Day Trading & Forex Forum

FTSE Beater

Hi All

This is taken from the Datafeeds thread
Hi Guibert

I may be wrong, but I don't think it would be possible. The data that comes out of the LSE would be traded prices only, which accounts for a lot of "dud" data and as the only data comes from the LSE, there is no other option.
You could program your backtesting package to eliminate the "dud" data, but that wouldn't be easy.

With regards backtesting, we are a long way behind the US IMHO

guibert

Hello, FTSE Beater, and thanks for your reply. I agree with your comment re High and Low 'dud' data where it is recorded on the basis of executed trades rather than market mid-prices. I have no doubt that there are many people using sophisticated modelling packages on such inappropriate data. It's a wonder that Sharescope and other data providers have not cottoned on to the fact that their EOD data does not give a proper market picture for SETS and SETSmm shares. They could provide additional 'mid-prices' if they so chose. I guess that they don't do it because most of their customers who use OHLC are unaware of the problem.

As for my original request for suggestions on somehow obtaining proper 'mid-price' Highs and Lows, I am still hopeful. The LSE actually produces official "End of Day Summary Data" which lists these figures. Problem is that these summaries only go back to April 03 as far as I know.

If all else fails, then I would go down the laborious route of actually extracting the prices from Best Price tick data. However, I haven't managed to find any affordable and extensive source of such data for UK shares. Suggestions very welcome.

theknifemac

Guibert

Can I ask why you want mids rather than actual traded prices for the highs and lows ?

Also tick data normally means the stream of actual trades at which the stock ticked or printed. Some tick data will include the bid and offer at the time of each trade. If you want the history or ever change of best bid / offer then you may find this difficult to acquire.

Stew

guibert

Theknifemac

A very good question.

My observation is that very often maximum and minimum traded prices do not refect the true extent of the market. If the 'High' trade price was a Buy, there is no guarantee that it was at the maximum 'Best' Buy price in the market during the day (there may have been no Buy at that maximum price). The 'High' trade price may not even have been a Buy, it could perversely have been a Sell at a Bid price, depending on how the market was moving and the frequency of trades. Similarly for the 'Low' trade price. This introduces a degree of randomness into the daily market range as you never know how closely the trade based Highs and Lows approximate the real market range. This can affect a number of modelled situations, the simplest being the modelling of a SL setting on an open position. The SL may have been triggered in reality even though the Trade based High or Low indicates otherwise. The same is true if you are monitoring for an entry signal at a certain price threshold. It all depends on what you are modelling.

If this sounds a bit like splitting hairs, then compare the Offer and Bid graphs for a few FTSE 250 shares against their officially recorded Highs and Lows over the past week. You will see that the differences can be very significant.

I should say that 'mid-price' based Highs and Lows are obviously not perfect either, as you still have to accomodate the issue of variable spread. However, they give a truer indication of market range than trade based prices. [Spread vagaries can be accomodated by certain modelling and trading techniques].

When I say 'tick data' I am referring to what is often called 'Best Price' tick data, or 'Level 1' tick data. This contains trades as well as the changes in 'Best Price' Offer/ Bid. The sample I have seen looked like it contained every incremental change in the Offer/ Bid. Perhaps very small changes were filtered out, but I wouldn't be worried about that. It seemed comprehensive enough to me.

Regards

theknifemac

Guibert

I see where you are coming from now, I have been attempting to get round those types of issues by using data from a lower time frame e.g. I am taking signals from 5 min bars and then using the one minute bars to check for execution of orders in the market. While this is not perfect, I think it solves some of the problems you mention. A better solution would be to have the actual tick data - trades and bids / offers but this is difficult / expensive to get hold of.

Of course you might then want to model the impact of your trade on the market - I know people of have tried to do this for institutions and it gets quite hairy !

Best of luck

Stew

jmreeve

guibert-

Everything you need is available including bid/ask data..
How much history do you need and on how many stocks?

Using quote data can also give to an inaccurate picture of price as bid/ask price is prone to extremes if market liquidity is poor but no one actually trades at these extremes.

__________________
Tomorrow anything may happen for no reason.

Rainmaker

Would 10 years' data for FTSE 350 stocks be possible?

Thanks in anticipation.

jmreeve

Possible but very expensive I think.
I believe Olsen and possibly FT Interactive data might have this.

__________________
Tomorrow anything may happen for no reason.