I'm considering going live with an algorithm but it does a lot of portfolio rebalancing across a small subset of ETFs essentially doing correlation style trading. While it has an impressive compounded annual return one concern is Quantconnect is reporting a very low unrealized gain. What sort of tax lot method is Quantconnect using to estimate unrealized losses? FIFO? 

I have a strong feeling this algorithm could possibly be a lot better in practice if I can model tax lots better. I'd like to see how it works out with LIFO, Highest Cost, and other Lot-Matching Methods Interactive Brokers uses - 

https://www.interactivebrokers.com/en/software/ibto/ibto/lotmatchingmethods.htm

Also, it'd be really useful if we had a way to see what percentage is long term vs short term gains. It'd be great to see how much tax drag it has in the dashboard too. My algorithm is intended to possibly run for a decade or more (backtesting has held up since 1960 at least through all markets), and so ideally I'd like to accumulate a lot of long term capital gains tax lots, while selling off any losses and minimizing short term capital gains. I.E. I want to mimize tax drag as much as possible.

Does IB expose nominating specific lots in their API? If so, it'd be awesome to manage these tax lots algorithmically too! TD Ameritrade has a great Tax Efficient Loss Harvester method - 

https://www.tdameritrade.com/education/taxes/what-is-a-tax-lot.page

 

 

Author