Which data to use?

[masmit]

I'm testing an idea that produces short term trades, typically 1 or 2 days in the market. I still have much work to do, but I'm not sure how to proceed, as I'm getting very different results according to whether I use back-adjusted, ratio-adjusted, or non-adjusted data.

With back-adjusted data, my last test produced a CAGR of 80%, max DD of 14%, and average DD of 3%.

With non-adjusted data, however, it was CAGR of 22%, max DD of 18%, and average DD of 4%.

The RAD contract fell between the other two.

I'm inclined to believe the non-adjusted (damn!), but since the sysytem tested uses indicators with lookback periods of between 40 and 95 days, I'm wondering about the validity of these tests.

Any thoughts from people here would be helpful.

Cheers,

Mark Smith

[Bondtrader]

I trade using backadjusted contracts. After 6 months of real trading, I compared the results in my real account, against a Tradestation simulation of the same period. The agreement was excellent. P.S. my data and contract-generation software is CSI UA.

[Javelin]

I'm curious what constitutes acceptable agreement between realtime and simulated results.

What are y'all using as criteria to accept/reject the realtime tracking of the simulated account?

Thanks,
J

[Bondtrader]

For each trade, check these eight criteria:

Same direction (long or short) in both accounts?
Same entry date in both?
Same entry order type (market, stop, limit) in both?
Same entry price in both? -- see below
Same exit date in both?
Same exit order type in both?
Same exit price in both? -- see below
Same gross profit (dollars per contract) in both?

In the general case, backadjustment will move the price levels around so it's important to find a way of verifying "same entry price" and "same exit price" despite different price scales. You can use your ingenuity to invent ways of accomplishing this; one idea is percent-of-candle-height.