Trouble getting a clean historical dataset
Hi everyone,
I have a strategy that generates buy/sell signals on stocks on a 1-day basis. The strategy employs a good deal of indicators with considerable lookback periods (e.g. 75-day SMA etc.).
Now the obstacle that needs to be coped with is: I keep getting major inconsistencies when employing the strategy on various PCs. That is, there are several PCs, each of which should have an accurate historical dataset for the stock prices. These datasets have been obtained via working with this and other strategies for quite a while, adding historical data to the database when needed via Yahoo / Google connections. The NT program settings on the PCs being in use are the same.
What happens is: I do the same analysis on two independent PCs, which both have a "correct" set of historical data. On the one PC the strategy tells me it has a current AAPL position of 60 pieces, whereas on the other PC it thinks it is flat. It might as well be strategy on PC#1 thinks it has 50 YHOO, whilst the PC#2 strategy thinks it has 95 YHOO. Therefore, the signals being generated differ on the PCs being used - pretty incomfortable to think that the signals don't stem from the strategy, but more likely rely on the underlying data...
Can anyone think of a procedure to establish a "clean" dataset, meaning I can do this procedure and then do the analyses on either PC without having to worry about the data possibly being different (I'm not saying wrong) from the other PC's data?
Regards
n.i.n.j.n.a.
I have a strategy that generates buy/sell signals on stocks on a 1-day basis. The strategy employs a good deal of indicators with considerable lookback periods (e.g. 75-day SMA etc.).
Now the obstacle that needs to be coped with is: I keep getting major inconsistencies when employing the strategy on various PCs. That is, there are several PCs, each of which should have an accurate historical dataset for the stock prices. These datasets have been obtained via working with this and other strategies for quite a while, adding historical data to the database when needed via Yahoo / Google connections. The NT program settings on the PCs being in use are the same.
What happens is: I do the same analysis on two independent PCs, which both have a "correct" set of historical data. On the one PC the strategy tells me it has a current AAPL position of 60 pieces, whereas on the other PC it thinks it is flat. It might as well be strategy on PC#1 thinks it has 50 YHOO, whilst the PC#2 strategy thinks it has 95 YHOO. Therefore, the signals being generated differ on the PCs being used - pretty incomfortable to think that the signals don't stem from the strategy, but more likely rely on the underlying data...
Can anyone think of a procedure to establish a "clean" dataset, meaning I can do this procedure and then do the analyses on either PC without having to worry about the data possibly being different (I'm not saying wrong) from the other PC's data?
Regards
n.i.n.j.n.a.
Comment