Analyzing trades - how is Return column computed?

[andreas-vester]

I am studying the docstring of vectorbt.portfolio.trades and I've got some questions.

>>> # Entry trades
>>> pf_kwargs = dict(
...     close=pd.Series([1., 2., 3., 4., 5.]),
...     size=pd.Series([1., 1., 1., 1., -4.]),
...     fixed_fees=1.
... )
>>> entry_trades = vbt.Portfolio.from_orders(**pf_kwargs).entry_trades
>>> entry_trades.records_readable
   Trade Id  Column  Size  Entry Timestamp  Avg Entry Price  Entry Fees  \
0         0       0   1.0                0              1.0         1.0
1         1       0   1.0                1              2.0         1.0
2         2       0   1.0                2              3.0         1.0
3         3       0   1.0                3              4.0         1.0

   Exit Timestamp  Avg Exit Price  Exit Fees   PnL  Return Direction  Status  \
0               4             5.0       0.25  2.75  2.7500      Long  Closed
1               4             5.0       0.25  1.75  0.8750      Long  Closed
2               4             5.0       0.25  0.75  0.2500      Long  Closed
3               4             5.0       0.25 -0.25 -0.0625      Long  Closed

   Parent Id
0          0
1          0
2          0
3          0

How do you compute the Return of individual rows? The first return of 2.75 seems to be clear, it should be the percentage gain with respect to initial portfolio value (=100), right? What about the subsequent rows?

Suggestion: Rename the columns to PnL in currency and Return %, if this is really what you intend to present. This will make things a bit more clear.

[polakowo]

Here's the function:

vectorbt/vectorbt/portfolio/nb.py

Line 5123 in f67c502

def get_trade_stats_nb(size: float,

Return is not displayed as a percentage (that it, it isn't multiplied by 100), but I'll consider doing so.

[andreas-vester]

I got it, you're computing Return as ret = pnl / entry_val. In that case Return is displayed as decimal value. I wouldn't change it.

My initital thinking was that Return has been computed with repsect to the portfolio value (i.e. 100 at the beginning of trading).

What about renaming the column into Return trade and inserting another column Return portfolio? In my view, that would be another valuable information. You could compute it as ret_portfolio = pnl / portfolio_value.

[polakowo]

@andreas-vester portfolio and entry/exit trades are separate entities that are loosely connected. Trade records are built exclusively from order records while portfolio is built from a combination of order records, initial capital, and other things. Storing third-party information in records reduces their flexibility. Such metrics should better be displayed by the portfolio object, which has a better view over all statistics. But I think it isn't worth the effort of adding another field when the user can easily compute it by themselves (trades.pnl.values / portfolio_value).

[andreas-vester]

Fair enough ;-)