Technology : New laptop

[Tuco]

I'm also considering a systems upgrade, so any help with the following from Peter (or anyone else in the know) would be much apppreciated:

Earlier this year there was an event I was trading in which I ended up having in excess of two thousand matched bets, and as a result the performance of BA was significantly reduced.

When I thereafter had bets matched, I imagine that the CPU had to re-calculate the exact position for each of the 40+ selections re-calculating all of the two thousand previous bets.

What resources does BA most heavily use? CPU? Cache? Memory? Graphics? Does this change depending on the type of trading you are doing, ie pre in-play or in-play?

If you were setting out to build an ultimate Dell Laptop soley for BA Trading, what spec would you look to build? ie would you choose Xeon over i7? 16/32/64Gb RAM? etc... (I already run twin top spec (but a few years old) Dell Laptops each powering 2 x 24" screens and would wish to retain running at least 4 screens.)

All/any advice in this regard would be much appreciated. Thank you in advance Peter

...just wondering if Peter or anyone in the BA development team could consider and respond to this request - thanking you in advance

[spreadbetting]

Two thousand bets !!!! Jeez.

Not surprised BA slows down with that many bets as the lot would get requested from the API on each request, that's a lot of data in JSON format especially if you're streaming. The loss in performance may well be the program having to retreive the data rather than calculate the pnl. There is an API request listMarketProfitAndLoss that provides the pnl figures direct from Betfair rather than having to calculate it yourself. Might be worth amending the program to call that in these type of situations rather than calculate it itself if it's impacting on performance.

[Hobsom]

BA calculations utilise CPU.
Cache, memory, HDD & GPU do not crunch numbers. A GPU may help if there are lots of graphs on multiple screens, but BA isn't exactly rendering in HD
An SSD may be faster than a traditional hard drive, but the speed differences wouldn't really be noticed if you're retrieving your data from t'interweb.
A fast broadband service connected via an ethernet cable, rather than wi-fi, is also a good option for speed and reduced 'data lag'.

The more cores that your CPU has the better. The more versions of BA you run simultaneously, the more CPU cores you'll need. At least that's how it looks on my 3 PC's.
My 8 core laptop runs about 10 to 15 times faster than my 2 core desktop with multiple versions of BA. No difference in RAM, GPU or HDD size or speed.

So, as far as BA is concerned, you should aim to obtain the highest number of cores at the highest processor speed that you can for your money.
The more cores you have, the more processor cache you tend to have, which also helps a little with calculation speed too.
So it's all down to the number of cores and the CPU speed combined as far as running BA is concerned.

Hope that helps.

PS. IMHO, two good price v performance laptops are: -
https://www.ebuyer.com/819842-hp-250-g6 ... kx90es-abu
https://www.ebuyer.com/805723-hp-250-g6 ... sy44es-abu

[LinusP]

Not surprised BA slows down with that many bets as the lot would get requested from the API on each request, that's a lot of data in JSON format especially if you're streaming.

I assume you mean 'not' streaming?

I regularly have close to 1000 orders per market and have no issues from streaming on my programs so I assume it is whatever BA is doing locally, could it be the local pnl calculation function?

[Tuco]

Thanks for the input fellas - I was thinking of a couple of much quicker laptops that those HP's shown hobson but take your advice on board and thank you for that. I'm still using a couple of old (but top spec) dual core Dell's hence the reason to get a couple of new ones which are in the region of £3k each.

Hopefully I'll get something from Peter and/or the development team in this regard soon...