A "day of the week" can be useful, if you have enough data backing it up, and you actually test the idea from the start. What many people do, though, is to run a system for a while, and after a couple of months say "Fridays I always win. Maybe I'll try double size on Fridays." With that approach, they are optimizing for day of the week after the fact, and probably without enough data. So, in that case I would not recommend it.
In using "day of week" approaches, I've started with the idea at the beginning of development. For example, I might say "every Friday, if the trend is up, I will look to buy Crude Oil near the close, and sell on Monday morning. The theory is that any natural disasters, terrorist attacks, wars/conflicts that start on the weekend might send crude much higher, especially when the general trend is up." My point is I try to build in the day of week idea before I actually test.
For position sizing purposes, I personally would not use winning percentage as the criteria. I'd use expectancy, or maybe largest loss, instead. The reason is that a 70% win system probably has small wins, and bigger losses. If you trade bigger size, and get hit with a loss, it will hurt a lot more.
But, every system is different, and it may be worth running some random simulations with different ideas, and see if one stands out.
I hope this answers your questions, if not please reply back.
If you have any questions please send me a Private Message or use the futures.io "Ask Me Anything" thread