Problem with back matched

I've had a problem today with back_matched.  Possibly a bug, but not sure.

In the space of 8 refreshes over 12 seconds the program has recorded my back_matched as going up then down then up then down, during a period in which I had no back bets match on the relevant selection.  Surely back_matched should not be reducing?

At the end of those 12 seconds, back_matched was back where it started (and should have been) but the false readings in the meantime caused some triggers to fire when they should not have done.  This happened on at least three separate occasions.

Any help gratefully received.

Hi Orzola,
If back_matched oscillates between 0 and original back_matched it is probably connection failure. In the past, I came up against the similar problém if I had too many matched and unmatched bets and used a short refresh rate. At present I have no experience with MF (and API) performance in such cases but under 'normal' conditions, this should (and in fact does) not occur.

mkas

Thanks mkas.

A lot of what you say corresponds to my betting - certainly I had lots of matched and unmatched bets.  Refresh rate was 1 second I think, but in practice processing speed for my trigger was meaning that it was taking about 1.5 seconds for each refresh.

However, back_matched on the relevant selection was not oscillating between 0 and original back_matched.  It never dropped below my actual back matched, but just recorded varying incorrect, excessive figures until it sorted itself out at the correct figure again.  During that period the trigger was firing bets (incorrectly, based on the incorrect back_matched the program was reading).

Does that still sound like a connectivity issue to you?

Thanks again.

Hi Orzola,

Please attach the relevant logs, i.e. the files where the program recorded the variable's value. It often happens that the logs are misinterpreted, so let me have a look.

Hi Oxa.  I'm afraid I don't have the logs any more (they have auto deleted).  Do you mean that I might have misread the log?  I'm sure that's not the case for two main reasons: 

1. I had the "log trigger expressions" setting enabled, and was setting a user variable "netback" each refresh.  Fairly obviously, netback was back_matched - lay_matched, so the back_matched figure was clearly stated in the log for each refresh.

2. The way my trigger reacted was entirely consistent with the program first misreading back_matched and then reading it correctly a few refreshes later.

Clearly I should have saved the logs.  I do have my notes from my post mortem on the misfiring trigger, and can put those on a spreadsheet and send them through if they will help?  They will show the relevant figures as recorded in the log on each refresh.  But if credibility of the reported issue is the problem, then that's not going to help you much I suspect?

But if credibility of the reported issue is the problem, then that's not going to help you much I suspect?

Oh, no, it's not a credibility issue, I'm sorry you thought it was! I wanted the log files not to make you give me proof, but to look at them and figure out what might be happening in the program.

How many markets were you monitoring and how much data you think you were requesting from BetFair at that time?
Was it a one-off issue or is it still bothering you? If so, can you please check the program logs for any error messages? Specifically the program logs, as opposite to the trigger logs.

Don't worry Oxa.  Worse things have been said to me! 
I was only refreshing one market at the time, but I was trading four selections on that market (a soccer match) and it was approaching kick off and bets were matching and triggers firing at a very frequent rate.  I think I have had this problem before, but this is the first time I have diagnosed it.  It seems to me that it occurs when the program is put under pressure by the frequency of trading.
I wasn't requesting a great deal of data from Betfair.  No more than halfway to the halfway mark shown in the Connection Monitor.
As to whether it is a one off; no, I don't think so, so what happened is certainly still worrying me.  However, since I had this problem, the domestic football season ended, and I have not been betting on as many selections in one market.  So I've not had a repeat, but I've not repeated the conditions which caused this problem, and probably won't do so until the Premier League resumes in August.
As to error messages, there were some exceeded exposure limit errors shown as I recall.  In fact I don't believe that these prevented the relevant bets from being placed.  I think that this is because my trigger was cancelling a bet at one price and placing the same bet at a different stake so not in fact increasing my exposure.  So the error message was, I think, a false positive.
I should probably have examined that aspect more carefully, but once I realised the program was not recording back_matched correctly I just focussed on that fundamental problem.