Hacker News Reader

Comment by danielharan | Hacker News Reader

A question for those who are trying to get better: how do you measure progress?

The biggest problem is selling stocks to suckers hasn't changed much in a century or so. Fred Schwed's hilarious classic shows absolutely nothing has changed in that business in 90 years other than minor terminology.

On the other hand, other than general job title and being in front of a keyboard and looking at a screen, almost nothing I did in 1981 is even remotely similar to what I do in 2013.

I started out with basic and Z80 assembly programming, then 6809 assembly programming, then OS-9 (kind of like unix, sorta..). Then downgraded to crude PC and dos generalist stuff in the late 80s to early 90s. I did lots of linux sysadmin work in the early (well, 93ish onward), mid, late 90s, then I took a detour for a couple years into being (mostly) a BGP operator at a regional ISP, then a phase of mostly DBA and data importer/exporter work, then went thru a CRUD web app phase, etc. Along the way I did a lot of microcontroller work mostly 68hc11 and a lot of the ancient and current PIC families. I also had a weird side trip thru the joys of mainframe work circa mid 90s, someday "modern" PC stuff could catch up to BAU in the mainframe world as of 1994, but probably not. And I did some microwave RF work which does actually relate to this discussion when you program a PLL synth using a microcontroller or implement stupid sequencing tricks or even just monitor a GPSDO. I simply don't have 32 years experience in anything. Adding to the weirdness I never really stop anything, I still do enough linux sysadmin stuff to stay current but probably not advance much beyond where I was which admittedly was pretty advanced. And I never stop learning, I still think its hilarious to implement project euler problems in new languages, scala being my amusement of choice right now. I can't count how many Fred Brooks style silver bullets I've seen, dodged, or laughed at in the last three decades.

On the other hand if you were a stockbroker pitching RCA in 1923 to rubes, its not much different to pitch GRPN to rubes in 2013.

How does Algorithmic trading fit into this model?

Your average 40 year career stockbroker knows there's some young nerds from the physics, math, and CS departments who were hired into the mostly B-school grad company, but he doesn't work with them or hang out with them or have anything to do with them other than maybe 3 or 4 levels up the org chart they share a boss. So for an upper level manager HFT is a "new thing" in the biz but not so much a broker. And from an upper level manager perspective they're not totally new so much as basically expensive automated hyperactive daytraders.

Yes I'm sure there exist rare individual exceptions but very few experienced brokers pivoted out of sales and human trading/consulting into algorithm design.

I re-read articles I read and didn't understand and if now I do, I learned something that allow me to better understand a wider range of subjects. If not, I didn't progress.

I retry puzzles that I failed to complete in the past and if they appear easier to do, I progressed. It worked for some Euler Puzzle that I couldn't begin to understand and a few months afterward they seemed trivial.

Redoing anything you struggled to achieve (or didn't achieve) could be a proof of progress, as long as it's actually understood.

Progress I think is context dependent. In competitive activities like sports, the performance statistics speak for your progress. On the other hand programming, being more like a craft, would measure progress by the quality of the work you produce and also the ease with which you do it.

Does your old work look like shit?

Well yes, but so does the current stuff :)

But now you are in the position where you know your work looks bad. Imagine when you were younger and said, "This, this right here is better than anything anyone has ever written."

At least now you have a decent smell test.

Reply on news.ycombinator.com