I expect that this has already been addressed but, I've been comparing the results of some of the published strategies, using different data sources, ... WL, ThinkorSwim, and IQFeed.
I'm experiencing very large variations from one data source to another, using the exact same strategy components. I am careful to standardize my selections so that there are no variations which might influence the outcomes, and I select variables which will not produce any NSF.
And, to be clear, these differences are not small, they are actually quite significant, and they vary widely. But, they are never the same from comparison to comparison.
Has anyone else found similar testing results to be true?
Does anyone have a possible, constructive explanation for this?
Thx
I'm experiencing very large variations from one data source to another, using the exact same strategy components. I am careful to standardize my selections so that there are no variations which might influence the outcomes, and I select variables which will not produce any NSF.
And, to be clear, these differences are not small, they are actually quite significant, and they vary widely. But, they are never the same from comparison to comparison.
Has anyone else found similar testing results to be true?
Does anyone have a possible, constructive explanation for this?
Thx
Rename
Firstly, backtests on WL's integrated Wealth-Data and 3rd party's Norgate Data will differ from most other sources due to being free from survivorship bias: Avoid Survivorship Bias with Dynamic DataSets.
As for others, no data feed is the same.
As for others, no data feed is the same.
Gosh, apart from the dynamic/historical DataSets Eugene referred to..
Back adjustments - does your provider adjust for these? (With Norgate Data all are optional. Other providers, not so.)
a. regular dividends?
If so, you shouldn't apply dividends to backtest results in Preferences
b. special dividends?
If so, you should apply the regular dividends to backtest reults.
c. Other corporate actions? Spinoffs? Merges? etc.
If not, you'll probably see unexplained "cliffs" in your data.
Closing Price
All EOD providers [that I know of] use the settled close, so there's little variation on this price.
Opening Price
All EOD providers [that I know of] use the first full-lot trade on any exchange except for Wealth-Data (and Wealth-Data alone). Wealth-Data gives you the Primary Market Open, which is a no-slippage price that you will get with a Market-On-Open order.
High/Low
Wealth-Data (and Wealth-Data alone) applies a statistical filter for outliers. Some highs may be lower and some lows may be higher than all other sources.
More on Wealth-Data at https://www.wealth-data.com
Back adjustments - does your provider adjust for these? (With Norgate Data all are optional. Other providers, not so.)
a. regular dividends?
If so, you shouldn't apply dividends to backtest results in Preferences
b. special dividends?
If so, you should apply the regular dividends to backtest reults.
c. Other corporate actions? Spinoffs? Merges? etc.
If not, you'll probably see unexplained "cliffs" in your data.
Closing Price
All EOD providers [that I know of] use the settled close, so there's little variation on this price.
Opening Price
All EOD providers [that I know of] use the first full-lot trade on any exchange except for Wealth-Data (and Wealth-Data alone). Wealth-Data gives you the Primary Market Open, which is a no-slippage price that you will get with a Market-On-Open order.
High/Low
Wealth-Data (and Wealth-Data alone) applies a statistical filter for outliers. Some highs may be lower and some lows may be higher than all other sources.
More on Wealth-Data at https://www.wealth-data.com
Persuant to my entries above... below is a 10 Year .PDF comparison of the backtest basic scorecard for the Published Strategy "One Night", using an exact same dataset of the DOW30 in TD Ameritrade, WealthLab, and IQ Feed. I thought this would be a good strategy to comparison in order to isolate any variations in backtest results produced by the data sources themselves. As you can see, the TD Ameritrade and IQ Feed results were very "close". Can anyone explain why the WealthLab results are so different relative to the other two
Also, why do I show any "short" results at all? This strategy does not produce short signals, I thought.
Thanks
Also, why do I show any "short" results at all? This strategy does not produce short signals, I thought.
Thanks
7679-One-Night-a-Strategy-Compare-of-different-Dataset-providers-pdf
They’re all “WealthLab” results are they not?
If you mean “Wealth-Data” then we already explained the differences.
If you mean “Wealth-Data” then we already explained the differences.
Do you think you could kindly offer an explanation for this previous question?
"Also, why do I show any "short" results at all?
This strategy (One Night) does not produce short signals, I thought."
Long and Short results are always displayed in the metrics report. Even if there are no trades it can be handy to see things like interest in cash
Your Response
Post
Edit Post
Login is required