No, that shows that it would have failed 100% of the time had that strat been implemented in the past. It does not indicate how it might perform in the FUTURE.
Most reliable data source with NinjaTrader
- Thread starter Nan11
- Start date
No, that shows that it would have failed 100% of the time had that strat been implemented in the past. It does not indicate how it might perform in the FUTURE.
to each its own. We would never trade anything without backtesting it first.
Maybe. But it does not tell you anything about the future. Listen carefully, there is a lesson to be learned here. It is towards the end.
What do you suggest instead of backtesting?
For Gods's sake, FORWARD testing! The market psychology changes year by year. With that said, you realize that what happened in the past may not happen in the future! Certainly, these past 4 years have proven that. You can backtest to the beginning of the market in the 1600s, but unless you are a 100-year timeframe trader, you will never find satisfactory results!
The dynamics of the markets changed too much and too quickly over time these past 5 years for backtests to be a reliable indicator of what the market might do in the future. Ergo, backtests will be unreliable for a successful system for the future.
Backtesting lets you weed out systems that will almost certainly fail in forward testing and can show what systems worked well over a long period.
And, if you believe what worked in the past will randomly work in the future, forward testing is useless too.
Backtesting lets you weed out systems that will almost certainly fail in forward testing and can show what systems worked well over a long period.
And, if you believe what worked in the past will randomly work in the future, forward testing is useless too.
I will respectfully disagree. Because backtesting can only show what will happen if market conditions from the past continue into the future at the same rate. We know that is not what happens. Grr.
there's more to it than just randomly throwing together some indicators and backtest it on random markets and that's where he get's stuck^^Backtesting lets you weed out systems that will almost certainly fail in forward testing and can show what systems worked well over a long period.
And, if you believe what worked in the past will randomly work in the future, forward testing is useless too.
OK I'll bite: how is data quality defined and measured? - in what units? How is live performance based on data quality defined and measured? How will competitive data be obtained.
First of all, data quality depends on the edge you want to test. If you want to test intraday, you'll most likely need tick data and these are expensive. Minute data does not work because it depends on the data collector how the minute is defined and which data is included in a period.
EoD doesn't matter that much, margin of error can be larger.
You most likely need to setup your own database and compare it to the data you bought from an external source. If you have more ticks than external datasource the data you bought might be filtered or snapshotted.
Competitive data will always be collected from a direct exchange connection, never from a broker or data provider, especially intraday. It's a ton of work and really expensive, which is why everything on intraday basis is dominated by larger shops. You have no chance doing that as retail with retail platforms.
If your edge is 20cts on average and you obtained data from IB which is a snapshot of SIP, you can asume that you have no edge at all because you only tested on 2/3s of the ticks that actually happened on exchange.