Wayne, Congrats on your excellent backtest and hope it performs as well for you. Unfortunately real-life is never so predictable, your exections will not match your backtest in a high-frequency system. If your using FOREX here are somethings to consider: 1) Backtests assume trade @ price = fill. You can assume price trade throughs your limit to guarantee fills to mitigate this risk. 2) Slippage does occur for trades during news events. Can simulate this if you have bid/ask for your entire sample. 3) Retail FX broker data is usually presenting AVERAGE of bid/ask and only buys on ask/sells on bid [Oanda at least] So further testing as per rule #1 needs to be done to adjust for this. 4) Missed trades in real-life can have a butterfly effect on the rest of the trades, i.e. the backtest assumes your filled, your still looking to buy in real-life, while your waiting-more trades are executed.... Apologies if you've already incorporated these risks into your tests. I have many different strats on futures that would make me a multi-millionaire if it wasn't for these 'inconveniences' although there ARE systems that are profitable after adjusting based on your writings....
No need to apologize. Most people assume slippage impacts strategies heavily and it does on MOST strategies but not all. I have experience with real live discretionary day trading as well as running some systems automated. So I know personally (from pain and success) how slippage works. For example let's say the market is in a big rally for a couple minutes. The Forex spread goes from the average of 1 pip to 5 or 6 pips or even more. Bid ask spread of 6 pips means that for example sellers are asking 106.06 for USD/JPY while buyers are bidding 106.00 and the market is fast climbing after news report. Suddenly, it climbs and the the spread jumps to 10 pips. Sellers are now asking 106.20 and buyers bidding 106.10. If you now want to BUY at the market or on a stop market, you're in from some real PAIN. First of all you're paying a 10 pips spread. But even worse, by the time your order gets filled it will probably be at 106.25 or 106.30. (Ask me how I know?!) You're only a tiny bit better off with a limit order but still in a bad position. Let's say you immediately place a limit order to buy at 106.20 you may never get filled because by the time you place the order, the market may already be at 106.22 or more. ========================== However, let's discuss SELLing into the exact SAME market conditions. ========================== Remember, this is a rapidly rallying market. So if you place a market order to SELL you will almost certainly get filled at 106.10 or more due to the over supply of buyers. But it gets even better if you use a limit order. Let's say you set your limit at 106.15. Since the market is moving it will likely fill at 106.15. But even if the market moves past 106.15 before it fills. Some buyer will certainly pick up your order at your discount limit. You may wonder, what if the market stops suddenly before filling at the 106.15. Then you learn the importance of scaling into a position variably and you'll learn that I was already short at 106.00 and if it went to 106.25 I would short again, etc. The point most people miss is that slippage always depends on what direction the market is moving. Don't you hate developing strategies that work with the market? Slippage can eat you alive. Anyway, the system you read about above ALWAYS goes opposite the direction of the market. So when the market rallies with many more buyers than sellers, it sells to those buyers. Never a problem getting filled there. Then when the market corrects and those buyers are getting stopped out with market orders at a panic (<--that means they'll pay any price) then we mop them up with limit orders. Point is, trade against the market (but with the prevailing trend), and forget about slippage entirely. Sincerely, Wayne
Hi, First of all, what do you mean by "random distribution of prices"? Do you mean distribution of price changes at regular time intervals? If it's so, then it is impossible to create a profitable method on assumption that time series is random and independent. On top of that, market time series are non-stationary due to drifts and fat tails. The only way to exploit random normal distribution (I assume you're talking about normal distr. since you are excluding fat-tails) I can think of is a martingale strategy with a big risk of distribution regime shift from normal to fat-tailed.
Listen, I stumbled onto this strategy while trying an idea. I'm stuck trying to explain it. How is it possible for to have a strategy with 99.99% trade accuracy and never a losing month if the markets are random? Of course, they're unpredictable. Question is, how is this possible. I have come up with an explanation that only make sense to me and the others who use this same strategy, it seems. Think about it, even when I curve fit strategies in the past (like every beginner) I never got 99.99% accuracy and zero losing months. And this strategy has ZERO optimization or curve fit. It only has 3 very simple rules. However, the platform I built has a very sophisticated infrastructure which automatically handles things like variable positions sizes and multiple time frame data within the same strategy. So it would be non-trivial to do this tradestation, Neoticker, NinjaTrader, etc. It took weeks of work to get all these crucial features built in. But after having that kind of power under my belt, I led to this strategy. How to explain 99% accuracy? I didn't believe it at first. I still have a feeling of disbelief but I found other traders on this forum like Maestro (search on him) who say they are trading by using randomness for years with never a losing month. What strikes me is that randomness is so predictable. I would never understood it or believed it until seeing it myself. It truly is the "holy grail". Now that I understand it, the rules are so simple, I can look at any chart of any instrument and know when to increase or decrease my positions to exit with profit. Watchout, I don't mean to predict. I never now from one second to the next if it will go up or down. I also heard many traders saying to trade never based on predicting but on what that market it doing now. I never understood that till now. Now it seems so simple and automatic. How many times have I heard that you have to watch prices on the screen for a minimum number of hours before you can "get" how the market works. It's possible I simply hit that critical mass that my brain finally recognized that pattern. I think it's impossible to take any short cuts. This strategy is so easy, I could do this discretionary but I just love automation and don't have the patience to sit for hours in front of a screen developing a position. And that boredom would be necessary to get the timing right. Let the computer do it I say. Thanks for reading. It's therapeutic somehow. And I hope this at least encourages others to never give up. Finally, I certainly do not expect you to believe me. And I still respect you if you don't. Sincerely, Wayne
Your posts are very interesting, but also very puzzling. I have enough expertise to program and automate anything, so that's just trivial for me. The interesting part is what's your idea. I can think of only three ideas of how to achieve 99% consistency in profitability: 1) Through finding strong and persistent dependencies in market and external data. But it really isn't as simple as "only 3 rules", so I guess this is not a case. 2) Martingale betting. As long as price change time series remain normally distributed and stationary it is doable. But martingale strategy is sooner or later severely damaged by fat-tail events. This might be part of your approach. I'm thinking like this because you mentioned fat-tails cause you troubles. 3) Frequent uncorrelated positive expectancy trading. It is true and mathematically provable that the more trades you place in a given period of time, the bigger probability it is to end up profitable. This is only true if the following is true: a) there is no correlation between trades (no serial correlation in the trades within single trading method, and no correlation between trades of different trading methods.) b) all trades have positive expectancy (after commissions and slippage) I guess your approach is a combination of the 2nd and 3rd. I cannot be sure, only you can tell.
I can tell you that a system exists that has 100% profitable trades in backtesting. No BS. It does not matter whether you assume the data is random or non-random: irrelevant. No TA necessary, only past, present, and future price (exact future being unknown of course). It is also extremely likely to maintain the same performance going forward (P = 100% based on historical information).
Comment away, but, if you really want to impress the audience, let us know when you've reverse engineered the system or proven it's not possible.. Anyone who's followed my posts should know I'm not prone to fabricating tales. I'm more in the verification and deconstruction camp (although I'm known to conjure a hypothetical scenario every now and then to make a point).
This the where my trading has taken me, and it is not perfect BUT results show there is lots of truth in your statement. At times, I simply wonder WHY more people have not gone down this road ...
Actually the whole "System Development" thread by Alan Crary is a detailed tutorial about this approach.