I am new to this site and have just listed my system here for subscription. However, I am not clear if I am able to enter back-dated trades (to create a back-test), or if I can only enter new, live trades going forward. Can anyone answer this question?
Also, any ideas with how to get answers other than at this forum would also be appreciated, as I have tried and cannot get a reply from Matthew or the site people.
[LINKSYSTEM_63551210]
Hi Peter,
C2 doesn’t accept any backtest data (for good reasons), just the realtime orders you enter. They build your track record over time.
I used my own web page to publish backtest data. Have a link in your system description to allow people to find it.
Matthew will answer you mail - sooner or later - he is busy…
good luck!
Peter:
We don’t offer the ability for you to enter backtests as part of your trade history. One important aspect of the Collective2 site is that we put a lot of emphasis on public go-forward commitment. You need to commit to publicizing your results before they occur, and then once they’re on the site, they remain there for all to see. The problem with backtests, of course, is that they suffer from hindsight bias – only people who have great backtests will submit them. The rest of the tests end up in a garbage can under someone’s desk somewhere.
The best way to get help is to write to our help desk at help@collective2.com
Hope this helps!
Matthew
My 2 cents. You can trust only your backtests. Even then, you could have these problems:
1) you could have bugs
2) you could look into future/present (bugs, or improper bactesting)
3) one backtest might not be enough (depends on your system) e.g. if I run 100 bactests, I will get various results, and some are giving very bad results, some very good results, but on average they do great
the most important one is 2) improper backtesting. Have you ever seen the data on which you backtest? You might say no, but if you train your algorithm on a few stocks with data until 2010 but you backtest on some other stocks also until 2010, you are doing it wrong.
And even if you did the backtesting properly, how many times you did it? if you backtest more than once on a certain period, it is not backtest anymore, it is (human) training (you will make decisions to change the algorithm based on backtest results - and sometimes it is enough to make this a couple of time to fit the data)
Edison wrote:
> if you backtest more than once on a certain period, it is not backtest anymore
And this is the most important piece of wisdom when developing trading systems. It took me several years to realize just how important it is…
