Hypothetical Performance to 1995 or 1999 for all?

Hello C2,

Why does C2 not require all systems on C2 to show a hypothetical going back to say 1999? Why is it tied only to when the system was put on C2?

There is no way to know with confidence how so many of these systems will do in down turns - the past should be there for all to see.

Can this please be improved?

Thank you,

Impossible, of course. Many instruments traded did not exist back that far, and many strategies are discretionary, not algorithmic.

1 Like

Hi, Robert -

You ask a great question, and I think it points out how much work I have to do to show new users the power of Collective2.

Of course, any Strategy Developer can create a hypothetical “backtest” of a trading system, going back to the beginning of time. With today’s powerful computers, and the great software that has been developed by companies like TradeStation and NinjaTrader, it’s simple to “discover” a strategy that does really well over many years or even decades.

But let’s peek behind the curtain.

In order to “discover” that strategy, the developer (or, more likely nowadays, the software he or she is running), has to “try out” thousands, or millions, of possible strategies that don’t work. These are simply thrown away. You and I never see those track records.

Then, lo and behold, after millions of permutations, the Holy Grail Trading Strategy is discovered! The backtest is published for all to see. The strategy is then marketed and sold.

And then the Holy Grail strategy doesn’t work.

Because what has happened, in most cases, is that the software simply “fitted” a strategy to the already-existing historical data. And, yes, it fit the data wonderfully, but it had no predictive power going forward.

That is why C2 provides an important benefit to investors searching for trading strategies. Collective2 does not allow someone to post “historical backtests.” Any monkey at a keyboard can create a wonderful backtest. But these are meaningless.

What really matters is “go-forward” results. On C2, you must post your trade signals as they happen. And then we track the results of these trade signals. So you can trust that the hypothetical track records you see here on C2 are not the result of backtesting, but rather are real go-forward out-of-sample results.

That’s a big, and important, difference.

If you are interested, about 9 years ago (!) I wrote a short e-book explaining this in more detail. If you are willing to overlook the horrible 2010 typography, and the freaky screenshots of what C2 looked like back when I ran it from my bedroom* then you can download the free book here:

“How to Lose All Your Money Trading”. (Free e-book. Click to download.)

Hope this helps.



(1) Yes, back in 2010 I ran C2 out of my bedroom. Today, as the company has grown, I have moved to much fancier digs in the family room.


Exactly. Back-fitting is dangerous.
And to echo @v1Trader 's comments, many instruments weren’t traded or the data is tough to get, as well as the fact that many strategies are discretionary so it would be impossible.
Forward walk-thru testing is the only real way to go. C2 accomplishes that.

Hi Matthew - I agree and understand that back testing can be used to game - but it is very hard, nay dangerous, to never the less judge any system on just a year of actual trading let alone 3 - so what is one to do in your opinion? It would mean there are only a very few strategies on C2 I would at all be willing to back - and only those that have been here since its very early days.

I believe the curve fitting conundrum you speak of has been solved by a number of groups. Suggest a discussion with Mike Barna at Trading System Labs in San Francisco. If I understand this correctly C2 does not monitor the systems in say the same way that say Futures Truth does…but are tracking the receipt of signals from the creators and thus the ability to run raw back testing on a system entrusted to you on your own is not how C2 is setup.