I definitely concur with having to test it over a broader time period. Ill have to get my hands on historical ES data as AMP only offers 2012 up till now I believe.
The $120 was net profit this week per contract. the stats I posted were from the raw data (~400 trades) that was then inputted into the excel journal available here on futures.io (formerly BMT). Using a starting balance of $5000, stop of 4 ticks and 2% risk the ending balance until end of may was ~$31,000. Not taking into account differing market conditions that might arise and jsut focusing on the specific time period I think that the test statistic Stop is conservative, taking 4 tick stop outs every time instead of my 3 tick volume threshold stop that I currently use. I think I said earlier, I'll take the 3 tick stop but if i get slipped to 4 its accounted for. After all, my original theory and idea for the system was to only get stopped out if price moved greater than 3 ticks -- touches of the 3rd tick would hopefully not get triggered.
In the end, putting my money where my mouth is will let me know if it works. I think $40-50 trades are a relatively small price to pay to find out if a strategy works...especially in comparison to CL strategies which require 10-20 tick stops!!!!
Unfortunately you won't get off so easy! You have to take it into account because you will encounter different conditions
You dont' have to worry about slippage and fills in real markets. The thing you have to worry about is the fact that your system won't work when the market conditions change. Also you will be totally screwed if a limit-move type event happens while you're trading. Stop losses and slippage are the least of your concerns in that case.
However, congratulations on your successful first step. Many people don't get to that point. Just realize that a successful first step is most definitely not a successful strategy without doing more to make sure it works in different conditions.
Can you explain your comment a little bit? I'm not sure if you are talking about a specific case, or just in general.
My experience is that slippage and fills are always important, and definitely need to be accounted for and worried about. I've had plenty of great look strats that failed once you accounted for proper slippage, or realized that fills were impossible.
I would have to agree that slippage and fills are one of the most important things that a trader has to be concerned about once he/she has established and traded his/her method.
The $120 was, this past weeks net profit per contract. What I meant by saying not taking into account differing market conditions was I tested the method over the given time and I posted the statistical data for that time period.
No, my method wont work in all methods but I like had said it was thought up and meant for ES and the characteristics of the market (thick bid/ask and smaller ADR) in saying that, I guess there would be less concern about slippage since you wouldnt necessarily be worried about someone slapping a 500 lot and price moving 5 ticks like in CL. Like Kevin said fills and slippage can really destroy a system. That is why I've decided to continue my tests with 4 tick stops instead of 3. The stats I gave were with a full 4 tick stop in the market conditions at the time, which I think are pretty "decent". Max draw down over the entire time was under 15% I believe with a max loosing streak of 4 if I remember correctly.
mwtzzz, I'll be sure to check out the post you linked me.
I just find it so exciting having developed a method that seems to play out in a reasonably profitable manner in forward tests and is mechanical enough for me not to really worry whether I should take the next trade or not...real cash will be another story. I could post up a sweet looking noobish excel spreadsheet of the millions that could be made in only a few years, but we have all seen those beautiful equity curves before hahaa.
Since it looks like you backtested from Jan - Aug -2012, and forward tested since then, you should definitely keep a VERY short leash on this strategy. Meaning, I don't think you tested for enough time/market conditions to make any meaningful conclusions. You might have enough trades, but you definitely don't have enough different market conditions.
Here is a good example of testing a short period, then testing a longer period, and how the conclusions are radically different...
But, my bias is to test with a lot of data (5-15 years). So, I'll always discount short term tested results, unless...
IF you can develop a protocol/method that senses when a strategy is going bad, before it goes bad, then you could theoretically trade this strategy until that point, then just switch to another strategy you have waiting. If you do it right, you'd ride the "wave" of each little strategy, then hop on the next, and the next, etc.
The key would be in the switching method, not the actual strategy itself (I have found it relatively easy to find a good 6 month tested strategy, but exponentially tougher to find a 6 year tested strategy).
The following user says Thank You to kevinkdog for this post:
In CL the bid/offer spread is at most two or three ticks in normal conditions and the profit objectives I use in my trades makes this spread a negligible portion of the gain. My profit target are usually from 11 to 40 ticks. I have never concerned myself with whether my offer gets filled at the bid, or vice versa. My primary goal has been to make sure I have a trading strategy that is much better than a "breakeven" one.
when I enter or exit a trade, I am satisfied with a market order - I don't care about the specific price, I care about getting in or out.
However, I certainly understand if you're working with a tight strategy, proper fills would be a concern.
You definitely have to backtest more. Usually a method that works well in one time frame (which could be months), does not in others. Unfortunately that's just the way it is, otherwise it would be easy to keep the ball rolling.
But your 15% max drawdown is good. Try to keep it around there as you tweak and modify your system.
Also don't dismiss my discussion about adverse market conditions and the need for options. All your hard work with real money over six or eight months, can go down the tubes with one or two days of extreme conditions.
this is just as difficult as developing a single strategy that works well in all markets. Which is to say, for all practical purposes, cannot be done. Again, an unfortunate thing, otherwise making money would be so much easier.
This is absolutely true, and as I mentioned in the other thread, it's difficult to tell when the market changes. My recommendation is to develop a strategy that works well in one kind of market. Then modify it so losses are limited in the other types of markets, while still allowing it to produce profit, albeit slower and fewer.
Last edited by mwtzzz; February 17th, 2013 at 10:58 AM.
Hi Stefan, backtesting by hand is not the same as livetesting by paper trading nor is any of these the same as live testing with real money. The other problem you will have too is real trading with a small account is not the same a real trading with a larger account. That aside I would encourage you to put aside the papertrading and backtesting as soon as possible and start real trading. nothing focuses the mind more than having real money on the line. Of course I would alse STRONGLY suggest you start with a very small account and number of contracts (preferably 1 contract) and trade at this level for a lot longer than you want. This is the only way to grow and perfect the art of trading. Best of luck to you.
Backtesting with Excel is not real backtesting.
Excel has its own limitation.
And real trading is real-time, instead of in intervals of 5-min, or 10-min, or 1-hour, or 1-day.
In between the time interval many things can happen that will cause your system to whipsaw like mad, get stopped out like mad, etc.
Results can be ultra super wonderful with Excel (due to error in formula or others) even inclusive of commission and slippage, but vastly a different world if you do it using tick data (or second/minute data) with C/C++/Python.