A couple of years ago I tried to do some simulation of trading strategies with historical data, and had some problems with Yahoo!'s data. Yesterday I updated that data, and suddenly the results of my simulations were entirely different.

After digging for a little bit, I came across some changes in the data for BLT.L. For 4th February 2010, 2nd and 5th April 2010, and 4th May 2010, Yahoo! finance now has the open, low, high and close values of 2301. For some of those dates that's a big outlier. What's strange is that the data I have from Yahoo earlier has believable values for 4th February, which match the values Google provides now. The data was taken at the start of March 2010, so pre-dates the other days which now have strange values.

On a different theme, the opening values in 2003 are quite different to the values I had previously. Here's a brief sample. New values are the first quintuple, old ones are the second quintuple:

There doesn't seem to be any pattern to these changes, but clearly some of them are significant enough to make a big difference to a simulation. I am hoping to get the chance to investigate some other potential sources of price data soon.