Retrospective - API Release To Prevent Minimum Bet Abuse - 19th June

[Sansa]

Assuming that an exact 25% is supposed to be valid. Then for £0.10 @ 1.16 maybe their calculation ends up getting 0.015999 due to some rounding down, instead of 0.016.

I wonder what tomorrow's planned maintenance outage is for?

[geoffw123]

Aha thanks for correcting me on this, I just removed my fixes and tested it again with £0.10 @ 1.16. You are correct, it is a 25.00% error and it is rejected. I had naively assumed Betfair would have been consistent on how 25.00% behaves across the tick range.

Betfair if you are reading this, you have a nasty inconsistency in your implementation. Please correct this, or explain to us why it is correct as is. Thanks

P.S Yes Sansa, I was thinking same thing, I guess they have got a small rounding error bug, that they missed in their haste to get this patch out.
Was it a Superman film, where richard pryor made millions exploiting some banks cent rounding loophole ? I think some naughty Betfair user must have been watching that film recently

[WTPooh]

Aha thanks for correcting me on this, I just removed my fixes and tested it again with £0.10 @ 1.16. You are correct, it is a 25.00% error and it is rejected. I had naively assumed Betfair would have been consistent on how 25.00% behaves across the tick range.

Betfair if you are reading this, you have a nasty inconsistency in your implementation. Please correct this, or explain to us why it is correct as is. Thanks

P.S Yes Sansa, I was thinking same thing, I guess they have got a small rounding error bug, that they missed in their haste to get this patch out.
Was it a Superman film, where richard pryor made millions exploiting some banks cent rounding loophole ? I think some naughty Betfair user must have been watching that film recently

There is no inconsistency here. 25% and above is always rejected.
When you lay 0.01 @ 1.8 you are getting 20% lower then expected. Everything above 20% is rejected.

[oldBloke]

There is no inconsistency here.

Lots of inconsistency.

try placing a lay of £0.02 at 1.4 - works

then try placing a back of £0.02 @ 1.4 - rejected

[geoffw123]

This is getting confusing now isnt it ?
I think we can all agree, it would have been better if Betfair had solved this issue in a different way. (Or post more frequently on this thread at least)

WTPooh, have you tested with the ReplaceOrders command? Moving a lay order of £0.01 up from lower odds to 1.8 is quite happily accepted That is a 25% inflation after rounding from 0.008 to 0.01.

Interesting comment I have just read on this issue from the BetAngel forum. (Some BA users commented that their patched version seems slow, which obviously doesnt make sense with the limited nature of this rounding fix)

We know that this emergency fix had to be done at betfair as they were worried about the stability of the whole exchange if they didn't intervene, so perhaps they've adjusted other internal settings to ease the strain on their servers.

[Hefi]

If they have problems with the servers why don't they let the applications directly bet under the minimum stake with the precision of two digits (so the real minimum stake would be 0.01). Placing a bet under the minimum bet is 3 operations which is definitely more "spam" than with only one operation.

[TOM2908]

yes this is absolute rubbish, betfair just round up or down!!!!!!!!!

[Sansa]

You can place Lay £0.10 @ 1.16 via the API, but you cannot do it via the website. Obviously the process is different i.e. API uses size reduction, the web uses size increase. I don't know if that was the case yesterday.

[BetfairDeveloperProgram]

Hi All

Regarding the below (and other similar scenarios)

Lots of inconsistency.

try placing a lay of £0.02 at 1.4 - works

then try placing a back of £0.02 @ 1.4 - rejected

Apologies, we are aware of these issues and ETA for the fix is early next week. We'll confirm once this is in place.

Kind Regards

Neil

[Sansa]

Thanks Neil.

In preparation for the fix can you tell us what should be accepted or rejected.
e.g. Should equal or greater than 25% be rejected (as implied by BDP in the first post), or is it greater than 25%?

[BetfairDeveloperProgram]

Hi Sansa,

Should equal or greater than 25% be rejected (as implied by BDP in the first post), or is it greater than 25%?

Yes, its actually <20% and >25% rather than <=20% and >=25%

KInd Regards

Neil

[TOM2908]

can i just ask why laying 10p at 100 is classed as below minimum, the liability is 9.90, would be the same as backing 9.90 at 1.1 and that is allowed

[Hefi]

Can someone give me some example about when we need to check the 20% and when we need to check the 25% of the rounding? I'm a little bit confused :S

[Sansa]

Here is my take on what should be happening.

0.01 @ 1.80 = 0.0080 Rounded = 0.01. 25.00% More. Accepted (Profit/Loss increased by 25%)
0.01 @ 1.81 = 0.0081 Rounded = 0.01. 23.46% More. Accepted (P/L increased by less than 25%)
0.01 @ 1.79 = 0.0079 Rounded = 0.01. 27.85% More. Rejected (P/L increased by greater than 25%)

0.01 @ 2.24 = 0.0124 Rounded = 0.01. 19.35% Less. Accepted (P/L reduced by less than 20%)
0.01 @ 2.26 = 0.0126 Rounded = 0.01. 20.63% Less. Rejected (P/L reduced by greater than 20%)

If the percentage difference between the Actual profit/loss and the Mathematical profit/loss is an increase of greater than 25%, or a decrease of greater than 20% then the bet will be rejected.

[ozknows]

Hi Neil,

Both your formulae don't work in all cases, if you take some price/size combinations that are exactly on -20% or +25% some are allowed and some are blocked.

Price: 1.01 Size:0.80 Diff: 25% Allowed satisfies >=25%
Price: 1.10 Size:0.25 Diff: -20% Allowed satisfies <=-20%

but

Price: 1.01 Size:1.25 Diff: -20% Blocked satisfies < -20%
Price: 1.05 Size:0.25 Diff: -20% Blocked satisfies < -20%

there are other examples of this.

Can you give us a working formula that works in all cases and/or a list of price/size combination from 1.01 to say 3.0 and whether each size is allowed or blocked (Sizes up to £1.60 should be okay) ? For example £0.01 is allowed at 1.80 but is blocked again between 2.26 and 2.58.

We are trying to apply fixes but have inconsistent info which is often contradictory and has edge cases that doesn't work.

Thanks.