In my own trading, I use fixed fractional position sizing - that is, while the monetary value of risk per trade varies as my equity goes up or down, I risk the same amount in percentage terms.
A while back, a good trader friend of mine experimented with varying the percentage risk per trade based on a look back period of performance.
Both methods have been used by successful traders and Market Wizards. Both have strengths and weaknesses.
In my own case, by risking the same in percentage terms on each trade, my monetary risk on each subsequent trade will be reduced if I go on a losing run, and in percentage terms my drawdown will be higher than someone who cuts back on their position size.
Against that, once I start getting back into some winning positions, I should be able to recover from those drawdowns quicker by still risking the same percentage risk per trade, compared to someone who has completely shrunk their position size using a smaller percentage risk per trade.
When my friend implemented this change, he found that, while his performance in terms of R over subsequent months was positive, his actual monetary performance was negative. In this period, he found that his biggest winning trades (expressed in R) were achieved using the smallest percentage position size!
As a very basic example of this, picture a scenario where you suffer 10 consecutive losing trades of -1R using say 2% risk per trade, followed by 2 winning trades of +10R using only 1% risk per trade.
In terms of R, over that small sample you have made a net gain of +10R, but in percentage terms you would be break-even, and in monetary terms with the negative compounding effect you would end up making a loss!
It is this very reason why I have shied away from varying my position size in percentage terms. To me, using a look back period of prior performance is a lagging indicator. And it is another layer of complexity that I want to avoid. I much prefer to keep things simple...
Of course, this suits me whereas other traders may find more benefit in varying their percentage position size.
No less a trader than Paul Tudor Jones had this to say in Market Wizards:
"I will keep cutting my position size down as I have losing trades. When I am trading poorly, I keep reducing my position size. That way, I will be trading my smallest position size when my trading is worst."
A few years back, a systematic trader I know in Europe told me about his experiences using variable position sizing. He had got his hands on a trend following system where the position sizing metrics meant he was trading far too aggressively - the pyramiding element allowed under the rules meant excessive risk was being taken.
Initially the system did fantastically well. He got into some Yen-based foreign exchange trades which developed some nice trends. However, the system then went into a sharp drawdown, losing those profits and more. The results being achieved were far too volatile.
He contacted the creator of the system and was advised to use variable position sizing so that, once he got into a losing period, the position sizes would be shrunk to help protect his capital.
Before implementing this, he decided to backtest this change, and to his horror, found that the system would never recover from those losses. The reduction in position size meant that the subsequent profitable trades weren't big enough to get the system out of the drawdown.
So you can see how what works for one trader and their method may not work for another. There is no universal right way or wrong way to do it.
This is another example of how one element of a trading plan may be completely different between traders, yet either method may have a positive effect of their performance.