Suppose I have backtested my strategy. 800 wins and 800 losses. With each trade having an average r/r of 2:1. However each trade has a different r/r. So one trade might be a r/r of 3:1 and another a r/r of 1:1. Should the kelly still be calculated as (2 (average r/r) * 0.5 (win rate) - 0.5 (loss rate)) / 2 (average r/r) or is that not correct?
Kelly formula works for bets with even outcomes like coin tossing which is typically not the case in trading - as you say. Therefore calculating a Kelly % may not tell you a lot. You can read Ralph Vince's book "The Leverage Space Trading Model" on the topic if you want to dig deeper.
I was waiting for the professional traders to give you their answers. Since none showed up, let me give you an amateur retail's perspective: I accumulated thousands of option trades over the past 6 1/2 years. When I looked at my average win:loss & risk:reward ratios, they were actually quite stable. Each individual trade may have different probabilities but I used the blend win:loss and R:R ratios to calculate my Kelly. The key is to have a positive expectancy or it is moot. The other is to not bet full Kelly. For me, it was actually consistent with what folks said: Not to risk > ~1%-2% of total capital per trade. Good luck.
Separate them into different setup/category. Each has their own Kelly value. 1. Bet each of the setup differently. 2. Pretend u have multiple accounts, growing each of them individually.