Defining characteristics of high quality historical data

Hello to you,

This will be a multipart question and greatly appreciate your advice.

1) Regarding historical FOREX data. I hear lots of people saying they are looking for high quality historical data. What does this mean exactly? How would one know the difference between high quality historical data and low quality historical data?

2) I also hear people suggesting that I should use historical tick data to build up the candlestick charts for higher time frames as opposed to just downloading the higher time frame OHLC data. Can someone tell me why this is ?

Thanks in advance for all the responses.

Edb500

1. A lot of broker historical data has gaps with no data or spikes which didn't occur. I can remember taking a screen shot of a live chart and a few days later looking at the historical data and it was unrecognizable.

2. Tick data is the only way to obtain the true spread. Otherwise you have to use the current spread which may not even be close. Plus the playback order of data is accurate. Interpolated ticks are just a guess.

@CodeMeister

1) Thats worrying. I plan on using Dukascopy for historical data which I've heard is quite good. But if I look at old data how am I able to determine if there are spikes there that shouldnt be?

2)Ok so for building historical data why is the spread (difference between bid/ask) so important?

Thanks again

Haven't used Dukascopy myself, but hear that its mostly good quality. Spikes will show up visually or in a very unusual win or loss. Typically I only do a three month backtest, so a visual check is sufficient. Missing data is more common and more obvious.

Spread is very important and subject to wide swings around session openings and news releases. If the spread widens, it can take out your SL and using OHLC data would never show that. If you dont trade those times, its not as important.

Anyone has tried to export from IC market MT4 platform? how's the quality? I tried compare with Dukascopy data, it's quite different.