SP trend following System

Discussion in 'Strategy Building' started by acrary, Oct 14, 2002.

  1. wdbaker


    I am not at liberty to post the code for that system and I've found that people don't like it when you post results for a system that you can't divulge. I also don't want it to take away from this thread, this really seems to be a decent system.

    #21     Oct 15, 2002
  2. acrary



    You use a buy/sell stop and enter if it touches the channel. Likewise with the exits.

    hit n run,

    In 2000, the max. contracts traded at one time was 37.
    In 1999, the max. contracts traded at one time was 20.

    Most of the time, it traded less than 1/2 the max. contracts for each year.

    The portfolio size to start is a function of your risk tolerance. Mine is pretty low. I don't mind 5-10% drawdowns per-year, but I never want to lose more than 20% of my portfolio in a year.
    The size I chose to start was 150k because it limited the max. drawdown to less than 20% of the account. That would translate to a 30k account for 1 Emini. If you could withstand a 40% max. drawdown, you could cut the starting account size in half. Anything less than 15k for an Emini and I think the risk of ruin will be too great. I'll post the 10 year drawdown chart so you can get an idea of how deep the drawdowns have been.
    #22     Oct 15, 2002
  3. thanks a lot for the great info.

    in case of the 15k per contract, the stop will be 2% or the 5 bar exit, ( what comes first) right?

    -Here's the 11 year results from 1990 - 2000 using continuous intraday data.-

    confused here, do you use intraday 30 min es chart or the continuous data for trading the system?
    #23     Oct 15, 2002
  4. acrary


    "in case of the 15k per contract, the stop will be 2% or the 5 bar exit, ( what comes first) right?"

    No, the stop is always the 5 bar exit. The 2% is used for position
    sizing. In the 15k case since you could withstand double the drawdown and you'd be trading 1 contract longer.

    The amount at risk would be the larger of .02 * 15k = $300 or
    the range from the highest high of 15 periods - the lowest low of 5 periods for a long entry. If the range between the high and low is 3 points or less, you'd be trading 2 contracts to start.

    "confused here, do you use intraday 30 min es chart or the continuous data for trading the system?"

    I use the back-adjusted continuous intraday data for backtesting only. A live feed with 30 min. charts would be used for real trading.
    #24     Oct 15, 2002
  5. acrary


    Yes, I posted some stuff over there.

    The difference between trading off market character and an edge is the difference between night and day. Market character refers to the past history of the market. Assumptions are made that the past will be similar to the future. Traders fool themselves into believing that if I test with enough past history and look for robustness, it'll work just fine. People that trade off character are trading off of historical tendencies. This is like trying to drive a car with the windshield blacked out and driving by looking out the back window. Imagine a car on a road in the desert. Most of the time it's straight with a few turns. By watching the side of the road and making constant adjustments they could probably drive at 20-30 mph without a problem. Now the market character changes and they're in the mountains. Hairpin turns, windy roads, dirt roads, etc. The methods used in the desert no longer work. The driver would have to slow to 1/2 the speed to have a chance of survival. If they kept looking out the back and drove as fast as in the desert, they'd be off a cliff in no time. No wonder people say psychology is important. If you're trading like this, you're always worrying about the market changing.

    Edge trading is based on finding non-random conditions in the market and using them to trade. Unfortunately, edges are not always perfectly defined, so we may not get 100% perfect trades. This is like having a small spot open in the windshield from which to see through. If the edge is small, maybe we can only see 50 feet in front of the car. This is enough to allow us to go much faster in the desert and drive with confidence in the mountains. To the trader with an edge, accounts compound much faster and with lower drawdowns than market character. The size of the edge can be measured. If you're interested in how, check the edge thread under the psychology forum.

    I have a edge in the SP markets that I found using weekly data. All the trades are done by buying/selling on Monday at the open and exiting on Friday on the close. No stops! You would expect any method like this to have had trouble in the past two years with the long side trades. Afterall, we've been in a downtrend. Here's the long side trade summary for 2001 and 2002 (up to the present).
    • long.gif
      File size:
      7.6 KB
    #25     Oct 15, 2002
  6. acrary


    Now that you've seen what happens trading against the trend, here's the summary of the short side trades for the past two years. Again, only weekly trades (Short on the open Monday and out on the exit on Friday). No stops! This is only using 1 contract per-trade with no position sizing. Once you've traded with an edge, you'll never settle for trading with character of market again.
    #26     Oct 15, 2002
  7. acrary


    How do I know this is an edge and not curve fit to the data. I tested the entries against random entries with 5 day holding periods for each trade (the longs and shorts were tested separately). The longs beat over 90% of the random entry and hold possibilities for each of the past two years. The shorts beat over 80% of the random entry and hold possibilities. As long as they continue to beat more than 60% of the random entries, I have confidence this is not fit to the data. I also did a blind test with the same edge on the ND weekly data to see if it might apply to that market. Here's the two year summary of trades for the ND market. As you might imagine the edge wasn't nearly as strong in the ND market as the SP (though it still did well).
    #27     Oct 15, 2002
  8. Good stuff, sir. Still I'm kinda missing the distinction here. You're giving an example of "day of the week" or "seasonal" kind of a strategy. Then you apply that random entries "test" to it. It survives the test, and therefore you deem this strategy to be an edge rather than chracter of the market. So, do I understand correctly that - the distinction between chracter and edge is the ability of a given strategy to pass the mentioned random entries test?
    #28     Oct 16, 2002
  9. It struck me last night, while dozing off to sleep, that what you are calling an Edge, I'm looking at as probability.

    If I've got this right, the point is, Going long/short on the appearance of a particular set of circumstances carries a higher probability of success, than making the same move randomly at any other different time. Having figured out the particular circumstances that trigger the trade either long or short (or not at all) Given the high probability of success overall the profits will flow over time.

    That's the kind of approach I've been taking, but with changing markets in crude at best technology that ain't easy. I Need automated backtesting badly. - I spent 5 hours last night working something out that a computer would have done in 10 seconds...

    #29     Oct 16, 2002
  10. Brother,

    No stops? Just buy on Monday and exit on Fri.? What a joke. Ever hear of risk management? What if we get a terrorist attack and markets are against you %10??

    In the real world.. who in their right mind would trade such a system? This is another one of these.. looks great on paper statistical gigs..

    I still dont understand the difference between edge and market characteristics.. seems to me like you are just spinning around the same thing.

    "Edge trading is based on finding non-random conditions in the market and using them to trade."--- HUH?

    How can you prove something is non random in the market? This is impossible... its like saying you know for sure who GOD really is. If something was truly non random.. it would work 100% of the time. The only non random thing about the market is that it always changes. May professors have come up with exotic formulas to test randomness.. but at the end its all just a waste of paper.

    "Assumptions are made that the past will be similar to the future"-- Any type of test that uses historical data whether looking for the "edge" or "market characteristics" is based on this premise. The fact that you are backtesting and using historical data contradicts your logic. In reality, what you are really doing is manipulating historical data to find a way to make money.. this uses assumptions.

    Any professional trader puts their emphasis on money mgmt.. not trying to find the holy grail (edge) like yourself. The holy grail is money management! Trading is not about manipulating past data and running different tests and optimizing variables. Its about using sound money management principles.. that's it!

    don't mean to be rude.. but this is bizarre.

    #30     Oct 16, 2002