It takes enormous effort to create an algo with positive expectancy and then have infrastructure to run it.
Between our ears is where the problems lays, it's no so much external. Yes, very short term trading is difficult for positive expectancy but the market bias is upward longer term. Infrastructure need not be enormous, I run my profitable system using tablets, and a laptop, no need for additional monitors. The tablets are just so I can lounge around anywhere and keep an eye on markets, just makes for more comfort rather than sit at a desk and a tablet is easier to handle while relaxing than holding a laptop. But if I were forced to, 1 laptop would do the whole job. In everything it pays to downsize rather than complicate. Some months ago I wanted to design a weedless fishing hook. I've spent months thinking about it and finally yesterday created one which uses one hook and a couple of items available at a dollar shop, so simple it could be assembled from scratch while sitting in a boat in about 2 minutes. Point is, with a creative mind anything is possible, think outside the square always.
It depends on what we create and our style of trading but generally speaking, we don't need a lot of complicated infrastructure for automated trading and if we are concerned about power outages we can use a VPS. I run 2 rather large size EA's on 18 MT4 trading accounts (for testing) on an I5 Machine with 8Gb of RAM. The EA's are very complex and each trades 26 pairs so there is a lot of data crunching going on and the machine handles it fine. The only times I hear the CPU fans running hard is sometimes during early NY open.
Interesting. What's your career background which allowed you the skills to build fx EA's, or, did you buy them? You mention 18 trading accounts for testing, what is the theory behind 18 accounts? Regarding risk, do you run overnight or unattended, how do you cope with unexpected glitches like power or computer outages or slow modem response, even heat which Sydney recently suffered from?
The important thing is to have a strategy which has +ve expectancy. The EA building itself can be outsourced. I have an Engineering background but I have not written the code myself, I employed various programmers over a few years. The strategies work well but I need many accounts because I'm testing upscaling (experimenting with trade size and testing performance and draw down on different size accounts). The PC's are Running 24/7 and are mostly unattended, I do monitor the market with the mobile and sometimes I manually close some trades but that's all I do. Trading many pairs allows me to reduce my trade size and therefore reduce my risk. Risk with this strategy is small and even if the accounts were offline for 1 or more days the overall account performance would not change by much. Those 18 accounts are demo and that particular PC is not power backed up but my live accounts are on a different PC powered up by normal 240V mains power and backed up by off grid power. City mains power is usually very clean but I'm scared of electrical storms so for power glitches I use 2 mains filters, 1 located inside the switchboard and 1 under my trading desk. Hot summer days should not cause power outages because a large percentage of Sydney homes have solar panels so electricity demand on hot summer days is not as high as it was a few years ago. On our roof we have a 6.6KW solar system and on hot summer days it generates around 40KW of power. In winter, on sunny days it was producing around 20KW. Slow modem response? I have not seen slow modem response for some years now. Last year we had ADSL2 connection. Now we have a 50Mbit NBN connection and I have not experienced any disconnections but I do have 4G mobile data available as a backup.
I was fortunate to meet by chance a programmer who did my coding years ago, however since, I largely taught myself. It is better when you code your own work, often an algo requires numerous adjustments and addons, so to get the job done it's quicker and more convenient. Any algo I write may take several weeks to months to create and fine tune. When I asked someone else to code for me, I was always conscious of stretching their patience too far with constant requests for alterations. I too have an engineering background, retired now. I live in Perth WA, with NBN, I'm getting an outage sometimes in evening infrequently but I'm not trading evening hours. The speed issue, not much noticeable different re anything I did in the past with ADSL, but International phone calls on VOIP is definitely much improved on NBN.
Yes it can be done if someone is willing to learn. oh nice... I love your time zone. Sydney time sucks for trading the NY session. I wonder why that is, are you inside or outside Perth? I don't know about Perth but some NBN areas are connected to the net via Satellite and this can be problematic. The morons who designed the NBN, decided to use Satellite to link the Australian regions together but they made a major mistake. They chose KA band for their Satellite uplinks and this frequency band suffers a lot from atmospheric conditions and rain fade. So if there is a lot of rain or thick rain clouds over an NBN uplink or downlink site there will be signal degradation and the overall NBN performance for the affected areas will drop and it could even turn to a complete outage.
Basically I was forced into learning how to code. It happened some years ago when I designed a trend following stop loss and entry system which my mentor coded upon request. It had a fault in the code which I asked him to rectify. Anyhow he insisted there was no fault in the code, so in frustration I resorted to spending some hours (days actually) attempting to decipher the Amibroker code. I repaired the fault and the elation was intense, from then on I just decided to teach myself. One advantage I have is I love maths and consequently really enjoy the algo coding challenge.