Strange difference in live vs delayed prices

[splendid2k]

I'm really confused over the market prices on selections in GB horse racing.

The link below shows 2 sample plots of (different) horse prices in the last 30 minutes before start of market.

All price data recorded using the delayed key is much "noisier" vs the live version and I can't understand why.

http://imgur.com/a/lyHxK

The only difference between the plots is me switching my live and delayed API keys in the scraping script.

Has anyone come across this before? What am I doing wrong!?

[StefanBelo.]

Delayed app key requests actually do not show real data on betfair market, use only live key.

[only1jake]

It would also be because the delayed key has a variable time delay, so the data you are getting will be jumpy. Which is fine for testing, but if you need reliable data use the live key

[KJIB]

Which data is "jumpy", (and by how much)...

It would also be because the delayed key has a variable time delay, so the data you are getting will be jumpy. Which is fine for testing, but if you need reliable data use the live key

Is there information on which data is "jumpy" and by how much?

I want to do some 'bot testing but need to know what data fields make sense to not waste my time.

So far, I it looks to me like these fields from MarketBook have massive "noise" on them time wise (i.e. the results come from a random time from anywhere over "a number of minutes" ago):
* status
* In play(?)
* Volume matched
* £ available
* Last match time

The runner back/lay info also does this, can anyone confirm?

I'd guess that Betfair just have a small-ish number of MarketBook samples from the last few minutes for each market id and you just get a random one of those so all fields in it are unreliable for many realistic 'bot testing scenarios?

I tried to filter out the noise by using the last match time and only updating when it's newer than the previous newest value but that just kills the update frequency, making the data near useless for most things I want to test.

[only1jake]

I'm not sure about the first few you mentioned but ListMarketBook, ie. the back/lay prices are on variable time delay. You basically get what you said, a price from x time ago. So you can get a general idea of prices and trends but not the correct price at the time. So its fine to test basics but if you need to test viability or actual functioning of your bot then you should definitely get the live key. So basically what you said is correct

Also unless your bot has basis on last match time/price, you shouldn't need to use it as last matched price is not necessarily the price that is available.