No where to change benchmark?

[liyi-1989]

Say, I am using v1.3.0 and trying to use zipline.run_algorithm() to run the algorithm. If I do not change the source code, I cannot find a way to change the benchmark.

1. In the run_algo() source code, it essentially creates a TradingAlgorithm object by inputing an env variable.

2. This env variable, defined in the function body, is created without explicitly giving the bm_symbol argument.

3. Based on the definition of the TradingEnvironment class, the env has the bm_symbol always equals to SPY by default (and load variable always equals to None).

4. In this way, the env.benchmark_return property is thus (always) created by bm_symbol='SPY'.

5. Thus, when TradingAlgorithm uses it (env), it always uses the benchmark_return created by SPY.

(6. btw, adding context.set_benchmark(symbol('XXX')) in the initialize() function is not useful, since it runs after the above environment process.)

So, it looks like the only way to change the benchmark from SPY to something else is to:

• change the code of TradingEnvironment class, the default of bm_symbol
• or, when creating the env variable in run_algorithm() source code, explicitly add the bm_symbol argument. (and probably add a bm_symbol argument for the run_algo function)

(in either way, we have to change the source code)
Am I right?

[marketwiz]

Yes

…

On Fri, Sep 28, 2018, 7:23 PM Yi Li ***@***.***> wrote: Say, I am using v1.3.0 and trying to zipline.run_algo to run the algorithm. If I do not change the source code, I cannot find a way to change the benchmark. 1. In the run_algo() source code <https://github.com/quantopian/zipline/blob/v1.3.0/zipline/utils/run_algo.py#L206>, it essentially creates a TradingAlgorithm object by inputing a env variable. 2. This env variable, defined in the function body <https://github.com/quantopian/zipline/blob/v1.3.0/zipline/utils/run_algo.py#L154-L160>, is created without explicitly giving the bm_symbol argument. 3. Based on the definition of the TradingEnvironment class <https://github.com/quantopian/zipline/blob/v1.3.0/zipline/finance/trading.py#L77>, the env has the bm_symbol always equals to SPY. 4. In this way, the env.benchmark_return property is thus created <https://github.com/quantopian/zipline/blob/v1.3.0/zipline/finance/trading.py#L100-L104> with bm_symbol='SPY'. 5. Thus, when TradingAlgorithm use it <https://github.com/quantopian/zipline/blob/v1.3.0/zipline/algorithm.py#L557>, it uses the benchmark_return created with SPY. So, it looks like the only way to change the benchmark from SPY to something else is to: - change the code of TradingEnvironment class, the default of bm_symbol - or, when creating the env variable in run_algo() source code <https://github.com/quantopian/zipline/blob/v1.3.0/zipline/utils/run_algo.py#L206>, explicitly add the bm_symbol argument. (and probably add a bm_symbol argument for the run_algo function) Am I right? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#2311>, or mute the thread <https://github.com/notifications/unsubscribe-auth/APkkZGMVU01NlltrXBWWY_YTNIcxHwY5ks5uftmZgaJpZM4XAR4e> .

[ssanderson]

Apologies for the slow reply to this issue. On Zipline master, you can now pass benchmark_returns or benchmark_sid to TradingAlgorithm. There's still no good way, however, to control the benchmark when running via the Zipline CLI. I'm tracking that issue in #2627.