The devil is in the details

Easy:
getting real-time data from a financial exchange

Harder:
ensuring that data is correct

Harder Still:
interacting with the exchange and sending markets/orders cancels etc.

Harder Than That:
maintaining a book with offsetting positions

Really Really Really Hard:
dealing with partial fills for new positions and/or exits

Approaching Impossible:
crossing those positions internally to save a trip and the spread, including partial filling, waiting for confirmation that outstanding markets

Worthy of Arthur:
doing all that, without loops, and a subsytem to minimize paying the spread via fancy bid-> agress conversion logic.

E.g. imagine you are long 5@100, looking for the market to goto 110$ in the next hour. Then you get short 10 @101, expecting the market to goto 99 in the next 30 minutes, in 30 minutes the market is 100- what do you do?
Lets say you cross internally. You take the 5 longs off the book, so you cancel the offers for those 5 longs- but wait- while thats happening the market goes to 107/109.

Now instead of 2 positions imagine 100, and instead of 1 price per position- imagine N distinct prices- and a market that jumps all over the place.

Some useful financial functions

thought these might be handy to a few people. I can't take credit for most:

Drawdown
drawdown:{[x]v:u?max u:(maxs x)-x;(u v;x?x[v]+u v;v)};

This returns a vector, the drawdown value, the index of the start and end of that period.


EMA
ema:{[n;x]b:1-a:2%n+1;c:(sum n#x)%n;((n-1)#0n),c,c{[a;b;x;y](a*y)+b*x}[a;b]\n _x};

Smooth Moving Average
smavg:{[n;x]((n-1)#0n),i,{(z+(x-1)*y)%x}[n]\[i:avg n#x;n _ x:0^x]};

Max Consecutive Losses
MaxConsectiveLosers:(max count each "0"vs raze string pl<0)

Convert Tics to Float and visa versa

f2tic:{[x]a:x - floor x;b:a%(1%32);c:floor(a -(floor b)%32)%(1%256);if[c=4;c:"+"];raze raze string(floor x),"-",string floor b, string floor c};

tic2f:{("I"$n#x)+((8*"I"$x n+1 2)+"0123+567"?x 3+n:x?"-")%256.}