Discussion in 'Trading' started by yobo, Sep 25, 2009.
your question is to simplistic, every software offers
this, tradestation, ninjatrader, prorealtime, multicharts, esignal,
more detail are needet.if you are a beginning
programmer i would recommend an easy language
platform, tradestation would do fine
If you really believe in backtesting and won't believe me that testing ideas with streaming prices is better, I suggest you consider testing ideas in real time as well just to see. You should see after a while what the limitations of backtesting are.
"You should see after a while what the limitations of backtesting are"
what are the limits??
I backtested a moving average crossover system that seemed to make money on paper. When I traded it with small amounts of money, I had something like 13 or 14 consecutive losses at one point. What were the problems I saw? First, when entering on a moving average crossover, you end up getting at the opposite end of the candlestick meaning you get the worst price. Looking at the charts, you'd think you'd get in exactly where the moving averages crossed. I guess I'd have to post a screenshot to show that. Another problem I saw with exponential moving averages, when the moving averages cross and continue in the same direction then the price makes a tempoarary retracement which causes the moving averages to cross which would mean you'd get out in real time. But then the moving averages "re-open" when the price takes off again. When the moving averages cross much later on, you'd have the illusion that the profit would have been much larger. And that those larger "profits" would have paid for several smaller losses which would give you the impression the system is profitable.
So, to summerize, two problems I saw are: #1. The entry points are not as favorable as the charts are implying once they are formed (in both directions). #2. Once the charts are formed, the profits seem larger than they would have been.
I did however like the fact I could see in backtesting how many consecutive losses you could have. I sometimes saw 3 to 5 consecutive losses for "profitable" systems and 10 consecutive losses for losing systems. Of course, even if a method makes money, I'd prefer assuming 10 consecutive losses in a row could happen when thinking about money management.
to the last poster with the ma issue's
you should set your software to enter a trade on the CLOSE of the bar, that is AFTER the crossover is confirmed and even then you should enter by a limit or stop-order. ofcourse your way you will get prices
that are in real world not do-able, it makes no sence to enter intrabar if the ma is not done crossing.
testing on close orders or next bar market orders is not enough, you will never get that close or open without paying slipage
the limits are the people not knowing how
to work with the software and deal with its shortcomings, not the software itself. just use limit orders and re-run your test. this is the only way replicate realtime fills
Right -- it sounds like you're just using an inferior backtesting platform. The code and data has to support the same time granularity used in actual trading. I get very high fidelity between test and actual results at the individual trade level. The real problem is how to chose rules for the future based on results from the past, as has been discussed many times here and elsewhere.
backtesting generates ideas. Newbies backtest dozens to hundreds of ideas. Then they see a few that "look promising" However, ANY large sample will always produce a few "on top just due to random, statistical reality. Rarely does it mean they have value. And there are other problems that can contribute to systems that look good.
At minimum, you need Lengthy (blind - no peeking) walk-forward testing under different market conditions to show ANY validity of a backtest. And even then, all you have is a possibility.
The realities of live market trading tend to leave trader ideas in ruins. Just because something worked under back and forward testing, and even if it is a true edge, it does not mean that the market won't deal you 15 losses in a row on that "edge."
Few traders have the moxy even when they find an edge, to apply good money/trade/portfolio/account management to see it through for even a couple of years.
There are limitations, often very critical, and flaws people do not become aware of until it is too late and they have sustained big losses due to taking the integrity of the results for granted.
I suggest everyone who is using backtesting to read this paper. The author will not reveal the identity of the programs but one can guess that. These programs, the author believes, their limitations and flaws, were a major contributing cause to losses by newbies in the period mid 1990's to the early part of this decade.
I don't think the poster who tried to point out the flaws in my backtesting really understood what I did. But, I don't want to take the time to debate it. In any case, the previous post reminds me of the ads I saw for TDAmeritrade bragging that you could use their software to do backtesting as if they were implying you'd be a "sophisticated trader" by getting an account with them and doing some backtesting.
I have an idea though you can always try. I'm not ready to give away my ideas yet, but I can tell you the latest ideas I decided to test with streaming prices that showed a lot of potential were developed by looking at the flaws of the previous ideas I had tried which had not worked. So, if you test ideas that do not work, if you look at several different ways of correcting the problems in those methods, maybe one of those strategies will show potential.
Separate names with a comma.