay the day after market closes and see if I can figure out what I saw and see how to make the bot see the same thing..
  4. jimmyrey


    I followed one of your previous journals - it was painful

    Best of luck with this one...

    ...but as we both know luck should have nothing to do with should have an edge.

    I hope you do.
  5. Yea I have had SEVERAL very painful futures journals in the past. One of these days one of them will work out... each time I start a journal I feel optimistic that this time will be the one... Something feels different about my approach this time which really has me believing that this one is it.

    At the very least my smaller leverage will keep my daily swings more manageable which should help. Part of my problem with futures was that the swings would be SOO exstreme that I could turn the bot off at absolutely the worst possible time..
  6. The afternoon session is now over:
    Shares traded: 214 for ~+$3

    Total day:
    662 shares traded for -$3.14

    All in all scratch for the day. However, I can't complain too much considering market was bearish most of the day.. Made a good bit of trades and after commission and everything only lost a little more than $3.. So, can't complain TOO much..

    Although, I saw several times in the morning and the afternoon where I should have finished the day up $100+... But the strategy was slow to recognize changing conditions. Will have to replay today several times this weekend and find what I can see but the bot can not...
  7. Friday was not a bad day. Basically flat when using a long only strategy during a down market. Can't complain much with those results. However, I did see some room for improvement in 2 areas.

    #1 I saw it takes an obviously bad trade early in the morning. One of its stocks gaped up and started selling hard immediately. The strategy bought this position within 5 minutes of the market opening. Which is a big no no. The spread was very wide and stock was experiencing intense gyrations which are common within first 10 minutes of market opening. So, I added a filter that should eliminate entering during those conditions.

    #2 I was able to clearly see the market turning at several points but the strategy was unable to identify the turn. I have replayed the day over and over trying to put into code what I saw, but was not quite successful. So, I have not yet fixed this issue.

    Monday should be interesting.
  8. Just another day. Not much to report... Mostly quiet slow grinding up day. The stocks the strategy watches for all pretty much gapped up... then ran a little... then stayed pretty flat for most of the day... This strategy requires a pullback to enter... by time the stocks started making pullbacks there was no big thrusts remaining.. when the stocks would make a move that would generate an entry there was no follow through... Too much of today's action happened before the market even opened. Disappointing mostly due to loss opportunity. Nothing to worry about... Keep chugging along.

    306 shares for -$25.24
  9. Are thoses 300 shares on many trades or just 1-2? If over 4-5, do you think comish is going to eat you up?
  10. The 300 shares is from 4 trades of which I paid a total of $8 in commission. My $25 loss included this commission cost.

    Right now commissions play a slightly larger role than normal in my P/L. However, once things are on track and I scale up the $$$ allotted to each stock commissions will become less of a factor.
    #10     May 18, 2009