Hi,
you need anaconda 32-bits to make things work on your 64-bits machine. I have that set-up, and for simple strategies performance are similar to my linux box.

Thanks Florent.
I was hoping there was a 64 bit win version so I don't have to rebuild my python env. Guess I will create a VM and build a new 32bit Python env.

64 bit won't help much with the performance issue on Zipline, zipline was designed with Quantopian as its main focus, so at each step of the simulation it has all the risk, performance, portfolio attributes available for you, even if you dont use them they are computed. For back testing you only want them at the end and don't need to compute on each tick. I have a stripped version of zipline where performance is much better I posted on the zipline group. I am working next on cache of TA and history since if you are optimizing you can compute the TAs and history once and just fetch from memory for the whole optimization. Should bring the performance close to expected. For simple portfolio optimization I use python QSTK. I am even considering making a wrapper for QSTK so I can write zipline like strategies test on QSTK and run on zipline after optimization is done. Which is pretty much what I do now but with a lot manual work involved.

Thanks for pointing this out Lucas. I'll try to replicate your perf improvements.

Hi Lucas,
I've made your suggestion as a parameter to the 'sim_params' variable in Zipline so that we may propagate the changes:

fast_backtest = True  
algo = TradingAlgorithm(initialize=initialize, handle_data=handle_data, capital_base = 10000, fast_backtest=fast_backtest)  

it's here: https://github.com/florentchandelier/zipline/tree/wip-ImproveBacktestSpeed

Speed is increased by a factor of 19 !